Essays on the Philosophy of Tyler Burge
edited by Martin Hahn and Bjørn Ramberg
Reflection...
22 downloads
1150 Views
2MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
Essays on the Philosophy of Tyler Burge
edited by Martin Hahn and Bjørn Ramberg
Reflections and Replies
Reflections and Replies Essays on the Philosophy of Tyler Burge
edited by Martin Hahn and Bjørn Ramberg
A Bradford Book The MIT Press Cambridge, Massachusetts London, England
© 2003 Massachusetts Institute of Technology All rights reserved. No part of this book may be reproduced in any form by any electronic or mechanical means (including photocopying, recording, or information storage and retrieval) without permission in writing from the publisher. This book was set in Times Roman by SNP Best-set Typesetter Ltd., Hong Kong, and was printed and bound in the United States of America. Library of Congress Cataloging-in-Publication Data Reflections and replies : essays on the philosophy of Tyler Burge / edited by Martin Hahn and Bjørn Ramberg. p. cm. Includes bibliographical references and index. “A Bradford book.” ISBN 0-262-08315-9 (hc. : alk. paper)—ISBN 0-262-58222-8 (pbk. : alk. paper) 1. Burge, Tyler. I. Burge, Tyler. II. Hahn, Martin, 1954– III. Ramberg, Bjørn T. B945.B7684 R44 191—dc21
2003 2002075361
10 9 8 7 6 5 4 3 2 1
Contents
Acknowledgments Contributors Introduction
vii ix xi
1
Burge, Descartes, and Us Calvin Normore
1
2
Anti-Individualism and Scepticism Barry Stroud
15
3
When Swampmen Get Arthritis: “Externalism” in Burge and Davidson Martin Hahn
29
4
Burge Thought Experiments Keith Donnellan
59
5
Anti-Individualism, Indexicality, and Character Joseph Owens
77
6
Competence with Demonstratives James Higginbotham
101
7
Implicit Conceptions, Understanding, and Rationality Christopher Peacocke
117
8
Burge on Mentalistic Explanations, or Why I Am Still Epiphobic Fred Dretske
153
9
Mental Paint Ned Block
165
10
Mental Content and Hot Self-Knowledge Bernard W. Kobes
201
11
Phenomenal Intentionality as the Basis of Mental Content Brian Loar
229
12
Internalist Explorations Noam Chomsky
259
TYLER BURGE REPLIES
289
13
Descartes, Bare Concepts, and Anti-Individualism: Reply to Normore
291
14
Some Reflections on Scepticism: Reply to Stroud
335
vi
Contents
15
Davidson and Forms of Anti-Individualism: Reply to Hahn
347
16
The Thought Experiments: Reply to Donnellan
363
17
The Indexical Strategy: Reply to Owens
371
18
Tracking Perspectives: Reply to Higginbotham
377
19
Concepts, Conceptions, Reflective Understanding: Reply to Peacocke
383
20
Epiphenomenalism: Reply to Dretske
397
21
Qualia and Intentional Content: Reply to Block
405
22
Mental Agency in Authoritative Self-Knowledge: Reply to Kobes
417
23
Phenomenality and Reference: Reply to Loar
435
24
Psychology and the Environment: Reply to Chomsky
451
Tyler Burge Bibliography Secondary Sources on Tyler Burge Index
471 475 495
Acknowledgments
We would like to express our gratitude to all the contributors for their generosity and, especially, their patience as the delays in getting this volume to the press mounted. The greatest thanks by far must go to Tyler Burge who has participated in almost every aspect of preparing the volume, right from the initial conception, and has contributed to it so generously. We would also like to thank the Social Sciences and Humanities Reasearch Council of Canada for supporting the conference in Vancouver, some years ago, that was the initial impetus for this book, and to Simon Fraser University for providing a Publications Grant to help with the final stages of the manuscript. Special thanks go to Robert Sinclair for his help in compiling the bibliography and to Jodi Lough for her work in preparing the manuscript.
Contributors
Ned Block: Professor of Philosophy and Psychology, New York University Tyler Burge: Professor of Philosophy, University of California, Los Angeles Noam Chomsky: Institute Professor, Professor of Linguistics, Massachusetts Institute of Technology Keith Donnellan: Professor Emeritus of Philosophy, University of California, Los Angeles Fred Dretske: Senior Research Scholar in Philosophy, Duke University; Bella and Eloise Mabury Knapp Professor in Philosophy, Emeritus, Stanford University; Professor of Philosophy, Emeritus, University of Wisconsin–Madison Martin Hahn: Associate Professor of Philosophy, Simon Fraser University James Higginbotham: Professor of Philosophy and Linguistics, University of Southern California Bernard W. Kobes: Associate Professor of Philosophy, Arizona State University Brian Loar: Professor of Philosophy, Rutgers University Calvin Normore: Professor of Philosophy, University of California, Los Angeles Joseph Owens; Professor of Philosophy, University of Minnesota Christopher Peacocke: Professor of Philosophy, New York University Bjørn Ramberg: Professor of Philosophy, University of Oslo Barry Stroud: Mills Professor of Metaphysics and Epistemology, University of California, Berkeley
Introduction
The volume you hold in your hands is a festschrift devoted to the work of Tyler Burge. As is the custom with such volumes, we asked various philosophers to contribute papers that either comment on Burge’s work or, best of all, relate it to their own. The caliber of these contributions speaks for itself. As is also the custom, we asked the focal figure to respond to these papers. What is unusual about this volume is the generosity—both in extent and in content—with which Tyler Burge approached the task at hand. The section of Burge’s responses thus comprises the most sustained, detailed, and interconnected body of writings Burge has published in one place. Those of us who have had questions about both the fine points of Burge’s philosophy and the overall lay of the land will find many answers here, expressed in the congenial setting of a philosophical conversation. The volume has been long in gestation. Some of the papers date back to 1995, most of Burge’s replies to 1999. We trust you will agree with us that the book was worth waiting for. In this introduction we offer a brief sketch of what we take to be Burge’s most important contributions to philosophy, followed by an overview of the contents of the volume. But perhaps it would not be inappropriate to start on a more personal note and relate two small details of life at UCLA in the late ’70s and early ’80s. In my (MH’s) first year as graduate student there I took a seminar from someone named “Tyler Burge,” of whom I had heard vaguely, regarding an article that had just gone to press entitled “Individualism and the Mental.” Being an ambitious young lad, I wrote my term paper on the central article itself instead of choosing a less tendentious topic. I analyzed and criticized as only a first-year graduate student can, and I came to the conclusion that, while it seemed an important insight that we can think with concepts we do not fully understand, the further conclusion that mental states are not locally individuated would be difficult for the philosophical community to accept. Considering all this, Tyler Burge wrote very gracious comments on that embarrassing piece of work, as is his wont. But by that last remark, his laconic comment was “We’ll see.” And see we did. The other small detail is quicker to tell as well as more revealing of Burge’s philosophical temperament. Graduate students like little vignettes that put their much-admired teachers in a nutshell. The one we had for Tyler Burge was that when asked a particularly revealing and pivotal question, he would think, hesitate and then start with “Well, this is a minority view, but . . . ,” whereupon an extremely detailed and subtle answer would follow. Indeed, Burge has defended a whole series of “minority views” in his career, some of which have become part of the accepted mainstream of philosophy. It is not that he regards the challenge of orthodoxy as an end in itself. On the contrary, he has stressed both in his writings and elsewhere the importance of paying close attention to common explanatory practices and of plausibility as a criterion in evaluating philosophical proposals. But like other great thinkers, Burge sees that his first responsibility is to get things right, to think through problems without undue regard for what the accepted answers have
xii
Introduction
been. The thinking and the hesitation in that vignette are, in fact, more important than the spoken words. If there is a single characteristic that marks Burge’s work, it is not the determination to go against the mainstream, but subtlety. The minority views are a happy by-product. Tyler Burge’s work can be divided roughly into four areas. The present volume is largely concerned with only one of these, albeit the most important both in volume and in influence. Burge’s early articles—on mass terms, proper names, demonstrative constructions, aggregates, etc. (see the bibliography of his works at the end of this volume)—are all concerned with the philosophy of language, or more precisely, with formal truth theories and the proper rendering of the various terms in question. The papers constitute an important contribution to the project of extensionalist semantics for natural languages. In this volume, only Higginbotham’s contribution is concerned with this early body of work. And the present volume is entirely silent concerning the second area of Burge’s research, although his work there has been widely reprinted and commented on elsewhere. This is the area of the semantic and epistemic paradoxes. Of much greater relevance to this volume’s main themes, although not directly addressed except briefly in Peacocke’s contribution, is Burge’s extensive work on the philosophy of Gottlob Frege. His most widely read article in this area is probably “Sinning against Frege,” in which Burge challenges the common view of Frege as primarily a philosopher of language. But the whole body of work presents an interconnected and subtle interpretation (and, to a large extent, defense) of Frege’s philosophy of mathematics and more broadly his rationalist epistemology and metaphysics of truth, meaning, mental content, a prioricity, and other related concepts. It is Frege’s work, and Burge’s own interpretative work on it, that inspires the robust notion of intentional content and that of a concept (Fregean Sinn rather than Begriff in this context) underpinning so much of Burge’s seminal work in the philosophy of mind. Moreover, it is Frege’s sophisticated rationalism that inspires some of Burge’s recent epistemology. The bulk of the present volume, however, as well as of the commentary in the philosophical literature at large, concerns Burge’s work centered on his anti-individualism or, as it is often called elsewhere, externalism about mental content. The seminal paper, “Individualism and the Mental,” established Burge as one of the central figures defending “externalist” views—often mentioned in one breath with Saul Kripke and Hilary Putnam, especially in the early literature on the topic. Thus, on a superficial level, Burge was first identified with the many “externalist” foes of autonomous psychology, or of what Putnam called “methodological solipsism.” It is true that “Individualism and the Mental” defends what was then the minority view that functionalism, and any other physicalist theory according to which an individual’s mental states supervene on her physical states, must be mistaken. The protracted debate between Fodor and Burge and the sheer bulk of other
Introduction
xiii
literature devoted to the topic bespeaks of this view’s historical importance. But much of the early reactions to Burge’s anti-individualism failed to make distinctions he had already either made or hinted at in the original paper, and as the view develops in later work, it becomes clear that Burge is no mere provider of intuition pumps designed to undermine established physicalist assumptions. Burge’s aim is nothing less than the solution to some of the most trenchant problems in metaphysics and, especially, epistemology: the nature of objectivity and norms, the force of sceptical arguments, aprioricity, the nature of the self, and so on. Central to all these is an account of intentional content and its component concepts. The main argument of “Individualism and the Mental” centers around a variation on Putnam’s famous Twin Earth experiment (1975). Putnam was concerned to show that what was “in the head” of a competent speaker of a language was insufficient to determine the extension of certain terms, ones referring to natural kinds. Thus a speaker who is fully competent with the term ‘water’ may be psychologically indistinguishable from her twin on Twin Earth who is competent with the Twin English term ‘water’ while the two terms differ in extension. The natural conclusion that the terms thus also differ in meaning is tempered in Putnam by his insistence that natural kind terms are in fact indexical. If that were the case, then, while it may be that the twins’ psychological states do not determine the extension of the terms they use, the terms do not necessarily differ in linguistic meaning. The point that indexicals (and other “directly referential” terms) determine an extension only given a context had, at the time, already been well established in the work of Kaplan, Kripke, Donnellan, and others, and Putnam’s externalism has often been considered to be an extension of this point about direct reference. Stephen Stich’s early work, criticizing autonomous psychology, makes explicit the alleged externalist consequences of understanding the semantics of indexicals in this way. Burge’s modification of Putnam’s thought experiment is extensive and subtle. A patient goes to the doctor complaining of arthritis in his thigh. The doctor corrects the patient, informing him that arthritis is a disease of the joints and so, whatever the correct diagnosis, arthritis is ruled out. On Twin Earth, the Twin English term ‘arthritis’ has in its extension a larger range of skeletal ailments, including ones that can occur in the thigh. The twin patient—identical to the earthly patient in every nonintentional way—thus does not mean arthritis by his term ‘arthritis’. He does not need to be corrected by his doctor. And, most important, he has never had a thought involving the concept of arthritis. Where would he get such a concept, after all? The most immediate difference between the Putnam and Burge thought experiments is that, whereas for Putnam the twins are in identical psychological states (which fail to determine the extension of the terms), Burge’s conclusion is that they employ different concepts and so have different intentional contents, and therefore are in different psychological
xiv
Introduction
states. Burge’s anti-individualism concerns thought, not linguistic meaning. Furthermore, it would seem that, whereas in Putnam’s work meanings vary between the two Earths because of facts concerning the speakers’ physical environments, in the Burge case the environments are the same while the social facts of linguistic convention vary. It soon became clear, however, that Burge’s views were both more subtle and more “minority” than this initial analysis suggests. In “Other Bodies,” Burge presents a version of the thought experiment that is closer to Putnam’s in that it uses variation in the environments to establish the crucial difference in twins’ psychological states. At the same time, however, he makes it clear that he means to be making a point about de dicto and not de re beliefs, and he explicitly rejects Putnam’s suggestion that indexicality can be used to account for the Twin Earth phenomena. There is thus no doubt that Burge conceives of his enterprise as something quite different from an extension of Kripke’s (and Kaplan’s) “externalist” insights. But the issues of indexicality and of the relationship of Burge’s anti-individualism to his own earlier work in “Belief De Re” are complex and interesting ones, as can be seen in the extended discussions in the contributions by Normore, Donnellan, Owens, and Hahn, and in Burge’s responses to these. Subsequent variations on the Twin Earth thought experiment and Burge’s extensive discussion of them make it clear that there are further deep differences between Burge’s pervasive anti-individualism and the externalist views he is often associated with. Burgean Twin Earth arguments can be constructed not just for natural kind terms, but also for terms like ‘sofa’ or ‘contract’ and, indeed, for almost any empirically applicable concept. Furthermore, it is not a condition of the experiment that the twins lack any kind of expert knowledge; it suffices that the best available theories of the nature of the referent happen to apply on Earth but not Twin Earth or vice versa. In fact, no single factor is necessary because “[t]he slack between an individual’s perspectives, physical processes, and individual behavior, on one hand, and the objective environment beyond the individual, on the other, takes many forms.” Anti-individualism is a reflection of this fact and of the way our empirical concepts are grounded in aspects of the objective environment. Other arguments can, according to Burge, be used to extend the central lesson of these thought experiments—that we have thoughts composed of concepts that the thinker understands incompletely—to categorical notions like justice as well as to mathematical notions. What Burge proposes, then, is nothing less than a redrawing of the notion of a concept as it has come to recent philosophy from the empiricist tradition. The story of what it is to have a concept, to understand it, and to explicate it, has slowly been emerging in Burge’s more recent work, starting with “Intellectual Norms and Foundations of Mind” in 1986. Peacocke’s and Normore’s contributions and Burge’s responses carry on this central stream of Burgean inquiry.
Introduction
xv
The debate concerning anti-individualism has also had the salutary effect of reintroducing several central philosophical topics into the forum, reformulating the key issues and stimulating new theories. One such issue is that of self-knowledge. If the individuation of a person’s thoughts depends on factors external to her and thus not accessible to introspection, what happens to first-person authority? Burge defends the compatibility of a version of first-person authority with anti-individualism in “Individualism and SelfKnowledge.” Bernard Kobes gives a different account of the compatibility in the present volume. Another epistemic issue connected with anti-individualism is the response to scepticism. Putnam and Davidson, in very different ways, have claimed that a more or less direct argument against the sceptic can be constructed from “externalist” premises. Burge has always claimed that, although there is a connection, the issue is more complicated than it seems. There is some discussion of this in Normore and Hahn, while Stroud directly addresses the relevance of anti-individualism to scepticism. Two further issues arose out of the debate between the defenders of individualistic psychology and those who accept Burge’s anti-individualism. Defenders of functionalism and other versions of theories that assume the local supervenience of the mental on the physical have claimed that only on such an account can one make psychological notions scientifically respectable. The reason is that psychological states have to play causal roles—both in relation to other psychological states and to physical ones. But causation has to be explained in terms of local, internal events—most plausibly the firing of neurons. Thus psychological states have to be internal or local states. Burge has responded, in a nutshell, that although causation is indeed local, this is fully compatible with the distal individuation of the relevant states or events. Dretske agrees with Burge about this but thinks more needs to be said about how externalistically individuated content can play local causal roles. The other crucial part of the response to Burge’s arguments has been a series of attempts to define a notion of narrow or “internal” content. The strategy is to accept Burge’s arguments as far as they go, but show that there is another, robust notion of intentional content that is immune to such arguments. Brian Loar has made important contributions to this stream of debate, the latest of which is offered in the present volume. One of Burge’s great contributions to contemporary philosophy is, in fact, methodological. Burge has argued that there is a great difference between having mastery of a concept and being able to explicate it. He has also urged that we can discover things about our concepts by paying close attention to examples and to our own linguistic practices. These are theoretical views, of course, not methodological ones. But they are consonant with a methodology Burge repeatedly defends in this volume and elsewhere. The Twin Earth thought experiments are just that. They are not deductive arguments from a set of principles. Any principles a philosopher might produce must yield to what we say in the face of examples; our practices and responses are more reliable than our attempts to
xvi
Introduction
capture them with definitions and theories. That is why, in the end, plausibility is such an important philosophical criterion. As we have already suggested, a pervasive feature of Burge’s philosophical argumentation is a deep respect for the patterns and norms embodied in existing explanatory practices. Of course, intellectual practices are not simply given; philosophical questions are raised, decided, or suppressed already in their characterization. Nevertheless, this attitude toward the relation between metaphysical questions and intellectual practice broadly conceived does provide constraints on philosophical options. Philosophical theory is for Burge itself a form of explanation; it aims to give systematic accounts of essential features of our intellectual practices. In several places, accordingly, Burge distances himself from popular contemporary metaphysical positions and preoccupations on the grounds that they find no support in ordinary explanatory practice. The critical potential of philosophical theory draws on the ability to discern the rationality of ordinary intellectual practice. A similar dialectic is present in Burge’s conception of the relation between systematic philosophical theorizing and the examination of the work of the great predecessors of contemporary philosophical discussion. Burge expresses a fundamental aspect of his own contribution to philosophy when he states, as he introduces his response to Calvin Normore, that an aim of his essay is “to provide an example of how history of philosophy and contemporary philosophy can combine for a discussion that yields rewards that neither can easily produce alone.” In fact, Normore’s paper, “Burge, Descartes, and Us,” which opens this volume, itself illustrates exactly the dialogical approach to the history of philosophy that Burge champions. Normore’s subtle reading of Descartes’s account of representation provides a corrective to popular current views of Descartes precisely by bringing out a motivation, which, if Normore is correct, bears immediately on a central challenge to current theories of perceptual representation. Embarking from the reading of Descartes that Burge offers in “Cartesian Error and the Objectivity of Perception,” a reading suggesting that confusion about the modal scope of claims to first-person authority with respect to our thoughts misleadingly provides an individualistic drift to the thought experiments of the Meditations, Normore then briefly considers the externalism of Kripke. His concern with this juxtaposition is to articulate the “tension between a view that would expand the direct reference picture centred on demonstratives, proper names and kind terms to perception and a view that insists that perceptual states specify objects as.” Normore observes, “One central issue in contemporary philosophy of mind and of language has been focused precisely here—how to preserve the insight that the reference of a thinker’s thoughts and words is wholly fixed by relations, usually causal relations, to a world outside the thinker (and so dependent on that world) and at the same time preserve the strong intuition that
Introduction
xvii
some concepts and most, perhaps all, perceptual states could misrepresent the very things they represent.” In the second part of his paper, Normore reads the distinction between the formal reality and the objective reality of an idea (Meditations 3) in light of this question—indeed, as driven by it. Descartes emerges as a robust externalist, albeit one differing in important respects from contemporary varieties. In the third and final part—drawing surprisingly on recently unearthed evidence—Normore offers an interpretation of the relation between the thought experiments of Descartes and Burge, which culminates in a substantive challenge to Burge. Descartes, Normore suggests, relies on the distinction between what Normore dubs the object of an idea and its target to avert a particular threat, to which Normore’s Descartes is instructively alive. This is the threat, in Normore’s words, of “a slippery slope between the view that some of our basic precepts are veridical a priori and the view that we cannot be mistaken in any of our perceptual representations. Somewhere along this slope,” Normore continues, “Burge will dig in his heels, and it would be very helpful to know where and why.” In his comprehensive and wide-ranging response to this essay, Burge not only clarifies his understanding of Descartes’s views and of the relation between these and his own. He provides a carefully constructed account of his view of the nature of thought, revisiting in the process many of the metaphysical themes of his work to date. The basic commitments underlying Burge’s rejection of individualistic accounts of content emerge unambiguously: “It is objectivity and lack of omniscience, not natural kinds or community, that lie at the root of anti-individualism.” Cartesian themes are pursued also in Barry Stroud’s contribution, “Anti-Individualism and Scepticism.” Like Normore, Stroud takes off from the arguments of “Cartesian Error and the Objectivity of Perception.” “Our perceptions,” Stroud takes it, “are in large part perceptions of things as they are.” Still, we must ask about the status of this conclusion, and we may wonder whether it provides sufficient reassurance against scepticism about perception. And if we find that anti-individualist views of perception offer some degree of reassurance against the sceptic, it is a further question, as Stroud emphasizes, whether similar reassurance can be had with regard to belief and knowledge in general. Burge’s handling of this last issue is the focal point of Stroud’s inquiry. For Burge holds that “there is no easy argument against scepticism from anti-individualism and authoritative self-knowledge” (Burge 1988, p. 655, quoted by Stroud in this vol.). “What, then,” wonders Stroud, “are we to make of Burge’s own reassuring-sounding conclusion about perception—that we are ‘nearly immune from error’ in that case?” The question is what would constitute an adequate reply to the sceptical challenge. Burge, suggests Stroud, does not think transcendental reassurance about our perceptions or beliefs generally constitutes an adequate reply. Rather, “it appears that Burge would
xviii
Introduction
like to prove, of some of the particular things that philosophical scepticism would say we don’t know, that we do or can know them after all.” Can anti-individualism be deployed to this effect? Certainly, Stroud observes in agreement with Burge, “that would be no easy argument.” But there is a prior question. “Do we,” Stroud asks, “have to prove such a thing, or in any other way prove that scepticism is false, in order to oppose it on anti-individualist grounds? Is the general reassurance provided by anti-individualism not enough to block any serious sceptical threat?” This sets the task for the remainder of Stroud’s paper, in which he subtly considers the question in light of Davidson’s famous claim that belief is in its nature veridical. We must admit, Stroud concludes, that it remains, in general, possible for any body of belief to be largely or massively in error. However, anti-individualism ensures that it is not possible for a believer consistently to find that possibility to be actual with regard to any particular set of attributed beliefs. Responding to Stroud, Burge returns to the question of what degree of reassurance against scepticism anti-individualism about perception offers. At least this: “it does seem impossible that no perceptual representations as of the physical world ever are or have been veridical representations. This impossibility is itself a necessary truth.” The latter point is essential. “The antisceptical interest of anti-individualism stems partly from the presumption that it might be known through apriori reflection.” It is critical for Burge to make it clear that anti-individualism is taken not as a view about our practices of interpretation or belief attribution, but as a view of the nature of belief—that it aims for necessary truths about belief. Accordingly, for Burge, antisceptical responses should not be confined to arguments for conclusions regarding what we can or cannot consistently believe. Rather we should aim “to show in non-question-begging ways that it is reasonable to believe what the sceptic doubts. . . .” This is a matter of determining what are “the legitimate starting-points for reason,” the basis for considerations that will not beg the question being raised. For Burge, the heart of a response to the sceptic must consist in (carefully distinguished) necessary truths about conditions for the having of (not the attribution of) certain sorts of content, about reason, and about the nature of epistemic warrant. Such a response, in light of the threat of question-begging, will substantially be a matter of apriori reasoning. But even a non-question-begging apriori refutation of the sceptic along such lines need not constitute a proof that the sceptic is wrong. Stroud’s juxtaposition of Burge’s views on scepticism with Davidson’s is hardly coincidental. Both strongly and systematically oppose individualistic theories of mental (and linguistic) content, and for that reason appear to stand together on one side of what many consider one of the key divides in contemporary philosophy of mind. Yet, the differences in their respective conceptions of what constitutes an acceptable—even a possible— answer to the sceptic’s challenge, suggests that the relation between the views of these
Introduction
xix
two philosophers is complex. Although both are often cited as champions of an alleged family of views of content often referred to as “externalism,” their common opposition to internalist conceptions turns out to be grounded in very different philosophical priorities and motivations. This, at least, is the diagnosis offered by Martin Hahn, in “When Swampmen Get Arthritis.” Hahn takes issue with a common view of Davidson and Burge, which has it that they are in accord with regard to what is often called “physical” externalism, whereas they disagree on the “social” variety. This view, if Hahn is right, gets the matter nearly exactly backward. With regard to the role of the physical environment in the individuation of mental content, agreement prevails between the two only so long as the externalist thesis is put in a very general way. Moreover, the arguments from physical or perceptual externalism to the social variety take a very different form in the two cases: “each depends on premises the other takes special care to deny: Davidson’s on his view that meaning and content are constituted by the process of radical interpretation; Burge’s on his view that we can stand corrected in our own idiolectical usage.” As Hahn sees it, to the limited extent that Burge and Davidson see eye to eye on the significance of the social environment for content individuation, this convergence is explained by the very difference in the respective arguments they mount from only superficially similar physical externalist premises. Hahn, distinguishing several anti-individualist theses, suggests that the critical factor in the deep differences between Burge and Davidson is the very different understanding they have of how concepts are individuated. The result is that both disagreements and agreements look different from those often supposed to obtain between the two. In his reply, Burge picks up on Hahn’s suggestion that Burge’s early treatment of de re attitudes is relevant for an understanding of anti-individualism. He begins by carefully sorting out his view of the relation between claims about the dependence of de dicto on de re attitudes, on the one hand, and, on the other, claims advanced by later arguments about the nonindividualistic constitution of the conceptual elements of propositional attitudes, whether de re or de dicto. Turning to the relation between his own views and Davidson’s, Burge argues that they differ pretty fundamentally not only with regard to perceptual externalism, but also concerning the role of the social in the individuation of mental contents. Though both advance a priori arguments for the constitutive dependence of (certain kinds) of thoughts on the presence of linguistic practices, the scope of their respective claims is importantly different. Unlike Davidson, Burge, as he points out, “does not take social relations to be constitutively necessary for having language or thought. . . . The arguments are not as global as Davidson’s.” Whereas Davidson takes his arguments to reveal constitutive features of language and thought, Burge’s arguments are designed to take explicit heed of certain contingent facts about our natures as thinkers and speakers. As far as Burge is concerned, he has given us
xx
Introduction
no reason to doubt that for some thinking being, though not for us, there could be thought without language and there could be individualistically constituted mental content. Where Davidson and Burge advance very different sorts of considerations for the dependence of thought on factors outside the thinker, Hilary Putnam and Burge have designed at least apparently quite similar thought experiments in the service of closely related arguments. The relation between the thought experiments in Putnam’s “The Meaning of ‘Meaning’ ” and Burge’s “Individualism and the Mental” is Keith Donnellan’s topic in his contribution, “There Is a Word for That Kind of Thing: An Investigation of Two Thought Experiments.” Donnellan believes they are importantly different, both in terms of the premises invoked, and in terms of the conclusions they support. “They may,” says Donnellan, “to be sure, have some of the same consequences for both the classical views of the mind and for psychology, but if so the routes taken are quite separate and their points of engagement with the classical view are also different.” In particular, proposes Donnellan, Burge’s experiment may leave available to the Lockean internalist certain defensive strategies that Twin Earth closes down. Putnam’s proposal invokes, according to Donnellan, a “revolutionary idea,” namely “that of a semantic rule that employs paradigms and their underlying nature, a nature that may not even be known to any users of the term.” By contrast, as Donnellan reads Burge’s argument, the latter involves not any semantic principle, “but rather principles about the attribution of beliefs.” And with respect to these, Donnellan’s intuition, at least, allows doubt whether Burge’s two duplicates with identically afflicted thighs really are entertaining different beliefs when they invoke the different ‘arthritis’-expressed concepts that their respective communities employ. Burge is not convinced by this suggestion, as his reply makes clear, insofar as his arguments are designed to make implausible exactly the sort of metalinguistic attribution that Donnellan suggests. Responding to the challenge in Donnellan’s comparative analysis, Burge deals in some detail with the relation between the two thought experiments. An important and often overlooked point concerns the fundamental argumentative strategy Burge employs to establish his anti-individualist conclusions. Donnellan holds that Burge’s argument rests on three premises: the possibility of incomplete mastery of concepts; that mastery resides in the usage of the speech community to which a speaker belongs; and, finally, that the different standards set by different speech communities imply that duplicates in separate communities will have incomplete mastery of different concepts. Burge does not challenge these premises. However, as he stresses, he also does not argue from them. As he puts it, “the thought experiments invite judgments about particular cases. They do not rest on general premises of the sort Donnellan articulates.” This point about argumentative strategy matters a great deal to Burge. It is supposed to ensure that his conclusions are not tied to any particular theory of the nature of thoughts or concepts. Indeed, it
Introduction
xxi
reflects a fundamental Burgean conviction about what we might call the direction of flow of philosophical credibility. In “Anti-Individualism, Indexicality, and Character,” Joseph Owens pursues a theme that arises also in the discussion between Donnellan and Burge; that is, whether an appeal to an alleged indexical element in the mental state contents of Burge’s twins can undercut the anti-individualist conclusions of his arguments. The temptation to try this strategy is not hard to understand, given, as Owens emphasizes, the dramatic epistemological and metaphysical consequences of the view of intentional mental states that Burge advocates. Nevertheless, the persistence of the approach requires an explanation, in light of the fact that the indexical interpretation of Putnam’s Twin Earth example on the face of it is extremely implausible and indeed has been abandoned by its original proponent, Putnam himself. Owens’s proposal is that the “approach remains tempting because of one particular theory of indexicals, that of David Kaplan.” The operative notion is Kaplan’s concept of character, which allows cognitive equivalence of thoughts even where they involve attitudes to different propositions. It is not hard to understand, therefore, the temptation to employ Kaplan’s scheme in order, in Owens’s words, “to defang the Twin Earth thought experiments.” In the second half of his essay, Owens sets out to show that the many philosophers who have yielded to it have been seriously misled. The essential point of Owens’s carefully constructed argument is that the notion of character itself cannot be individuated individualistically. In broad agreement with Owens’s line of assault on one prominent source of resistance to anti-individualist conclusions, Burge uses the occasion to elaborate further his views on the indexical strategy as a response to his anti-individualist arguments. He observes, along the way, that the anti-individualist point about the natures and identities of mental states is not a claim about supervenience and should not be construed in terms of some kind of supervenience failure. Such failure, Burge points out, “is logically compatible with individualism.” James Higginbotham’s “Competence with Demonstratives” seems to delve further into the more technical aspects of theory of language, apparently retreating from the metaphysical and epistemological concerns that are central to Owens’s discussion. Higginbotham takes off from a tension in our aims “when we attempt to express, from where we are, what a person says or thinks or wonders about from where she is, was, or will be.” The reporter of another’s expressed thoughts will, in formulating the report, be facing an ineliminable conflict, argues Higginbotham, “between the aim of preserving perspective and the aim of preserving reference, [a conflict that is] induced by the distinction between those aspects of language that belong to the setting-up phase and those that belong to the phase of saying things. . . .” The problem is that the devices available to us as reporters to
xxii
Introduction
ensure that reference and thus truth conditions and content are faithfully preserved, will of necessity obliterate some of those features of the speaker’s perspective or setting that she originally exploited in order to secure the very reference we are trying to preserve; this is unavoidable, to the extent that our perspective, and hence also the contextual properties to be exploited, differs from that of the speaker. This conflict, however, is not a problem for communicators; it is a challenge to theorists. In Higginbotham’s words, “We do convey perspective through the use of language; but we do not, and cannot, do it by saying what it is.” The question is, then, how do we systematically and illuminatingly capture this achievement of our ordinary linguistic practices? Higginbotham develops an answer that builds on Burge’s early work on the accommodation of demonstrative constructions in theories of truth for natural languages. In the process, he probes core issues regarding the constitution and delineation of linguistic competence. Higginbotham’s essay, in Burge’s words, deals with “fundamental issues in the philosophy of language, attractively disguised through the medium of low-key technical issues.” Or perhaps one might say that Higginbotham illustrates with unusual vividness how much is to be gained, in the intersecting areas of philosophy of language and philosophy of mind, when one resists the temptation to treat the resolution of technical problems and the exploration of fundamental issues as independent tasks. While generally sympathetic to Higginbotham’s handling of the conflict—and to his criticisms of attempts to neutralize it—Burge is not convinced that the original speaker’s perspective is excluded from the meaning of the reporter’s report quite to the extent that Higginbotham’s diagnosis suggests. Then, we may wonder, to just what extent is it excluded? That is, in Burge’s view, an open question. Still, he takes Higginbotham’s subtle discussion to establish that in “a satisfying investigation of language . . . more should be treated than recent semantic theory has allowed itself.” One issue that arises in Higginbotham’s and Burge’s discussions about what meaning conveys is how exactly to conceive of what is loaded into the Fregean notion of sense. Burge’s treatment of this topic is explicitly invoked by Christopher Peacocke, as one source of inspiration for his account of the theoretical benefits of what he calls implicit conceptions. In his contribution, “Implicit Conceptions, Understanding, and Rationality,” Peacocke offers a novel and significant extension of the explanatory scope of his influential account of conceptual capacities. How are we to characterize the conceptual resources feeding reflection that may lead us to endorse a new principle, that is, one that cannot be deduced from principles we already hold? A challenge here is to find a characterization that will, on the one hand, make it apparent how reflection may lead us rationally to conceive of or endorse a new principle, yet without, on the other, making that conception or endorsement an inevitable consequence of the ascribed prepositional attitudes. What we need is a way of characterizing our understanding of the constituents of the content of the
Introduction
xxiii
principle that strikes this balance. Endorsing the notion of implicit conceptions, Peacocke argues, allows us to account for, rather than merely affirm, the rationality involved in our coming to conceive or accept a new principle—“the phenomenon of new principles,” as he calls it. More generally it serves, as Burge puts it in his response, in “the explanation of reflective understanding,” by allowing us to account for “an individual’s improvement on his understanding of terms or concepts like chair or limit.” In the second half of his essay, Peacocke reaps some philosophical gains of his implicit-conception conception. In a wide-ranging discussion he moves from a curtailment of conceptual-role accounts of meaning, by way of a consideration of points of contact (and disagreement) with Wittgenstein’s conception of rule-following and understanding, to, finally, an articulation of a rationalist conception of our knowledge of basic axioms. Peacocke, in his essay, recognizes the deflationist temptation to dismiss implicit conceptions as an obscurantist way of describing inferential dispositions, and sets out to inoculate us against it. Burge is sympathetic; there is clear congruence between Peacocke’s line and Burge’s long-held view that we are committed to allowing thoughts comprised of concepts that the thinker understands imperfectly. While raising questions about how it is to be taken, Burge is not troubled by the general idea of implicit conceptions as real structures in our psychological make-up. However, Burge suggests, other kinds of unconscious structures may be involved in accounting for some of the phenomena that Peacocke describes. It seems to Burge “doubtful that implicit conceptions explain all the phenomena that Peacocke is concerned to explain.” Moreover, the issue here is not confined to the nature of the implicit structures invoked in accounting for a thinker’s coming to an explicit understanding of—articulating a rule for—a concept he uses. The suspicion Burge raises is that any conception of reflection producing conceptual knowledge, which treats this process as entirely a matter of bringing what is implicit to consciousness, will fail to do justice to the complexity of our concept-applying abilities. The extent of philosophical common ground between Peacocke and Burge is attested to in part by the fineness of grain with which their divergences come to articulation—most notably, perhaps, in the illuminating exchange, drawing on Leibniz and Frege, of the nature of our warrant for fundamental principles. A different sort of philosophical confrontation is introduced by Fred Dretske, in “Burge on Mentalistic Explanation, or Why I Am Still Epiphobic.” This essay and Burge’s response focus on an issue that marks one of the fundamental fault lines in contemporary philosophy of mind, where philosophical intuition is sharply divided. Nevertheless, dialogue across it can be illuminating and progressive, as the present exchange illustrates. Dretske’s problem is this: given the matters of philosophical substance on which he and Burge agree—in particular, about the broadness of content and the locality of the physical transmission of causal influence—then how is it that Burge can be so sanguine about the causal efficacy of mental states? Given anti-
xxiv
Introduction
individualistic individuation, what accounts for the explanatory significance of the particular content of a belief? How, Dretske asks, “can the fact that it is a belief that P (not Q or R) explain whatever effects the belief has on the believer’s bodily movements and, thus, behavior?” Keeping company with Burge, unlike many other naturalists, Dretske will not resort to an account of the casual role of intentional states in terms of their narrow or individualistic features. But Dretske knows the temptation of those who do so resort. To bring it out, he elaborates several vivid examples, among them this: though four quarters will get you a Coke from the machine, it is not the quarterness of the quarters to which the Coke machine responds. Rather, what cranks its chains are narrow features of the objects inserted. And just as the property of being a quarter is an epiphenomal property invoked in ordinary accounts of monetary-machine interaction, so belief content is an epiphenomenal property invoked in causal explanation of behavior. Although, as Dretske says, he disagrees with this latter claim, he is “sympathetic with the feeling that something has to be said to distinguish these cases from one another.” Burge, in his reply, reiterates two aspects of contemporary discussions of the perceived threat of epiphenomenalism to which he objects. The first of these is what he calls a “makebelieve posture,” an “artificial and histrionic . . . spirit of angst.” Second, and, one supposes, relatedly, Burge objects to “prevalent assumptions—centering on materialist metaphysics—about what sorts of considerations must be dominant in an explanation of why epiphenomenalism is false.” However, this negative attitude is directed at the cast of some of the current discussion, not at the philosophical project, per se, of inquiring into the nature of mental causation or mind–body causation. Engaging that project in a critical way, Burge devotes much of his response to the examination of the intuitions at work in examples supposed to induce epiphobic worries. Neither the materialist assumptions nor the arguments from them are, in Burge’s view, convincing. The relation between anti-individualism and the project of constructing a naturalistic or physicalistic conception of the mental, an explicit theme of Dretske’s contribution, is also an issue in Ned Block’s paper and Burge’s reply. “Mental Paint,” however, poses no direct challenge to Burge, and engages with Burge’s work principally in the dialectical use that Block makes of various thought experiments in charting our mental lives. Taking a stand on one side of a fundamental divide—in his words, “a chasm”—between theoretical views of consciousness, Block defends what he calls “phenomenism,” the view that “phenomenal character outruns representational content” (as well as the functional and the cognitive). The present contribution is primarily critical in thrust, and its target is representationism. This is the view that “the phenomenal character of experience is exhausted by . . . representational contents.” The least implausible variety of this false view is, Block suggests, the externalist one, which, accordingly, is the central topic of his paper.
Introduction
xxv
What is fundamentally at issue between Block and representationists such as Gilbert Harman and William Lycan? To focus the argument, Block distinguishes three kinds of things: first, the intentional content of an experience; second, mental paint, that is, mental properties of the experience representing the perceptual properties of the object perceived (the redness of the tomato); and third, mental oil, that is, mental properties of an experience that don’t represent anything at all. With regard to the second, Block holds that mental paint exists, that it can be attended to, and that its essence is not captured by its representational content. The last of these is the (weak) antirepresentationist claim regarding mental paint. The strong antirepresentationist claim concerns the third element, mental oil; there are such properties—notably in orgasms “(the ones that make the orgasm experience something one wants to have).” With characteristic verve, Block draws on his thoughtexperimental resourcefulness to wear down representationist resistance. In his reply, Burge is sympathetic. The persistence of representationism appears to him symptomatic of just that second feature of certain strands of contemporary discussion of the mental with which Burge registers his impatience (in his reply to Dretske): “the defensive maneuvers that Block considers and criticizes seem to me to derive from philosophical strategy—an attempt to make a programmatic metaphysical ideology, encompassing a scheme for confronting the mind–body problem, ‘come out right’—rather than from insight into the matters at hand.” In broad agreement with Block’s “ingenious and varied arguments” against the representationist construal of awareness, Burge emphasizes that this causes no dissonance in his conception of mind. Anti-individualism with respect to representational content combines without strain with the view that nonintentional phenomenal characteristics feature prominently in our mental lives. But there are subtle complications. Burge grants, for example, “that sameness of phenomenal character is compatible with difference in intentional content, and indeed difference in reference.” Yet he doubts “that sameness of intentional content—all the way down, at the finest intrasubjective level of individuative grain—is compatible with difference in phenomenal character. . . . The perceptual intentional content, the perceptual modes of presentation, will commonly be in some way, at some level, different, if phenomenal character is different.” According to Burge, this last is so because “phenomenal character is commonly an individuating element in the individual’s recognitional ability. So it is an element in individuating the intentional representational content of an individual’s perceptions.” But then, what of Block’s empirically grounded claim that phenomenal characters associated, for example, with basic color perception may vary widely from person to person? If we accept this point, and yet hold that phenomenal character plays a content-individuating role, do we not risk undermining the intersubjective applicability of perceptual contents? Burge’s discussion of this issue toward the end of his reply is methodologically significant; it
xxvi
Introduction
provides a clear view of his understanding of the point and the grounding of philosophical descriptions of our mental lives. In “Mental Content and Hot Self-Knowledge,” Bernard W. Kobes considers the relation between the view that much of our mental content is externally fixed and our sense of “the immediacy and authority of our knowledge of our own current conscious mental contents.” His starting point is a prima facie tension: “if S’s thought contents are fixed partly by conditions external to S, would it not be incumbent on S to investigate those external conditions in order to know what his own thought contents are?” What Kobes calls the “Burge–Heil reconciliation story” immediately meets the challenge in an important range of cases, namely those cases where self-attributions are made by virtue of employing the very content of the first-order belief. Developing “a revised and extended version of the account,” Kobes aims to answer “a number of doubts and discontents” that have emerged in response to it. The foundation of Kobes’s story is the idea of cognitive agency. In “hot self-knowledge,” the performative element in self-knowledge is construed along the lines of our knowledge of what we are going to do when we intend to do something. The significance of this is that the knowledge-state involved is “telic” rather than “thetic”; the “direction of fit” is world-to-mind rather than mind-to-world. On this conception of selfknoweldge, Kobes suggests, our entitlement in attributing thoughts to ourselves is derived from a mental act, insofar as it provides the subject matter of the thought. The power and fruitfulness of Kobes’s contribution is further evinced in the reply from Burge. Acknowledging his agreement with Kobes’s driving conviction, Burge notes that “agency is at the heart of our understanding of first-person authority.” But a full account of nonempirical entitlement to self-attributions requires us to take heed, in duly qualified and restricted form, of Spinoza’s view that “formation of belief is . . . not an act.” An account of the authority of self-knowledge that trades essentially on our taking the selfattributed belief to be formed through an exercise of intellectual agency cannot be the whole story. In his first extensive and systematic discussion of these issues—in “Individualism and Self-Knowledge” (1988)—Burge, as he puts it here, “concentrated on pointing in a direction for an account of the phenomenon.” The complexity of the phenomenon and the differences between various kinds of cases were never lost from view. In the present discussion, some of the grounds for the cautiousness of that paper are scrutinized. One significant focus in the Kobes–Burge exchange is a particular challenge to antiindividualist accounts of self-knowledge raised by Brian Loar. Loar once pressed the claim that the reconciliation story is threatened by an infinite regress, demanding an account of the entitlement to self-attribution at ever higher levels of thought. In his contribution to this volume, “Phenomenal Intentionality as the Basis of Mental Content,” Loar engages antiindividualism on a different ground, but his concern remains the same, namely to vindi-
Introduction
xxvii
cate, as he puts it here, the “intuitive idea that the intentionality of outward-directed thoughts can be internally determined.” But this idea “has run into serious trouble,” and the force of the anti-individualist case against it calls for strong countermeasures, as Loar makes clear at the outset: “Something theoretically novel (though familiar in experience) needs acknowledging. My homage to Burge in this volume will be expressed by my being driven to extremes.” This theoretical novelty, introduced in the context of a discussion that links up directly with Block’s paper, is the idea of phenomenal intentionality or directedness. Loar introduces it (in his section 4) as “a phenomenologically accessible feature of virtually all perceptual experience and of perceptually based concepts.” The key point for Loar is that this directedness is “an internally determined phenomenal feature” of experience, to be identified in the experience itself, whether, as in Loar’s example, of real lemons or lemon hallucinations; it is reference independent. Can it be employed to construct a notion of mental content that survives anti-individualist arguments? That depends, for Loar, on whether “we can hold constant phenomenologically accessible intentional visual qualia while varying all the properties that they represent things as having.” This would, in effect, mean that we sever the link between the characteristic directedness of (in this case) visual-perceptual states from objective features of the world beyond the perceiver. The heart of Loar’s case for the internalist conception of mental directedness is an appeal to an envatted twin brain. Allowing the “phenomenological coherence” of the idea of Loar’s “twin-in-a-vat having visual experiences intentionally equivalent” to Loar’s own experiences, we can see our way clear, if we follow Loar, to an internalist notion of mental directedness. In the last sections of his paper, Loar develops the view by extending it to other classes of concepts. In a reply that ranges across many of the themes raised in earlier discussions—with Donnellan on the nature and force of the thought experiments, with Higginbotham on reference-preservation, perspective, and context, with Block on representationism, with Owens on the temptations of indexicality-readings, to mention some—Burge makes it clear that he is not convinced by Loar’s constructive efforts. He remains fundamentally critical of the direction in which Loar wants to move our conception of the mental, and draws into question both the motivating intuitions and the force of the reasoning behind the proposal. Once again, Burge emphasizes that his anti-individualism is essentially argued from consideration of cases and not from principles, and he suggests that the dialectical significance of this is not fully appreciated by Loar. Moreover, although he takes the anti-individualistic nature of (at least many of) our concepts to be incontrovertibly established by consideration of the various thought experiments, Burge also holds that what is thus established is compatible with materialism, for instance (at least when construed in terms of constitution rather than identity), and indeed with many of the intuitive
xxviii
Introduction
things that Loar’s internalist wishes to say about the phenomenology of the mental life. And while there is fundamental disagreement between the two, Burge acknowledges that Loar “is on to something that needs better understanding. . . . Loar’s taking phenomenal experience seriously may be fruitful for our understanding of intentionality.” In this direction, suggests Burge, lie questions to which we may not, at present, have very good answers. The final contribution in the volume is Noam Chomsky’s “Internalist Explorations,” and the title, if nothing else, might have the reader expecting another direct challenge to Burge’s fundamental doctrine. The thrust here, however, is different. It is not Chomsky’s aim to construct a notion of intentionality that will save mental content from external individuation. Chomsky’s concern is rather to display the methodological motivation behind the kinds and types he proposes to employ in the study of cognitive abilities, in particular, of course, of linguistic abilities. “Internalism” in this context emerges as the name of a general sort of constraint on these kinds and types, a constraint derived from—or expressive of—Chomsky’s conception of linguistics as a branch of natural science. His arguments against “externalist” conceptions are arguments against a set of conceptions of psychological states that he regards as scientifically fruitless, insofar as they appear to entail that the states involved are not genuine states of individuals. Certainly, common classification of views as “internalist” or “externalist” creates the impression that Chomsky’s line of thinking is deeply at odds with Burge’s. As readers of Burge’s earlier work on Chomsky will know, however, the extent of substantive and methodological agreement between the two is significant, and Burge elaborates it here in some detail. Once again, the diagnostic unreliability of the “internalist”–“externalist” opposition is brought home to us; a lesson of the present volume is that this opposition tends to obscure, rather than reveal, significant points of convergence and conflict between competing positions in the philosophies of mind, of language, of psychology, and of linguistics. Chomsky’s essay provides Burge with an opportunity to display his own position—the nature of his claims as well as the nature of the warrant for them—from a revealing perspective. For example, one might have thought that Burge’s self-assertively philosophical conceptual investigations, relying on intuition and thought experiments to establish purported conceptual knowledge about the nature of the mental, would fly in the face of Chomsky’s steadfastly empirical approach to the investigation of cognitive structures. But this would be a misunderstanding. When Chomsky says, “intuitive judgments are data, nothing more,” this is not fundamentally at odds with Burge’s conception, though Burge thinks that “if this slogan is generalized, it greatly overstates the case.” When Chomsky goes on to say that judgments “might become evidence within the framework of some explanatory theory,” Burge may agree with him—in many kinds of cases. However, this
Introduction
xxix
cannot be the whole story, on Burge’s view. In support of this resistance, Burge elaborates his view of the role of reflection and of the a priori in the development of knowledge. It is in the context of this nuanced view of the role and nature of reflection that Burge’s anti-individualism must be assessed. Anti-individualism concerns, as Burge puts it in his reply to Chomsky, the question of “whether the existence and nature of certain psychological kinds depends necessarily on the existence and nature of certain relations to specific kinds or situations in the external world.” What reflection on the thought experiments establishes, if Burge is right, is that such dependencies are common. If this is so, it is a philosophical fact revealed by reflection; it does not depend on explanatory theory and general principles, but rather constrains them, as something they must be able to accommodate. Clearly, there are differences of emphasis between Chomsky and Burge, as well as disagreement on substantive matters. It is striking, however, that those conflicts of view emerge, in this encounter, on the whole as tractable ones. They are locatable—provided care is taken and sufficient fineness of grain is employed in the characterizations—within the shared context of psychological realism and a sophisticated philosophical rationalism, along the axes of the different but complementary forms of investigation that such a philosophical framework allows. And here is a feature characteristic of many of the exchanges in this volume: although agreement may be elusive, tractability is improved. One comes away from this volume with a better philosophical map, a more precise view of the pressure points in our current attempts to come to an understanding of our mental life.
1
Burge, Descartes, and Us
Calvin Normore I will first sketch what I take to be Tyler Burge’s view of Descartes and of how Descartes went wrong in the philosophy of mind. My focus here will be on Burge’s paper “Cartesian Error and the Objectivity of Perception.” Then I will sketch another, perhaps revisionist, view of Descartes. Finally I will try to show that the revisionist Descartes is struggling, not altogether unsuccessfully, with some problems that have been overlooked in recent philosophy of mind and that, once appreciated, cast the externalist–internalist debate in a new light. I Burge’s Descartes Burge begins his paper “Cartesian Error and the Objectivity of Perception”1 with the claim that “Individualism as a theory of mind derives from Descartes. It dominates the postCartesian tradition—Locke, Berkeley, Leibniz, Hume—up until Kant. And it has reemerged in the writings of Husserl and of many English-speaking behaviorists and functionalists” (p. 117). The individualism Burge has in mind is not easy to characterize precisely, but he proposes as a working account the following: Individualism is the view that an individual person or animal’s mental state and event kinds—including the individual’s intentional or representational kinds—can in principle be individuated in complete independence of the natures of empirical objects, properties or relations (excepting those in the individual’s own body on materialist and functionalist views)—and similarly do not depend on the natures of the minds or activities of other (non-divine) individuals. The mental natures of all an individual’s mental states and events are such that there is no necessary or other deep individuative relation between the individual’s being in states or undergoing events with those natures, and the nature of the individual’s physical and social environments. (pp. 118–119)
Burge thinks that individualism has a number of motivations but that the most powerful were and are a series of epistemic considerations that are focused by Descartes’s own thought experiments, particularly those in the First Meditation. Burge lays aside the details of the Cartesian thought experiments on the ground that they, for example, the dream hypothesis and the hypothesis of a deceiving demon, seem to bring in the physical or social environment in ways that make their relevance to individualism problematic. He proposes instead to focus on some core elements in the Cartesian thought experiments and, in particular, to “construe Descartes as capitalizing on the causal gap that we tend to assume there is between the world and its effects on us; different causes could have produced ‘the same’ effects, certainly the same physical effects on our sense organs.” He continues: “I will interpret him as conceiving a person as radically mistaken about the nature of the
2
Calvin Normore
empirical world. I shall see him as imagining that there is something causing the given person’s mental going on, but as imagining that the entities that lie at the ends of relevant causal chains (and perhaps the causal laws) are very different from what the person thinks” (p. 120). Burge wants to insist that, so construed, the Cartesian thought experiments do nothing to bolster individualism. They seem to bolster individualism because they seem to support an argument like the following: “a) Suppose that one imagines that one’s thoughts are subject to error in one of the Cartesian ways. b) One knows what one’s thoughts are, and they would be mistaken. But c) an anti-individualist position holds that in some of the Cartesian cases we would think thoughts different from those we actually think—we would be in different mental states. d) This conflicts with our authoritative knowledge about what (some of ) our thoughts are and would be” (p. 123). Burge thinks the Cartesian thought experiments seem to bolster individualism only if one neglects a crucial distinction. We can get at this distinction, he suggests, if we consider the thought experiments in the context of three principles, each of which seems compelling and which together enable us to derive a nonindividualist conclusion from the same thought experiments. These premises he sums up thus: “. . . our perceptual experience represents objective entities; perceptual experience specifies objective entities as such; and the formation of perceptual representation of perceptual intentional types is empirical” (p. 131). The strategy of the anti-individualist argument from these premises, in Burge’s words, is simple. the first premise notes a possible gap between a person’s physical states and intentional states, on one hand, and the state of the world that is seen on the other. Holding the relevant physical effects constant, we imagine different visible objects in the world, and different optical laws normally and regularly relating those objects to the person’s physical states. In such a case, it is clear that some of the person’s intentional visual states, at least some of those that specify objective entities as such, would be different. (p. 135)
The individualist as such need not reject Burge’s three premises, so we seem to have two arguments from the same situation to opposite conclusions. What has gone wrong? Burge’s conclusion is that what has gone wrong is our understanding of first-person authority. He suggests: “We are authoritative about some of our actual thoughts about the empirical world; and we can imagine those very thoughts being quite mistaken. Moreover, whatever our thoughts would be if the counterfactual situation were to obtain, we would be authoritative about some of them. But we are not authoritative about what our thoughts about the empirical world would be if the counterfactual cases were actual. That is a philosophical issue, not a matter of what one’s present mental events actually are” (pp. 123–124). This is a bit cryptic, but I think we can get at Burge’s point from another perspective—by considering Jean Buridan’s distinction between the possible and the possibly true. In his Sophismata the medieval philosopher Jean Buridan raises a puzzle case
Burge, Descartes, and Us
3
about the sentence ‘No sentence is negative’.2 On Buridan’s view of affirmative predication, an affirmative sentence is true only if its subject term refers. Thus “ ‘No sentence is negative’ is true” is itself true only if the sentence ‘No sentence is negative’ exists. But if that sentence does exist then it is false. Hence that sentence cannot be true. Nonetheless if sentences are written or spoken tokens as Buridan thinks then it is not difficult to imagine a world in which no sentence is negative. Hence things could be as the sentence has it but the sentence could not be true.3 Just as Buridan’s sentence can be formulated in our world and as formulated in our world can truly describe a world in which it could not be truly formulated, so Burge thinks that our thoughts formulated in our world can describe situations that would not be described by thoughts we formulated in those situations. Burge’s picture of Descartes is richer and fairer to Descartes than is usual among those who discuss him nowadays. It does, however, raise a line of questioning I would like to explore a little further before turning to my own account of Descartes’s view. One form of contemporary externalism is suggested by some remarks of Kripke’s in Naming and Necessity. Kripke’s discussion began from the theory of proper names and was designed largely as an attack on the idea that names had the sense of a description or set of descriptions associated with the name a priori. So Kripke has us imagine that the stories in the Bible about Jonah are indeed about Jonah but all false, so that we know nothing about Jonah save that these stories are about him. Kripke proposes to generalize this account to kind terms. Perhaps tigers have none of the features we normally ascribe to them. Perhaps they are reptiles not mammals, live in oceans not jungles, and so on. Nonetheless, if the baptismal encounters went well it is tigers about which all our beliefs are false. Kripke himself does not suggest that all our kind terms are bare in the sense that there is nothing we can see to be true about them a priori. Indeed his discussion of pain in Naming and Necessity seems to presuppose that ‘pain’ is a kind term and that there is an a priori connection between pains and their ‘feel’.4 But one might imagine a kind of externalism that does take all our basic categorematic concepts to be bare in this way. This form of externalism suggests a difficulty. If our basic concepts are bare in the sense that they connect us with objects but are conceptually independent of connections among those objects then it would seem that causal externalism is compatible with a strange form of scepticism. We can be sure that many of our beliefs are true, but in a sense do not know what those beliefs are. To see why, suppose that concepts are very like terms in a natural language—in particular, that they can be individuated in a way that is independent of what they are concepts of—so that in principle the same concept could apply to very different things or kinds of things. Now suppose I encounter a thing and form an individual concept of it. Suppose that on the basis of that encounter with it and some encounters with others (which are as a matter of fact of its kind) I form the concept of
4
Calvin Normore
orca. Suppose on the basis of those encounters and yet others I form the concept of animal—and so on. I may now be in a position to state some truths with confidence, for example, ‘If there ever have been any orcas and any animals then orcas are animals’.5 But my confidence in this truth is compatible with my having no idea what orcas are like. The reference of ‘orca’ and the content of the concept of orca are fixed by the causal chain that gives rise to their use—and we can even suppose that the relationship to the term ‘animal’ and the concept of animal is similarly fixed. I could not discover that ‘If there ever have been any orcas and any animals then orcas are animals’ is false. But all of this leaves open the possibility that I could be presented with the very orcas and animals from contact with which my concepts were acquired, and presented with them under optimal perceptual conditions, without my having the faintest suspicion that they were orcas or animals, and that I could be presented with rocks and galaxies under optimal perceptual conditions and have not the faintest suspicion that they were not orcas and animals. If my concepts are bare, nothing in them gives us a hook out to perception. What thought experiments like these seem to suggest is that percepts, the products of perception, are not bare. When we perceive things we perceive them as. If I see a bear off in the distance I may see it as little more than a black dot on the horizon; from much closer I may perceive it as a bear, perhaps even as ursus horribilis. Although it is not at all obvious that I can’t have a concept of Socrates that is of Socrates but isn’t of Socrates as anything, it does seem that I cannot perceive Socrates without perceiving him as. Perceptual states have representational content. I think that in this I am of Burge’s mind. Burge wants to insist that externalism as he understands it does not give us as basic perceptual states with contents like ‘whatever normally causes this sort of perceptual representation’ but instead gives us perceptual states with contents that specify particular objective types of objects, properties, or relations as such. Perceptual representations specify some particular objective entities (e.g., a boundary) as such (e.g., as a boundary). There seems to be a tension between a view that would expand the direct reference picture centered on demonstratives, proper names, and kind terms to perception and a view that insists that perceptual states specify objects as. A proper name cannot misrepresent that of which it is a proper name. It cannot misrepresent it because it does not represent it as this rather than as that. It doesn’t represent it as this rather than as that because its semantic function is wholly exhausted in picking out its referent. To misrepresent its referent it would have to have some further content. One central issue in contemporary philosophy of mind and of language has been focused precisely here—how to preserve the insight that the reference of a thinker’s thoughts and words is wholly fixed by relations, usually causal relations, to a world outside the thinker (and so dependent on that world) and at the same time preserve the strong intuition that
Burge, Descartes, and Us
5
some concepts and most, perhaps all, perceptual states could misrepresent the very things they represent. II
Descartes
This problem is, I suggest, one of the central issues motivating Descartes’s account of representation. Causal externalism of the “bare concepts” sort was an option available in Descartes’s world. It had already been worked out in the fourteenth century by William Ockham, Robert Holkot, and others. On this Ockhamist picture, a concept is a quality or act of the mind that is produced in the mind by causal contact with some thing or things in the world. The concept is individuated by the thing or things that cause it so that it— that very concept—could not have been caused by appropriately different things. The mental quality or act that is the concept could have been produced by other things, however, and had it been so produced it would be a different concept. It is, I think, a picture all of us find rather familiar.6 But the Ockhamist picture was not Descartes’s preferred view. Descartes begins from a picture of thought that was also worked out in the fourteenth century and indeed was advocated by Ockham before he abandoned it for a more straightforward causal theory of reference. It was kept alive within the Scotist tradition throughout the early modern period until, through Brentano, it directly influenced Husserl, Meinong, Sartre, and Roderick Chisholm. This is the theory that thought is a relation between the mind and an intentional object—an object that is an object of thought and that has a kind of reality specified by that very relation, namely, objective reality. Descartes spells out his attraction to this line of thought in the Third Meditation, where he distinguishes two ways in which we may regard an idea—formally and objectively. Regarded formally an idea is a mode of mind. Thus understood it is dependent on the mind of which it is a mode. No thinker, no thought. From this perspective all ideas are equal in what Descartes calls “degree of reality.” All have a lower degree of reality than any substance but are not nothing. But ideas properly speaking are also, Descartes claims, “as if images,” and, following the tradition that talked about intentional or representational or objective reality, Descartes supposes that an image is not merely the substratum in which it is (to speak in an extended way) embodied. Thus a painting is not merely a collection of bits of paint on a canvas such as might, by accident, be produced by a team of monkeys working in the British Museum. As Descartes writes: “Suppose someone said that anyone can paint pictures as well as Apelles, because they consist only of patterns of paint and anyone can make all kinds of patterns of paint. To such a suggestion we should have to reply that when we are talking about Apelles’ pictures we are not just considering a pattern of colors, but a pattern skillfully made to
6
Calvin Normore
resemble reality, such as can only be produced by those very skilled in this art” (CSMK7 214, AT III, 567). This idea, that there is more to a painting than the pattern of colors, is exploited by Descartes in his account of ideas. Just as a painting involves both an assemblage of colors and a picture, so an idea involves both a mode of mind and an object present with objective reality. This second element, the object present with objective reality, has different degrees of reality in different ideas. As Descartes puts it: “Insofar as ideas are considered simply as modes of thought, there is no recognizable inequality among them: they all appear to come from within me in the same fashion. But insofar as different ideas are considered as images which represent different things, it is clear that they differ widely” (CSM II, 27–28 AT VII, 40). Moreover, Descartes insists that the second element has enough metaphysical status to require a cause and that the sort of cause required depends on the sort of object present. The cause of this objective element in an idea must “have as much reality formally or eminently as the idea has objectively.” It is thus that he proves the existence of God in the Third Meditation. Examining his idea of God he finds there an element with so much reality objectively that only a perfect being could have that much reality formally or eminently. Thus the second element of his idea must have been caused by such a being and so there must be such a being. We think of truth and falsity as primarily properties of sentences, statements, and judgments. It is not always easy to distinguish ideas and judgments in Descartes, but in this area he is clear that it is judgments that are formally true or false and yet that there is some other material falsity in ideas when they represent a non-thing as a thing [cum non rem tamquam rem repraesentant]: so, for example the ideas which I have of heat and cold are so little clear and distinct that from them I am unable to learn whether cold may be a privation of heat or heat a privation of cold or both may be real qualities or neither. And because there can be no ideas which are not as if to be of things [Et quia nullae ideae nisi tamquam rerum esse passunt] if it is indeed true that cold is nothing other than the privation of heat, the idea which represents that to me as something real and positive would not be called false without merit and so of the others. (AT VII 43–44)
There has been much discussion of what exactly makes a materially false idea false. I want here to focus on a simple suggestion that hews close to Descartes’s text. An idea is materially false if it represents something as other than it is. This suggestion involves the thought that an idea does not simply represent something but represents it as something— and not always as what it is. In the text just quoted Descartes considers the possibility that a privation (say cold) might be represented as a positive mode. That an idea represents a thing as something distinguishes Descartes’s picture from the “bare concepts” view. Cartesian ideas are not bare bearers of reference; they have content. But the suggestion does not distinguish Descartes from Burge, who also insists that at least our elementary concepts present things to us as—this is the burden of the second premise
Burge, Descartes, and Us
7
of his argument for anti-individualism in the context of the Cartesian thought experiments. But Descartes has a very special view of how the representational character of an idea works. What I earlier called the second element in his account of ideas—what is present with objective reality—is what for Descartes provides representational character. There are for Descartes two sources of error in ideas. One is the mistaken supposition that there is something in the world that an idea is of; the second is the representation of something in the world as other than it is. Descartes, unlike most twentieth-century philosophers, is primarily interested in the second of these. His concern is less with what things there happen to be than with what there could be, and his major worry, even in the First Meditation, is not that we will be deceived into thinking that there are material pyramids when there aren’t but that we will be deceived into making categorical mistakes— thinking that a mode of mind like color is a mode of matter. He has this focus partly, I think, because so many of his concerns are mathematical and partly because his physics is less a detailed analysis of particular systems than a general account of how a world of matter in motion could go. There are two sources of error in ideas for Descartes because there are two ways in which an idea may be caused—it may be caused formally or it may be caused objectively. Let me pervert some terminology by calling the cause of the objective reality of an idea the object of the idea and the cause of the formal reality of an idea the target of the idea. Now we could proceed by distinguishing two sense of ‘about’, and say that what an idea is about is in one sense its object and in another sense its target. On this picture an idea would be veridical if its target was its object. In the cases where target and object diverged, an idea would misrepresent. Descartes does not take this line. He instead reserves the notions of representation and misrepresentation for what I have called the object of an idea. He agrees that God would be a deceiver if it did not regularly happen that the object of an idea was also its target and so concludes that it does regularly happen, but if (per impossibile) God were a deceiver in this sense the content of our ideas would be unaffected. Note that as we read Descartes, then, he is an externalist in the sense that the content of our ideas depend on their causes— but on the cause of the objective reality of the idea, not the cause of its formal reality. It is almost time to return to the thought experiments, both Cartesian and Burgean, with which we began, but first one more distinction. Descartes distinguishes ideas of true and immutable natures from fictitious ideas. Exactly what is at stake in this distinction is far from obvious, but at least in part it seems to be that from the ideas of true and immutable natures one can draw surprising conclusions whereas from fictitious ideas one cannot. The ideas of true and immutable natures are productive. The significance of this is that Descartes wants to insist that only the ideas of true and immutable natures need really be caused in us (objectively) by the natures themselves. The fictitious ideas we construct, and
8
Calvin Normore
we need not suppose that we get them right. Chimaeras and existent unicorns appear in our ideas but with no more objective reality than is found in the ideas of the true and immutable natures that make them up. The Descartes I have (re)presented above is deeply concerned about error. He thinks that we must take seriously the possibility that we are mistaken about the existence and even the nature of the physical world. I put the claim that way to emphasize that for Descartes the threat that we might have the nature of the world wrong is the deeper threat. Idealism would hardly touch Cartesian science, but scholasticism—which agrees with Descartes about the existence of the physical world but disagrees about its nature—would leave Cartesian science in shambles. This significant difference between Descartes and the modern externalists is connected with a second difference. Modern externalism focuses on the physical world. Thus modern externalists fare best accounting for our physical kind concepts and least well accounting for our purely mathematical concepts. Descartes’s externalism, on the other hand, is motivated by mathematics. His physics operates at a very high level of abstraction from particular phenomena, and there has been considerable and ongoing debate about whether he even has particular physical objects in his ontology. This emphasis on kinds (what Descartes calls natures) gives Cartesian externalism a flavor that seems peculiar to us. Descartes wants to insist that what natures there are depends on the will of God and that although God’s will is immutable God could have “made” different natures from the ones he did make.8 Had God made different natures, our thoughts would have been different in the sense that they would have had different objective contents. Would they have been different considered merely formally—as modes of mind? Descartes never addresses this question—he is not after all a late twentiethcentury philosopher—but there is nothing of which I am aware in his philosophy that would prevent him from admitting the possibility that had God made different natures but left our minds the same then formally the same mode of mind would have had a different objective content. In that case Descartes would be an externalist in a very strong sense. Even if this turns out not to be the case, and Descartes holds that in the counterfactual situation in which God makes different natures our modes of mind are also different, Descartes would not be an individualist in at least one strong sense: his thoughts and their contents would depend on something outside his mind. Descartes does not admit that we could think about natures that God does not create. But how in that case can Descartes seriously worry about whether he might be radically deceived even about the basic natures? If the objective content of Descartes’s thoughts is determined by the natures God makes, how could most of Descartes’s thoughts be deceptive? Does he not have available an analogue of the argument that we are not brains in a vat?
Burge, Descartes, and Us
9
I think the answer is that he does—just in case the idea is clear and distinct. If the idea is clear and distinct then there is no possibility of a gap between the objective reality of the idea and the way in which that objective reality is presented—what we may call, following Margaret Wilson, the representational character of the idea. That is what Descartes’s reflection on the cogito shows him. If the idea is not clear and distinct then we may be confused about what is given objectively in the idea—the idea may, for example, be materially false. This issue, of whether we can read the object of the idea off the idea itself, is, in general, prior to the question of whether that object might have been supplied by something other than the target of the idea. III
Thought Experiments Revisited
The core of Burge’s response to Descartes’s demon is that in demonic worlds our thoughts are different from the thoughts we now have. I guess that of the Ur demon world, the world in which there is nothing but the demon and the meditator, Burge would say that all our thoughts are about the demon or about ourselves. Descartes would not say this. Descartes would, I think, say that in the Ur world we have no general thoughts at all because there are no natures and so no general thought-objects. That is a very different but not an individualist thing to say. It is inelegant to distinguish where no distinction is needed, and one might think that Descartes’s distinction of the cause of the formal reality of an idea and the cause of its objective reality is just such a case. Certainly Arnauld thought so. But, as usual, Descartes’s distinction is motivated. We can see some of the motivation by considering Descartes’s problem of the two suns. Here we are very fortunate that, following up the trail pioneered by Prof. Armogathe in his discovery of Descartes’s thesis for the Faculty of Law at Poitiers, Dr. Deborah Brown was able to unearth a posthumous letter from Descartes to Burge in which this very case is the focus.9 Descartes thinks that we have two ideas of the sun. One idea (let us call it the Manifest Image) is of the sun as a small disk about the size of a Canadian dollar coin, a “loonie.” The other (the Scientific Image) is of the sun as a huge gaseous body many times the size of the earth. Both of these are ideas of the sun. But one of them is of the sun as a disk the size of a loonie the other is of the sun as a huge gaseous body. Descartes wants to insist that these are different ideas because if they are not then astronomy has not revised our conception of the sun—which seems to him false. They are different ideas and not different judgments (beliefs) involving the same idea—though I grant that this distinction is not always clear in Descartes. Let me now turn to the recently discovered letter from Descartes to Burge. Speaking of the Manifest Idea of the sun Descartes writes:
10
Calvin Normore
Let us perform a Twin Earth experiment on this idea. Suppose that the Twin-sun is in fact smaller than Twin Earth. Suppose also that it is composed of hot gases which are different from the gases of which our sun consists. On your view, given the difference in the suns, the idea of the sun that A has in the actual case would be different from that A has in the counterfactual situation. But do you really mean to say that because our idea of the sun is normally caused by a heavenly body larger than the earth that this information is somehow encoded in the idea? . . . What else could explain [your view] that there is a difference between the two ideas? . . . but of course, that is absurd. It would follow that astronomical reasoning has not corrected our idea of the sun. (p. 167)
Descartes’s reasoning here is a little compressed but perhaps could be unpacked this way. Burge claims that were our circumstances different our thoughts would be different. He also claims that “we have perceptual representations (or perceptual states with contents) that specify particular types of objects, properties, or relations as such” (p. 125). Burge does not say which of our perceptual representations have this character. He speaks of “our perceptual kinds and of other kinds that are taught by more or less immediate association with perceptually based applications” (p. 131). Descartes seems to be interpreting him very broadly as committed at least to the view that our representation of the Sun does. Thus, the difference between the Sun and the Twin Sun would be reflected in our conceptions of them. But this seems to require that our conception of the Sun have been different all along from our putative conception of Twin Sun and different in a way that is veridical. So it seems that our conception of the Sun was always truer than we knew and that astronomy corrects us less than we thought. Is this fair to Burge? It would seem not. Surely Descartes is mistaken in thinking that as complex a representation as that of the Sun must represent the Sun as such. Moreover Descartes is surely unfair to Burge in moving from perceptual representations (by which we perceive perceptual kinds as such) to concepts that are not perceptual and, in the case of the Sun, not even obviously of kinds. Descartes, it would seem, is here not so much arguing from premises Burge would accept as arguing out of his own picture—a picture that makes basic not perceptual kinds but true and immutable natures and that does not distinguish percepts and concepts as Burge does. But could the problem not be replicated at a level that Burge would accept? Burge considers a case in which there are two worlds differing in the minimal ways that would allow there to be small shadows in our world just where there are small cracks in the counterfactual situation, having us be in the same internal states while seeing the shadows that our counterparts are in while seeing the cracks. Burge suggests we not only see the shadows but see them as shadows. They see the cracks as cracks. Occasionally one of us might misperceive a crack as a shadow, but those are the exceptional cases. Shadows and cracks are fairly minimal objects but not so minimal that they have only the features of being shadows and cracks. Imagine that the counterfactual world in ques-
Burge, Descartes, and Us
11
tion has no dark objects other than these cracks. Can its inhabitants have the concept of ‘dark’ as distinct from the concept of ‘crack’? If they can conceive cracks as cracks, can they not conceive cracks as dark? If they can, then can they not conceive cracks as small, thin and dark? There seems here to be a slippery slope between perceiving a crack as a crack and perceiving the sun as a small disk the size of a loonie. But in that case there is a slippery slope between the view that some of our basic percepts are veridical a priori and the view that we cannot be mistaken in any of our perceptual representations. Somewhere along this slope Burge will dig in his heels, and it would be very helpful to know where and why. As for Descartes, for him all percepts are concepts and all concepts are “as if images.” On his view both the manifest idea of the sun and the scientific idea of the sun are of the sun, but the second is of the sun as it is while the first is of the sun other than as it is. For an idea to be materially true it must be of its target as its target is. This can be guaranteed when the idea is of a true and immutable nature as such, but when the idea refers to something existing (other than God) then error can creep in because the idea may not present the object as it is. Burge claims that for perceptual kinds the perceptual representation must present the kind as it is. Descartes will typically deny this, claiming that because there are two causes of ideas—their targets, which produce them as modes of mind, and their objects, which determine how the target is conceived—the possibility of error exists even in the perceptual cases Burge has in mind. He would, I think, challenge Burge to explain why these cases are privileged and suggest that there is no natural stopping point on the slippery slope from veridical perception of basic kinds to the impossibility of perceptual (and even conceptual) massive error. If the picture of Descartes just painted is correct, he is not an individualist in Burge’s sense but an externalist of a robust sort. But he is not, I think, an externalist of Burge’s own persuasion. One of the many important contributions Burge has made to recent philosophy of language and mind is to refocus the “externalist” debate from considerations about names and natural kind terms to categorematic terms generally. Burge’s examples of ‘arthritis’, ‘brisket’, ‘contract’ and so on not only point up the community’s role in meaning but show that the issue is not primarily about a special kind of term. Within a Cartesian framework these examples raise special questions because Descartes usually focused on (and is often understood to have exclusively focused on) what we might think to be natural kind terms. His ‘true and immutable natures’ hardly extend to ‘brisket’ and ‘sofa’. Yet Descartes was not altogether unaware of the fact that the world as we ordinarily conceive it is not altogether a product independent of us, and it seems clear he thought that on some occasions social factors broadly construed did enter into our conceptions of material things. To see how, we must attend to his account of material substance.
12
Calvin Normore
The orthodox view of the twentieth century has been that Descartes thought there were many distinct and individual minds but only one physical substance (res extensa). On this view the ontological status of parts of matter, not to speak of sofas and brisket, is a bit obscure. They must be modes of matter; but exactly how that is to be understood is not entirely clear. The competing view is that Descartes believed in many individual material substances just as in many minds. The problem for this view is to defuse a body of text in which Descartes makes clear that bodies are individuated by their motions. Since motion is a mode of matter, if bodies were substances then they would be made what they are by the very modes that depend on them for their being. That would turn Cartesian metaphysics upside down. My own view is that this problem has been misconceived. I think it plausible that Descartes thinks that there are indefinitely many material substances but not plausible that these are what he ordinarily calls ‘bodies’. The individual material substances are chunks of matter. But bodies have criteria of identity through time and often consist of different material substances at different times. The only kind of body whose identity criteria Descartes discusses is the human body, and although the question of the criteria of individuation of human bodies is a tricky one in Descartes, there is no reason to think that he regards the human body as anything social.10 What a human body is is for him a metaphysical matter independent of our purposes or desires. But Descartes sometimes speaks even more broadly of material things. We can see a hint of it, for example, in his comments on Regius’s Broadsheet: “we should note that in subjects which are composed of several substances, one such substance often stands out; and we view this substance in such a way that any of the other substances which we associate with it re nothing but modes of it. Thus a man who is dressed can be regarded as a compound of man and clothes. But with respect to the man his being dressed is merely a mode, although clothes are substances” (CSM I,299 at VIII-2, 351). This passage is puzzling in its suggestion that a substance may, without impropriety apparently, be regarded as a mode of another substance. From a metaphysical point of view though I think we must focus on the language Descartes uses here. First he is talking about a special case, the case of a composite of several substances. Second he does not say that the clothes are a mode of the man but that they may be viewed or regarded this way. This supports, I think, the suggestion that there is in Descartes a tendency to the view that we make objects by projecting criteria of unity. In the passage just quoted he speaks of our making a unity of the man and his clothes. In an interesting passage on love in the Treatise on the Passions of the Soul he speaks of love as “an emotion . . . which impels the soul to join itself willingly to objects that appear to be agreeable to it,” and he glosses the ‘willingly’ thus: “Moreover, in using the word ‘willingly’ I am not speaking of desire, which is a completely separate passion relating to the future. I mean rather the assent by
Burge, Descartes, and Us
13
which we consider ourselves henceforth as joined with what we love in such a manner that we imagine a whole, of which we take ourselves to be only one part, and the thing loved to be the other” (CSM I,356 AT XI 387). It would be tempting to go so far as to suggest that the kind of substantial union the human soul has with the human body involves some of the same factors. But we must be cautious. Descartes insists (notably against Regius) that the human being is a substance, and he never suggests that we make substances by thinking of things as substantial. He only suggests that we can consider or regard them as such. Indeed, in the reply to Regius’s Broadsheet he says that “When he says that the nature of things leaves open the possibility that the same thing is either a substance or a mode what he says is quite self-contradictory, and shows how irrational his mind is” (CSM I, 300 AT VIII-2, 352). So it seems clear that he does not think whether something is a substance or a mode in a strict sense can depend on our (contingent) way of regarding it. But Descartes does not always speak of substances and modes strictly. He often calls ordinary bodies substances, and he speaks of them (in letter 425) as getting their names from the way they appear to us. To the extent that that is a function of our interests we can reasonably suggest that what there is in the Cartesian ontology of everyday life is to some extent a matter of what we put there. Notes 1. Burge (1986), pp. 117–136. All page references are to this article unless otherwise noted. 2. Buridan, Jean, Sophismata, ch. 8 “Sophism” p. 2. See Hughes (1982), p. 34. 3. Buridan’s case also bears some analogies to a reply by Hilary Putnam to the suggestion that I might seriously doubt whether I am a brain in a vat. Putnam agrees that it is possible for me to be a brain in a vat. Perhaps tonight as I sleep some evil scientist will envat me. But I cannot, he suggests, coherently suppose that I am and have always been a brain in a vat. If I had always been a brain in a vat then the concepts I associate with ‘brain’ and ‘vat’ would not have been acquired by contact with brains and vats but (for example) by contact with electrical stimuli produced by the demon scientist’s computer. Thus they would not be the concepts of brain and vat, and so the claim ‘I am a brain in a vat’ would not mean that am I a brain in a vat but (perhaps) that I am an electrical pattern of type X related to another of type Y. Thus I can be sure that the supposition that I am and always have been a brain in a vat is false even if there is nothing impossible about my always having been a brain in a vat and nothing impossible about the sentence ‘I am a brain in a vat’ being true. See Putnam (1981), ch. 1. 4. Kripke (1980). For the discussion of Jonah, see pp. 67–68. For the extension to tigers, see pp. 119–121. For the connection between the concept of pain and the feeling, see pp. 146ff. 5. The antecedent is needed because (as Tyler Burge showed me in conversation) it is far from clear, on the picture described here, that one can even know a priori that one has a concept at all. If there are no objects lying behind one’s use of ‘orca’ or ‘animal’, or if the objects there are don’t fall into any genus, then ‘Orcas are animals’ might well fail to be true. 6. For more on this picture see my “Ockham’s Mental Language” (1990). 7. References to the works of Descartes are to the volumes and pages of C. Adam and P. Tannery, Oeuvres de Descartes, Paris, Vrin, 1964–76 (abbreviated as AT), and to the English translation by John Cottingham, Robert
14
Calvin Normore
Stoothoff, and Dugald Murdoch in The Philosophical Writings of Descartes, 1984–91. The first two volumes of this translation are abbreviated as CSM, the third volume, based on earlier work of Anthony Kenny, as CSMK. 8. This strand in Descartes’s thought—that natures and eternal truths depend on God—is found exclusively but persistently in his correspondence from the time of a letter to Mersenne in May of 1630 (AT. I, 151). 9. The letter has been translated in Dr. Brown’s dissertation “Swampman of la Mancha” and Other Tales about Representation (University of Toronto, 1992). Dr. Brown has been kind enough to give me access to the original text and I have taken the liberty of retranslating certain key passages to make the translation more literal. 10. The central text for Descartes’s account of the identity of human bodies is a letter to Fr. Mesland of Feb. 9, 1645 at AT IV 162ff (CSMK 241ff).
References Brown, Deborah. 1992. “Swampman of La Mancha” and Other Tales about Representation. Doctoral Dissertation, University of Toronto. Burge, Tyler. 1986c. Cartesian Error and the Objectivity of Perception. In Subject, Thought, and Context, McDowell and Pettit (eds.). Oxford: Oxford University Press. Buridan, Jean. 1966. Sophisms on Meaning and Truth, T. K. Scott (trans.). New York: Appleton-Century-Crafts. Descartes, René. 1964–1976. Oeuvres de Descartes. Adam and Tannery (eds.). Paris: CNRS. ——— 1984. The Philosophical Writings of Descartes, vol. II. Cottingham, Stoothoff, Murdoch (trans.). Cambridge: Cambridge University Press. Hughes, G. E. 1982. John Buridan on Self-Reference: Chapter 8 of Buridan’s Sophismata. Cambridge: Cambridge University Press. Kripke, Saul. 1980. Naming and Necessity. Cambridge, MA: Harvard University Press. Normore, Calvin. 1990. Ockham’s Mental Language. In The Historical Foundations of Cognitive Science, J. C. Smith (ed.). Dordrecht: Kluwer. Putnam, Hilary. 1981. Reason, Truth, and History. Cambridge: Cambridge University Press.
2
Anti-Individualism and Scepticism
Barry Stroud I take anti-individualism in this context to be the view that what thought a person is thinking at a particular time depends in part on that person’s present and past relations to what is so in the world around him. I think that idea is immensely important in understanding without confusion or paradox how human thought and experience as we know them are possible. There is no telling, so far, just how rich the consequences of that idea will turn out to be. In this essay I consider some of its apparent implications. What the view implies obviously depends on what it does and does not say, and that in turn depends on exactly how it is formulated and defended. Here I will enter into those crucial questions of formulation only as far as I think I need in order to examine some of the very general epistemological implications of the kind of view of perception, thought, and belief defended by Tyler Burge. I will have in mind as well the very closely related ideas of Donald Davidson. The view that certain relations must hold between a thinker and the world around him in order for anything he does or whatever happens to him to count as his thinking a particular thought finds a parallel in the perhaps more familiar point that certain relations must hold between a perceiver and the world around her in order for what happens to her on a particular occasion to count as her perceiving a particular thing, or her having a perception of a particular kind. In the basic cases—those without which we could not understand people to be having perceptions of certain kinds at all—the perceptions that are attributed to perceivers are perceptions of something that is so in their perceptible environment. This does not mean, of course, that everyone always perceives what is there and only what is there. But even when we misperceive or perceive some object or state of affairs that is not there, our having a perception of just the type we are having on that occasion (which is in fact false or illusory) also depends on our past history and the world in which the very perceptual capacities we are exercising at the moment have been developed and exercised. A different world, or a different past for us, could have meant that we were having different perceptions from what we are having right now, even if all the stimulations at our sensory surfaces remained just what they are at the moment. As Burge puts it, “most perceptual representations . . . represent what, in some complex sense of ‘normally’, they normally stem from and are applied to” (1986, p. 131). It is not easy to state in simple and adequate terms exactly what this implies about the reliability of human perception. But at the very least I think we can see that there is, and is meant to be, a reassuring point here. It is not that none of us is ever wrong or ever misperceives anything. It is not even that no one person could be more wrong than right in perception over a long period of time. But the view does require that for someone to
16
Barry Stroud
misperceive, or to perceive things as they are not, it must be possible for him or others to perceive things as they are. Things’ being the way they are perceived as being is in that sense a condition of there being perceptions of things being one way rather than another at all. Without enough “normal,” that is, veridical, perceptions of certain kinds, no one could be understood to have any perceptions of those or any other kinds, and so no one could even be understood to have nonveridical or illusory perceptions. There is a certain reassurance in the thought that if we ask “Could all or most of the perceptions we actually have misrepresent or mislead us about what they in fact stem from?” we must on antiindividualist grounds answer “No, they could not.” Burge puts this reassurance at one point by saying that in the case of perception “we are nearly immune from error in asserting the existence of our perceptual kinds, and of other kinds that are taught by more or less immediate association with perceptually based applications” (1986). There is a question here of how “nearly immune from error” we are, and even what different degrees of immunity amount to. These are some of the questions of adequate formulation I mentioned at the beginning. The precise character and measure of this immunity do not matter at this point. I think we get the idea, and it is enough to be going on with. Let us say that, given the conditions of identification of perceptual experiences as being of this or that type, or as having this or that content, we could not be largely wrong in perception. Our perceptions are in large part perceptions of things as they are. Even with the scope of possible error left vague in this way, we can go on to ask about the status of this conclusion and about its implications for certain varieties of scepticism about the senses and about the empirical world. Is the reassurance this view provides an adequate reassurance against scepticism about perception? That will depend on what scepticism is, and on what it takes to oppose it or to provide reassurance against it. But if this kind of view does have antisceptical consequences in the case of perception, whatever they might be, can those consequences be carried over directly, or at least developed along similar lines, to apply to belief and knowledge in general, and so to oppose epistemological scepticism everywhere? I am not sure what Burge’s answers to these questions would be. He certainly seems to accept what I have just said about perception. He thinks the anti-individualist view in that case provides what he calls a “qualified basis for the oft-repeated slogan that error presupposes a background of veridicality” (1986, pp. 130–131). He also draws a parallel between the external conditions of perceptual experiences of certain kinds and the conditions of thoughts and beliefs more generally. In fact he thinks the point about perception can help explain and make plausible his anti-individualist theory of thought; it can bring out the way in which entertaining a certain thought also has its conditions in the wider world. He holds that in both cases those external conditions’ holding is not something the perceiver or the thinker himself must know (or at least must know empirically) in order
Anti-Individualism and Scepticism
17
to have the perception or thought in question, or even to know that he has it. This last point is the key to his account of self-knowledge in general, or knowledge of one’s own psychological states. Despite this parallel, and the fact that he thinks we are “nearly immune from error” in asserting the existence of the things we perceive, Burge thinks, as he puts it, that “we are not immune from fairly dramatic and wholesale error in characterizing the nature of the empirical world” (1986, p. 131). He also thinks the “oft-repeated slogan that error presupposes a background of veridicality” is “sometimes misused” (ibid.). Since he appears to contrast the scope of the possibility of “fairly dramatic and wholesale error” about the empirical world with our near immunity from error in the case of perception, this might suggest that he thinks that slogan is “misused” in trying to show that we cannot be largely wrong in our beliefs about the world in general. He is inclined to believe that Quine and Davidson in particular “sometimes use this important idea [as expressed in the slogan] with insufficient discrimination” (ibid., p. 131n). Burge uses that same idea himself to show that we are “nearly immune from error” in the case of perception. Does he think that Quine and Davidson are indiscriminate in applying it more generally to all our beliefs about everything, even well beyond the limits of perception? Or is there some other way in which they are not sufficiently discriminating in their use of it? Davidson does conclude from what looks like the same general conception of the identification of thoughts and beliefs, as well as perceptions, that, as he puts it, “most of our beliefs are true,” even that “most of our beliefs must be true” (Davidson 1983, pp. 435, 431). He thinks the anti-individualism that guarantees the truth of most of our beliefs “serves to rescue us from a standard form of skepticism” (1983, p. 438) or, more cautiously, that it has within it “the makings of a cogent argument against some forms of skepticism” (1988, pp. 664–665). Perhaps Burge thinks that drawing that optimistic-sounding conclusion directly in that way is an insufficiently discriminating use of the important idea. He himself holds, to the contrary, that “there is no easy argument against skepticism from anti-individualism and authoritative self-knowledge” (Burge 1988, p. 655). What, then, are we to make of Burge’s own reassuring-sounding conclusion about perception—that we are “nearly immune from error” in that case? If that counts as an appropriately careful and discriminating use of the idea that error presupposes veridicality, does it not rescue us or save us from some forms of scepticism, at least about the senses or the things we perceive? If it is not possible for all or most of the perceptions we understand ourselves to have to misrepresent or mislead us about what they in fact stem from, is that not an argument against scepticism? I think (I’m not sure about this) that Burge’s answer would be “No”—that the general consideration, as it stands, is not an argument, or at least not an easy argument, against
18
Barry Stroud
scepticism. Why not? I do not think he means that although scepticism about our perceptual beliefs has been ruled out, scepticism about beliefs that go beyond what we can perceive remains a possibility. For one thing, that would seem to require some sharp and lasting and therefore dubious distinction between what we can perceive to be so and what we believe or hypothesize or infer to be so on the basis of what we perceive. It is dubious because such a distinction seems on the contrary precisely to encourage scepticism. It is accordingly more realistic to allow that the limits of what we can perceive can be gradually moved further and further outward as the sophistication of our knowledge and of our perceptual capacities increases, so that there would eventually remain virtually nothing that a sufficiently informed and perceptive person could not be said to perceive to be so under some circumstances or other. That would mean that whatever immunity from error we enjoy in perception is at least theoretically extendable to everything, or virtually everything, we believe about the world. Second, the traditional source of scepticism about our knowledge of the world in general, at least in the so-called modern era, has been the threat of scepticism about perception. If skepticism had somehow been blocked there, or had not even got off the ground, it is not easy to see what would be left of scepticism about our knowledge of the world in general. I think Burge must mean that the reassuring conclusion about the reliability of perception does not itself amount to an argument, or at least an easy argument, against scepticism, even with respect to our perceptual beliefs. Scepticism in this context is presumably the view that we do not know anything about the world around us, or that we have no reasons to believe the things we believe about it. If an argument “against scepticism” is an argument for the negations of that doctrine, it would have to show that we do know or have reason to believe the things we believe. If that is what Burge means by an argument against scepticism, then I agree that what I have been calling the reassurance provided by anti-individualism does not amount to such an argument. For one thing, even if it does imply that most of our beliefs about the world are true, or by and large true, it does not follow that those true beliefs amount to knowledge. That is a point Davidson concedes, although he continues to hold that anti-individualism nevertheless “serves to rescue us from a standard form of scepticism” (Davidson 1983, p. 438). I think the distinction is important. Providing the negation of scepticism and so knowing that it is false is not the only way to be saved from it. When Burge says “there is no easy argument against skepticism from antiindividualism and authoritative self-knowledge,” he is speaking in part about what he calls “transcendental” responses, in particular, Hilary Putnam’s attempt in the first chapter of Reason, Truth, and History to show that, given an anti-individualist account of meaning and belief, even a brain in a vat could not be largely wrong about what is happening in its environment. Burge thinks those considerations “do not do much to undermine
Anti-Individualism and Scepticism
19
skepticism” (1988, p. 655n). If he means that they do not show that we know that we are not brains in a vat and so know what is happening now in our environment, I agree. I do not think a valid argument can be found which a person might deploy, starting from antiindividualist premises, to take him to the conclusion that he knows he is not a brain in a vat, or that he knows most of the things he knows about the world, in a way that refutes philosophical scepticism. But again the question is whether anti-individualism must be shown to have that implication in order to have antisceptical consequences. Burge thinks that what he calls “transcendental” responses would at most provide only “ ‘general’ guarantees against skepticism” (1988, p. 655). Perhaps that is why he thinks they do not amount to a good argument against it. He does not say what a “general guarantee against skepticism” is or would be, or why it would not be just what we want. It is pretty clearly not what he wants. He contrasts transcendental responses with what he regards as a more promising line, which he says would actually “justify particular perceptual knowledge claims in the face of skepticism” (pp. 655–656). This suggests that a completely general anti-individualist reassurance about the reliability of perception would not be enough to counter scepticism as Burge wants to counter it. I have agreed that the reassuring general point does not imply that scepticism is false—that we do know or have reason to believe all or most of the things we think we know. But it appears that Burge would like to prove, for some of the particular things that philosophical scepticism would say we don’t know, that we do or can know them after all. The merely general reassuring-sounding conclusion implies nothing about any particular case. If Burge does want to prove, in particular cases, that we know—if that is for him a condition of success for an argument against scepticism—that would explain why he thinks “there is no easy argument against skepticism from anti-individualism and authoritative self-knowledge.” There is no easy argument from those premises that can be used to prove, in a particular case, that we know in that case. He hints that a non-easy, or at least a more complicated, argument to that effect could eventually be given. He thinks it would show that we can know, presumably in a particular case, that no demon is fooling us, or that we are not brains in a vat. We could know that, he suggests, “by inferring it from our perceptual knowledge” (Burge 1988, p. 655n). This suggestion is made briefly, and in a footnote, and is not developed further. He says there that “this is a complicated matter best reserved for other occasions” (ibid.). Well, this is another occasion. As I understand it, the argument would show, of particular perceptual knowledge claims made on particular occasions, that they are true or justified. That is what would make it more than a “general guarantee.” But it would have to reach that conclusion while remaining consistent with another feature of perception that Burge stresses, namely, that “in any given case, all of a person’s perceptual capacities . . . could in principle be mistaken about
20
Barry Stroud
the empirically perceivable property (object, relation) being perceptually presented” (Burge 1986, p. 125). This, he holds, is a consequence of the fact that our perceptual experience as we understand it is of things or relations or properties that are “objective”—they are as they are independently of anyone’s thoughts or experiences of them, or even of whether there are any mental phenomena at all. An anti-individualist or externalist theory of perception would presumably imply that, even in the face of that general possibility, if things are in fact a certain way on a particular occasion, and the person’s perceiving them to be that way is connected in the right way with their being that way, then the person is perceiving things as they are on that occasion, and in that way thereby knows that they are that way. If all of that were so, it seems, then given anti-individualism, scepticism would be false. The person would know. But that is still a completely general, or only conditional, claim. The stronger or non-easy complex argument that Burge envisages would presumably enable the person to prove, at that time, that what she perceives to be so at that time is so, and that she therefore does know the truth of what she perceives to be so. And she would have to prove it in that particular case even though in any particular case a perceiver can be wrong about how things are even though she perceives them to be that way. I agree that that would be no easy argument. This is a fascinating suggestion that I hope Burge will pursue. I am still far from certain that it is really what he has in mind. Rather than speculate about it further, I would like to raise a prior question. Do we have to prove such a thing, or in any other way prove that scepticism is false, in order to oppose it on antiindividualist grounds? Is the general reassurance provided by anti-individualism not enough in itself to block any serious sceptical threat? At the risk of being thought too friendly to so-called “transcendental” responses, this is the question I would like to explore. Davidson holds on anti-individualist grounds that most of our beliefs are true, or that they are by and large or for the most part true. The anti-individualist grounds he relies on are to be found in the conditions of what he calls interpretation—one person’s understanding and communicating with another—which involves the attribution to that other of thoughts, perceptions, beliefs, and other propositional attitudes. We can do that only by connecting the attitudes we attribute to people with circumstances or states of affairs in the world in which we as interpreters and they as interpreted interact. This inevitably produces a large measure of agreement among those who can understand and communicate with one another, according to Davidson. Because an interpreter relies on what he himself believes to be true of the world, he will find those he interprets to have largely true beliefs. Burge’s anti-individualism, even about perception alone, is also a theory of our practices of attributing psychological states and attitudes with determinate contents to people in the world we inhabit. The thought experiments he appeals to to support his anti-
Anti-Individualism and Scepticism
21
individualism turn on and exploit our capacities of belief- and attitude-attribution as they actually are. We are convinced by his “experiments,” if we are, because we find that in the situations he describes his subjects to be in we think it correct to attribute to them the attitudes that his anti-individualist theory implies that they have. What we are told is so in the world surrounding the subject helps determine what attitude we correctly ascribe to him, whether he knows it is so or not. If the thought experiments succeed, they show us that our attitude-ascribing practices, and so our understanding of perception, thought, and other mental attitudes, are anti-individualistic. The world and our thoughts about the world cannot come completely apart. In that sense we understand that we are “nearly immune from error” in our perceptual beliefs. To declare, as a consequence of this, that most of our beliefs must be true, could be just a way of announcing, on anti-individualist grounds, what one holds to be true of one’s fellow human beings. That is one way of understanding Davidson’s conclusion. “Given that they have got the particular beliefs I have ascribed to them,” he could be saying, “those beliefs must be by and large or for the most part true.” That too could be the kind of “immunity from error” that Burge thinks we must enjoy in perception. We could not understand people to perceive this, that, and the other as we do without acknowledging that those perceptions we ascribe to them are by and large veridical. That is a condition of perception-attribution on the anti-individualist theory. The people we understand to have perceptions of this or that kind do not just happen to have avoided error for the most part in their perceptual beliefs; they are largely “immune from error.” We find that they could not be largely wrong, given that they have got the kinds of perceptions we ascribe to them. But saying that most of our beliefs must be true, or that people couldn’t be largely in error in their perceptual beliefs, can also sound like a stronger thesis to the effect that beliefs and perceptions simply must be largely true or veridical—that that is a condition of anyone’s having beliefs and perceptions at all. This is certainly suggested by Davidson’s saying “belief is in its nature veridical” (1983, p. 432). That makes it sound as if beliefs simply could not be false, that no reasonably rich set of beliefs could fail to be largely true, that if there are beliefs at all, they must be for the most part true. That thesis, it seems to me, would threaten the objectivity of what we believe to be so—the idea that the truth or falsity of what is believed to be so is in general1 independent of its being believed to be so by this or that person or group, or even by all human beings universally. The stronger thesis could allow that such independence holds for each particular belief, or perhaps even for each relatively small subclass of beliefs, taken on its own. But it appears to rule out any such independence for all or most of a large set of beliefs taken as a whole. The stronger view seems to imply that the world would have to be by and large the way that any group of believers with a sufficiently rich and comprehensive set of beliefs believed it to be. Not just because human beings can reasonably be expected to
22
Barry Stroud
get things more or less right, but simply because that is what beliefs are. They are, in their “nature,” veridical. That stronger thesis, if it were correct, would be, or would provide, a foolproof antidote to scepticism, it seems to me, even though it would remain, in Burge’s terms, completely “general.” It does not directly imply the negation of scepticism—that we know or have reason to believe the truth of those things we believe—but anyone who came to realize that “belief is in its nature veridical” in the strong sense could easily conclude that he, or at any rate human beings generally, could not go for the most part wrong in believing whatever they believe. This kind of reassurance is what I think Davidson had in mind when he argued in “A Coherence Theory of Truth and Knowledge” that anyone who came on reflection to see that “belief is in its nature veridical” would have a reason to believe that most of our beliefs are true, and thus the threat of pervasive failure or defect in the grounds of our beliefs would then have been defused. We could see by reflection that if what we seek is the truth, we cannot fail, at least for the most part. This is very reassuring news. But like much good news, this strategy also has its negative side. As the independence or objectivity of what is believed fades, believing what is for the most part true becomes less and less of an achievement. It is something in which, on this stronger view, we could not possibly fail. Such are the embarrassments of a coherence theory of truth, and the perennial disappointments of idealism. I do not think the stronger thesis that “belief is in its nature veridical” follows from anti-individualism about thought, perception, and belief. Perhaps the suggestions in some of Davidson’s formulations that he is taking the phrase in that stronger sense are what inclines Burge to believe that Davidson sometimes uses the slogan that “error presupposes a background of veridicality” with “insufficient discrimination.” I do not know whether that is what he has in mind, but I do think it would be a misuse or an insufficiently discriminating use of that idea to derive from the slogan what I have been calling the stronger thesis.2 Anti-individualism is supported by our practices of attributing thoughts and attitudes with determinate contents to people in our world. It is a condition of our ascribing the thoughts we do that we understand them to be related in appropriate ways to the objects or states of affairs in the surrounding world we take them to be about. Without our own knowledge of and engagement with the world we could make no sense of anyone’s having thoughts or experiences of this or that kind at all. Human beings understand the mind antiindividualistically. They can identify the contents of minds only in terms of what they also take to be true of the independent world. But none of that implies that what people believe about the world is in fact true, or that the beliefs they inevitably ascribe to their fellow human beings are in fact for the most part true either.
Anti-Individualism and Scepticism
23
This is not a sceptical point. I do not mean to be casting doubt on any of our beliefs about the world. The point is only that, however overwhelmingly unlikely, even in some sense unthinkable, it might be that we are largely or even slightly wrong in our beliefs about the world, our not being wrong, or our having for the most part beliefs that are true, does not strictly follow from the conjunction of the fact, first, that we have a great many beliefs about the world, second, that we understand our fellow human beings to have a great many determinate beliefs with this or that content, and third, that we do or even must ascribe those beliefs on anti-individualist grounds. If those three things are true, no one could consistently find that his fellow human beings believe this, that, and the other about the world, but that those beliefs are all or for the most part false. The conditions of his ascribing those beliefs to them rules out that possibility. But it does not follow that the beliefs in question, or therefore in general any beliefs that any people could be understood as having, are in fact largely true. The stronger thesis about belief and truth is not supported by what supports anti-individualism. Anyone who thinks that people do have certain determinate beliefs, and who thinks so on anti-individualist grounds, will of course regard those beliefs as for the most part true. She will not see the people she interprets as holding the beliefs she says they have got unless she also sees them as having mostly true beliefs. But that is not surprising, if she sees those others as sharing her beliefs to a large extent. Every believer regards her own beliefs as true. That they are true, even if they are, is not something that follows from that alone. It does follow from what such a person believes—not from her believing it—that the beliefs she shares with those others are true. She believes many things about the world, and she believes that others also believe many of those things she believes. So from the totality of everything she believes it follows that most of those others’ beliefs are true. But that is not a special consequence of anti-individualism. It is a consequence of the fact that from any proposition p, and the proposition that someone believes that p, it follows that that person has a true belief. If you write down everything a belief-attributor believes, it will follow from everything you write down that the beliefs she attributes are for the most part true. There is then a weaker sense in which “most of our beliefs are true” and “error presupposes a background of veridicality” are legitimately supported by what supports antiindividualism. It says: we can understand ourselves to have false beliefs or illusory perceptions only if we understand ourselves to have a set of beliefs or perceptions that are largely or for the most part true or veridical. This still leaves unspecified the exact scope and quality of possible error in ways I mentioned earlier. But it retains the idea of a kind of reassurance provided by anti-individualism. To say not merely that we are for the most part not in error in our perceptual beliefs, but that we are “nearly immune from error” in those beliefs, would then be to say that we cannot consistently find ourselves to have
24
Barry Stroud
certain determinate perceptual beliefs without also finding ourselves not to be largely in error about them. Finding them largely erroneous would be sufficient for finding those beliefs not to be present after all. The idea that “belief is in its nature veridical” can accordingly be taken in the sense that, as we might put it, belief attribution is in its nature truthascribing (for the most part), or that attributed beliefs are necessarily regarded as, for the most part, true. Scepticism is the outcome of an effort to examine our beliefs about the world all at once. The line of thinking that eventually leads to it starts from what looks like the uncontroversial observation that all our beliefs could be false, consistently with our holding them. This so far is only what might be called a logical point. Much more is needed to get to a sceptical conclusion, but this is where the reasoning that typically leads to it starts. We are then challenged to say how we know that that admitted possibility is not actual— that it is not the case that our beliefs are not true even though we all believe that they are. The rest of the reasoning attempts to show, in one way or another depending on the case, why the challenge cannot be met. What I have called the stronger thesis about belief and truth would imply that this very first step is wrong. On that view, there is simply no possibility that a large and reasonably comprehensive set of beliefs is entirely or for the most part false. What the sceptical reasoning would start from is therefore on this view actually a contradiction. There could be no serious question of how we know that that alleged possibility is not actual. It is not a possibility at all, so there is nothing that we have to show is not actual. I have said that this view, if it were true, would still not refute scepticism in the sense of implying its negation. But it would decisively stop the typical sceptical challenge from getting off the ground. But it seems to me to deny too much, and so to be too strong, and too quick. I do not think it follows from anti-individualist premises about belief attribution. I think we must grant the completely general and abstract point that the truth of all or most or even any of a set of beliefs does not follow simply from their being held. Given that they are all believed, it is still in that sense possible for them to be false. That is the logical point. But what the slogan that “error presupposes a background of veridicality,” or the idea that “we are nearly immune from error” in our perceptual beliefs, do imply, in the sense in which I think they are supported by anti-individualism, is that anyone who understands people to have a set of determinate beliefs must take those beliefs to be for the most part true. She could not say “these are the beliefs they hold, and they are completely or for the most part in error.” So the abstract possibility from which the sceptical reasoning typically starts could pose no serious threat as applied to any particular set of beliefs anyone might consider. Not because it is in no sense a possibility at all that those beliefs are false, but because, even though it is in the weakest sense a possibility, it is not
Anti-Individualism and Scepticism
25
a possibility anyone could ever consistently find to be actual when specific, determinate beliefs are under consideration. The epistemological investigation of human knowledge, of which scepticism is one possible outcome, involves scrutiny and assessment of the particular beliefs we human beings have actually got. The possibility we are asked to consider at the beginning is a possibility with two parts, or conjuncts. It is the possibility that we have all the beliefs we now have in this, that, and the other aspect of the world, and that those beliefs are all or for the most part false. Anti-individualism with respect to our thoughts about and attributions of beliefs means that we could not consistently find the first half of the possibility to be realized without finding its second half not to be so. And if we found the second conjunct to be true, we could not consistently find the first half of the possibility to be realized. Attribution of the beliefs we attribute requires finding them for the most part true; finding a certain set of propositions for the most part false rules out assigning them as contents of the beliefs of people with whom one shares a common world. Our position with respect to that original possibility is therefore similar to a person’s relation to the apparently paradoxical sentence ‘I believe that it is raining, and it is not raining’. That is not something one could consistently believe or assert—but not because what it says is something that could not possibly be true. It is possible that I believe that it is raining when it is not raining. That is a possibility with two parts, or conjuncts. The first does not imply that the second is false, and the second does not imply that the first is false. It is in that sense a genuine possibility. But no one can consistently hold in his own case that that possibility is actual, that both conjuncts are true. This brings out the difference between something’s being simply inconsistent or impossible (which my believing that it’s raining, and its not raining, is not) and something’s being impossible for anyone consistently to believe or discover. I am suggesting that this distinction applies to the possibility that human beings have all the beliefs they now have, and that those beliefs are all or for the most part false. It is the impossibility of consistently finding or believing that possibility to be actual that I think follows from anti-individualism about our thoughts, perceptions, and beliefs. If that is so, then we cannot take seriously the possibility that the beliefs we take ourselves to have are for the most part not true. There is no serious challenge to our beliefs about the world expressed in the question “Given that the truth of your beliefs and the veridicality of your perceptions do not follow simply from your having them, how do you know that, taken all together, they are not all or for the most part false or illusory?” That is the question that the traditional epistemological challenge presses. Failure to answer it satisfactorily at any point leaves us eventually with scepticism. Applied to a particular case, we find in various ways that we cannot appeal to anything we already believe or to
26
Barry Stroud
any perceptions we might get or seek in order to answer the question, since at every step all of our beliefs and perceptions of the world are in question all at once. But given an anti-individualist understanding of beliefs and perceptions, for any specific set of beliefs and perceptions we take ourselves to have, the question cannot really present the serious challenge it might otherwise appear to lead to. Not because the challenge can be met by establishing independently—somehow without relying on any perceptions or beliefs—that our perceptions and beliefs are veridical or true and amount to knowledge of the world. Not because we could not possibly fail to be largely right in whatever set of perceptions and beliefs we happened to have. But because our considering the particular set of perceptions and beliefs that we are asking about in itself guarantees that we find them to be for the most part true. Does it follow from that that we know those beliefs are true? No, it does not. Does it follow from that that the beliefs in question are even largely true, whether we know it or not? Again, I think it does not. We must always admit that our beliefs could be false. Asked right now about the things we believe, we will of course say they are true. Asked whether we know them to be true, we will say, of many of them, perhaps on reflection, that Yes, we do know them to be true. But we are fallible human beings with beliefs about an independent world, and it is always an open question whether what we think we know is really something that we know, or is even something true. What I have described as the reassurance derived from anti-individualism does not conflict with that. Although I cannot consistently believe both that I believe that it is raining, and that it is not raining, I can still ask myself whether it is raining or not, or whether I am right in believing that it is raining. So, in the more general case, I can still ask, “Are those beliefs that I think people have true, or for the most part true, or not?” And just as I answer the first question by going outside and finding out whether it is raining, so the way to answer the more general question is to put myself in the best position I can for finding out whether the things I and others believe about the world are true. If I have not already done that, or even if I have, I might find on further investigation that some or even many of the things I believe are not true. I will then abandon those beliefs. But I might equally find that all or a great many of them are indeed true, and in fact so well supported that I do not hesitate to say that I know they are true. The open question has been, to the best of my knowledge and at least for now, answered. If I answered the question in that way, would philosophical scepticism about our knowledge of the world have been refuted? I would say “No.” Would what I say nevertheless conflict with scepticism, even if it does not refute it? This question is harder to answer. Would we still face a challenge to our alleged knowledge starting from the thought that all or most of our beliefs and perceptions could be false, and so demanding some reason to think that that possibility is not actual? Here I am inclined to answer, on anti-
Anti-Individualism and Scepticism
27
individualist grounds, “No, that is not a possibility we can take seriously; it is not a way we could find things to be.” Even if that is right, and reassuring, it does not amount to the negation of philosophical scepticism. But still I say, for the very best reasons we can find, that we know a great many things about the world around us. Philosophical scepticism is a negative outcome of an effort to assess all of our knowledge of the world all at once, and perforce from a position in some sense outside it. The denial or negation of that assessment, although perhaps reassuring if reachable, would presumably have to be made from that same outside-of-all-of-our-knowledge-of-the-world position. That is why I think we should not aspire to deny the thesis of scepticism, any more than we want to accept it. The great promise of anti-individualism as I understand it is that it would reveal how and why it is impossible for us even to get into that position with respect to any comprehensive set of beliefs or experiences we can recognize ourselves to have. We understand ourselves to have beliefs only by for the most part endorsing them. We could then consistently arrive at neither the assertion nor the denial of philosophical scepticism. If we could accept the fact that our beliefs about the world are none the worse for all that, perhaps final satisfaction would be at hand. Notes 1. I say “in general” because there are some things that must be so if anyone believes anything, e.g., that someone believes something. But for what does not fall into that special class of beliefs, the truth or falsity of what is believed is independent of its being believed. 2. I have argued this at greater length in application to Davidson’s theory in Hahn (1999). For its connection with the “transcendental” strategy once pursued by P. F. Strawson, see my “Kantian Argument, Conceptual Capacities, and Invulnerability” (Parrini 1994).
References Burge, Tyler. 1986. Cartesian Error and the Objectivity of Perception. In Subject, Thought, and Context. McDowell and Pettit (eds.). Oxford: Oxford University Press. ———. 1988. Individualism and Self-Knowledge. Journal of Philosophy 85(11): 649–663. Davidson, Donald. 1983. A Coherence Theory of Truth and Knowledge. In Kant oder Hegel? D. Henrich (ed.). Stuttgart: Khitt-Cotta. Also in (1986), Truth and Interpretation, Lepore (ed.), Oxford: Blackwell. ———. 1988. Reply to Burge. Journal of Philosophy 85: 664–666. Hahn, Lewis E. (ed.). 1999. The Philosophy of Donald Davidson. Chicago: Open Court. Parrini, P. (ed.). 1994. Kant and Contemporary Epistemology. Dordrecht: Kluwer. Stroud, Barry. 1994. Kantian Argument, Conceptual Capacities, and Invulnerability. In Parrini (1994).
3
When Swampmen Get Arthritis: “Externalism” in Burge and Davidson
Martin Hahn Introduction Externalism is not a single thesis. Even restricting ourselves to theses concerning the contents of mental states, the views generally called “externalist” do not form a particularly interestingly interconnected family of theses. Or so I shall argue. The locus of this argument will be the relationship between two “externalists” who are sometimes regarded as propounding very similar views, Tyler Burge and Donald Davidson. On closer examination, we shall see, the numerous agreements between Burge and Davidson either have little do with their “externalism,” or, when they do, are merely superficial. The details reveal instead that at the core of the two “externalisms” lie disagreements of the most fundamental kind. I begin by citing some passages from Burge and Davidson that make it seem that there is, indeed, a core of externalist agreement between them. Agreements, Broad and Narrower In seeking the common ground, we can start with Burge’s general definition of antiindividualism, which Davidson cites approvingly (Davidson 1986, p. 450): (1B) . . . the intentional content of ordinary propositional attitudes . . . cannot be accounted for in terms of physical, phenomenal, causal-functional, computational, or syntactic states or processes that are specified nonintentionally and are defined purely on the in individual in isolation from his physical and social environment. (Burge 1992, p. 288)
As we proceed, we shall see that this very general agreement does not amount to very much at all, for much depends on what one means by the exceedingly vague “accounted for,” how one conceives of the connections between the environment and the individual, nonintentionally specified, and what consequences one draws from these contentions. But as we dwell on some of these details, further prima facie agreements between Davidson and Burge emerge. Davidson and many other philosophers divide externalism into two varieties, physical and social, depending on the sort of factors that are claimed to figure in the individuation of mental states. Davidson further identifies the former kind of externalism, at least in his version, with what he calls “perceptual” externalism. He takes “perceptual externalism” to be central to both our folk-psychological conception of thought and to his theory. There are several striking points of apparent agreement with Burge on issues that Davidson takes to be at the crux of perceptual externalism. Whenever externalism is discussed, one of Davidson’s starting points is the recurrent theme that “in the plainest and
30
Martin Hahn
methodologically most basic cases,” the meanings (and beliefs) an interpreter will attribute will be constituted by the objects that are the normal causes of the utterances in question: “the objects of a belief [will be] the causes of that belief” (1986, pp. 317–318).1 To take another statement of this oft-repeated view: (1D) . . . in the simplest cases the events and objects that cause a belief also determine the contents of that belief. Thus the belief that is differentially and under normal conditions caused by the evident presence of something yellow, one’s mother, or a tomato is the belief that something yellow, one’s mother, or a tomato is present. . . . [T]he causal history of such judgments provides a major constitutive feature of their contents. (Davidson 1988, p. 195)
As Davidson himself notes, the view that in certain cases the contents of beliefs are determined, in part, by their normal causes is shared by Burge: (2B) . . . some visual presentations that represent objective entities as such must have the representational characteristics that they have partly because instances regularly enter into certain relations with these objective entities. Their carrying information, their having objective intentional content, consists partly in their being the normal causal products of objective entities. . . . That is why we individuate intentional visual representations in terms of the objective entities they normally apply to. . . . (Burge 1986a, pp. 40–41)
Another striking point of agreement is on the general view that in order for a subject to be credited with intentional states, that is, psychological states with content, the subject must be able to recognize and react differentially to objects in her environment. If there are any contents at all, some of them must be applied to objects: (3B) . . . if the subject is to be credited with having propositional attitudes he must indicate some ability to correlate his thoughts with objects those thoughts are thoughts of. Failing evidence of the ability to recognize such correlations, there is no adequate ground for attributing understanding of sentences or propositional attitudes. (Burge 1977, p. 348) (2D) The dispositions to react differentially to objects and events . . . are central to the correct interpretation of a person’s thought and speech. If this were not the case, we would have no way of discovering what others think, or what they mean by their words. (Davidson 1987, p. 450)2
We should note that the Burge quote predates his anti-individualist writings and is, furthermore, not a prefiguring of them but a distinct thesis. One might well hold, with Burge, that in order to have any thoughts, at least some of person’s thoughts have to be de re, while maintaining that such a dependence does not affect the way de dicto thoughts are individuated at all. A final point of seeming agreement concerning “perceptual externalism” is the thesis, already stated in (1D) above and repeated in several other places, that the contents of some thoughts depend on the causal history of the thinker. This is the part of Burge’s and Hilary Putnam’s view that Davidson takes to be largely right:
When Swampmen Get Arthritis
31
(4D) The point is that the correct interpretation of what the speaker means is not determined solely by what is in the head; it depends on the natural history of what is in the head. (Davidson 1989, p. 164)
Although this sounds like Burge’s view to Davidson, and indeed Burge does in some sense agree, he seldom articulates his anti-individualism in terms of the “natural history” of the thinker. The closest we will come to are certain passages in Burge (1981) explaining why it is that Twin Eartheans have no water thoughts and cognate passage elsewhere. For example: (4B) [Ali] lacks attitudes that can be correctly (truly) described with ‘aluminum’ in oblique position. . . . Ali has never had contact with aluminum . . . ; and no one in his community use a word that means what ‘aluminum’ means in English. (Burge 1982b, p. 285)
So much for what seem to me to be the best bets at straight-out agreement between Davidson and Burge on perceptual or physical externalism. Davidson sees himself as holding essentially the same position as Burge on these issues. Their differences, according to him, have to do with the efficacy of (and need for) Burge’s (or, more frequently, Putnam’s) Twin Earth arguments. We shall return to Davidson’s claim that no such complex arguments are needed below. I will also endeavor to cast doubt on the contention that Davidson’s and Burge’s views are anywhere near as close as Davidson and the carefully chosen quotations suggest. Davidson and Burge are often taken to disagree concerning social externalism, a thesis from which Davidson wishes to distance himself (1987, pp. 448–450; 1991, pp. 197–198). But, somewhat surprisingly perhaps, the evidence of at least broad agreement is as solid here as it is in the case of physical externalism, if not more so. Consider: (5D) . . . before anyone can have thoughts there must be another creature (one or more) interacting with the speaker. . . . The interaction must be made available to the interacting creatures. . . . For this to work, it is clear that [their] innate similarity responses . . . must be much alike. . . . A condition for being a speaker or interpreter is that there must be others enough like oneself. (Davidson 1989, pp. 198–199)
and (5B) In learning words, individuals normally look to others to set standards for determining the range of legitimate examples and the sort of background information used in explicating a word or concept. I believe this is a psychological necessity for human beings. (Burge 1989, p. 186) Although our kind-making abilities may differ in some instances, it reasonable to expect that they will be typically the same, especially with concepts that apply to entities of common perceptual experience. (Ibid., p. 185)
32
Martin Hahn
In other words, Burge and Davidson, broadly speaking, agree that in order to have the thoughts and linguistic meanings we have, we need other people whose perceptual and concept-forming abilities and dispositions are innately substantially similar to ours. As we shall note further, both regard this to be the case in spite of their further agreement that the proper subject of the study of semantics is the individual and her psychology. More agreement can be found, again pace Davidson’s assessments, concerning some of the consequences of an externalist view. Davidson makes a concerted effort (e.g., 1987, pp. 450–455) to show that his externalism does not have the consequence that thoughts are not “in the head” in some sense that would put the scientific investigation of their physical realization in doubt: (6D) . . . it has been argued [that] theories that identify mental states and events with physical states of and events in the body must be wrong. . . . [This] is explicitly claimed by Tyler Burge. . . . The argument assumes that if a state or event is identified . . . by reference to things outside of the body, then the state or event itself must be outside the body, or at least not identical with any event in the body. This is simply a mistake: one might as well argue that a sunburned patch of skin is not located on the body of the person who is sunburned. . . . (1989, p. 167)
But, although Burge does claim that his anti-individualism undermines certain kinds of identity theories—even token-identity theories such as functionalism—he agrees with Davidson about what would be a mistake: (6B) The failure of supervenience in no way casts doubt on investigations of neural or biological realizations of mental structures. The identity of a heart depends on its function in the whole body, on its relations to parts of the body outside the heart. In a crudely analogous way, the identities of some mental kinds depend on those kind’s relations to entities beyond the individual’s body. (1989, p. 178)
There are other agreements concerning the consequences of externalism as well: Firstperson authority is, according to both philosophers, compatible with the “externalism,” and for similar reasons (although, once again, Davidson is concerned to deny this). Both philosophers also agree that there is a sense in which all of our thoughts about the external world couldn’t be false if a certain form of externalism is true (though the consequences of this for the traditional problem of skepticism are a point of sharp disagreement). In what follows, I will not have the space to follow up on all of these points but will, instead, concentrate on what I see as the crux of the difference between Davidsonian and Burgean externalisms. Burge’s Theses Over an almost ten-year period, Burge presents at least four different versions of antiindividualism, supported by four different Twin Earth thought experiments as well as other
When Swampmen Get Arthritis
33
arguments. The anti-individualist theses are all different but complementary and interwoven into a complex view about thought, language, psychology, intellectual norms, and many other issues. For the present purposes, only the briefest of sketches of the views— presented as a rational reconstruction rather than chronologically—will be possible. Burge’s central idea is that there is a class of propositional attitudes with nonindexical contents that have the crucial feature that their content is, in part, individuated by the nature of their referents. Such contents are neither esoteric nor in the minority. They use concepts like dog, running, gold, cloud, insurance, edge, bathtub, and so (not very obviously) on. In fact, the class likely includes every concept with empirical application whatsoever, ranging from ones that figure in perceptual beliefs to highly theoretical ones in the natural sciences. For the perceptual ones, Burge argues, their identity depends, in part, on what normally causes the perceptual beliefs in which they figure (1986a,b). Twin Earth thought experiments are offered in which the internal, nonintentionally described states of two subjects are identical but the contents of their perceptual beliefs differ because Earth and Twin Earth differ in such a way that the two perceptions are not perceptions of the same kind of item: The causal laws or other background conditions differ in such a way that the normal causes of (nonintentionally) the same internal state differ as well. The background assumptions are that perception is objective in the sense that its objects are such that they do not necessarily co-vary with our discriminative abilities and that we individuate at least some perceptual representations in terms of the objective entities that normally cause them in members of a given species. The argument and its conclusion thus concerns only the most basic perceptual cases—ones whose processing is perhaps fully subpersonal and whose contents do not depend on the subject’s cognitive theorizing, language, or relations to other representations she might have. By contrast, the other three anti-individualist arguments concern much higher-level attitude contents, ones that enjoy linguistic expression and, for at least some of them, figure in scientific theories. The basic point about such contents is, again, that their individuation depends in part on the nature of their extension: what it is to have the concept of water is in part determined by the nature of the kind water and the thinker’s direct or indirect interactions with it. Burge argues that this is true of all empirically applicable concepts, and the particular Twin Earth arguments for the anti-individualist conclusion depend on this fact. So, in “Individualism and the Mental,” he argues that because we do not need to have complete linguistic mastery of a word in order to employ the concept it expresses in our thought, it is possible for the thoughts of two nonintentionally identical individuals to differ. The contents of our thought are individuated, in part, by facts about the language we speak and the community of speakers we live in. This is what is often called Burge’s “social externalism.” In “Other Bodies,” the more familiar point is made that such individuals may differ because they lack full knowledge not of linguistic meaning, but of
34
Martin Hahn
the nature of the extension of the concepts they employ. This thesis is an extension and reapplication of some ideas found in Putnam (1975) and is said to constitute Burge’s “physical externalism.” And finally, in “Intellectual Norms and Foundations of Mind,” he argues that in cases where there is no question of either linguistic or expert-knowledge lacunae, such individuals’ thoughts may differ because a piece of theory they hold applies to the extension of the concept used by one person but not the other. The details of these arguments can be found elsewhere; what is important for us is that, although Burge has views on these issues, the arguments are independent of any positive account of what concepts, natural kinds, meanings of words, and other key items turn out to be. The power of the arguments comes from our intuitions about the Twin Earth examples, and, indeed, the results of the arguments are used by Burge to reach further conclusions about thought and language in subsequent publications. There are two other arguments in Burge that will turn out to be of some relevance in the present context. In his (1977), Burge argues for a thesis about the centrality of de re propositional attitudes: We would not attribute any belief-contents at all to a creature who did not have at least some de re beliefs. Since de re beliefs are irreducibly relational (acts of application of a conceptual content to an object in the environment), Burge’s thesis here is in a loose sense an “externalist” one. The second argument concerns the relationship between two of Burge’s antiindividualist theses. In his (1989), he argues that the conclusion of his (1979) (so-called social externalism) can be derived from the thesis (“physical externalism”) that the individuation of the concepts we think with depends, in part, on the nature of their extensions. Davidson’s Theses To the best of my knowledge, there are two arguments for externalism present, more or less explicitly, in Davidson’s œuvre. One of them, we might call it the “orthodox” argument, is rooted firmly Davidson’s account of radical interpretation and its crucial role in assigning mental content: “Externalism makes clear how one person can come to know what someone else thinks, at least at the ground level, for by discovering what normally causes someone else’s beliefs, an interpreter has made an essential step toward determining the content of those beliefs” (1989, p. 195). To spell this out a bit more: We assign contents of thoughts and meanings of sentences to speakers by observing them in their environment and assigning reference to their expressions on the basis of the perceptual relations we observe them to have with objects in the environment. Although it is surprisingly difficult to find a place where he asserts the doctrine, Davidson seems to have long accepted the view that for a creature to have a belief with a content is to have its
When Swampmen Get Arthritis
35
utterances interpreted by someone.3 The closest I have been able to find to an assertion of this interpretationist view (which Davidson is reluctant to say is an antirealist one) in the early writings on radical interpretation is in his (1975)—an article whose main argument appears to be rooted in the view: We have the idea of belief only for the role of belief in the interpretation of language, for as a private attitude it is not intelligible except as an adjustment to the public norm provided by language. (Where “public norm” is glossed this way: interpretation requires the distinction between a sentence being held true and being true. Hence we need the notion of belief—the notion of a sentence being held true.) (1975, p. 22)
But then it follows, I take it, that the individuation of thought-contents depends on the (observed) normal causes of the beliefs whose contents they are. Or, to put it in a more Burgean way, that two nonintentionally identical individuals can differ in the beliefs they have if they are embedded in different environments (or, for that matter, if their interpreters happen to come up with different theories, since such theories are always underdetermined by the data). A version of externalism is thus built right into Davidson’s most basic theses about language and thought and, if one is a Davidsonian, one does not need elaborate twin earth arguments to show this. Davidson may thus be at least partially correct when he claims that he has been advocating an externalist view all along. It is a straightforward consequence of his interpretationist stance, together with the view that evidence is distal, not proximal stimuli, that whatever the nonintentionally described facts there are about a person, the interpretation of her utterances is not fully determined by them: Mental states do not supervene on local nonintentional states of the subject. As a bit of history of philosophy, however, this interpretation of Davidson’s early writings suffers from the same interpretive overenthusiasm as would attributing existentialism to Aristotle on the basis of his remarks concerning character and eudaimonia, or finding Hume’s view about causation in Descartes on the basis of what he says about the ontological independence of particular moments of the world’s existence. To attribute a view to an author, it is not enough to find that he held some theses from which the view follows. Philosophers, like other mortals, do not have their beliefsets closed under deduction. The procedure is even more suspect when, as in the present case, a statement of the premises themselves is difficult to find in the author’s writings. Thus, although what I am calling the “orthodox” argument for externalism is one that Davidson makes, or alludes to, in his writings in the ’80s and ’90s, the jury is out—at the very best—on whether he made the argument for externalism before Burge’s antiindividualist writings and, indeed, whether he even held the premises. The second argument for externalism found in Davidson is much more recent on any interpretation. It only makes its appearance in the papers where Davidson discusses
36
Martin Hahn
externalism explicitly. There, he often expresses the view that externalism is a thesis that is quite readily apparent if we pay attention to our attribution and interpretation practices and to how we learn the meanings of the first and most basic words. Let us call this the “learning” argument. The famous Swampman, for example, can’t mean what I do by the word ‘house’ . . . since the sound ‘house’ it makes was not learned in a context that would give it the right meaning—or any meaning at all. (1987, p. 445)
Something more akin to an argument along these lines appears, for example, in “The Myth of the Subjective”: It is a commonplace of the empirical tradition that we learn our first words . . . through a conditioning of sounds or verbal behaviour to appropriate bits of matter in the public domain. . . . This is not just a story about how we learn to use words: it must also be an essential part of an adequate account of what words refer to and what they mean. . . . the story entails consequences that seem to have been ignored until very recently. . . . two speakers may be alike in all relevant physical respects, and yet they may mean quite different things by the same words because of differences in the external situations in which the words were learned. . . . The point is that the correct interpretation of what a speaker means in not determined solely by what is in his head: it depends also on the natural history of what is in the head. (1989b, pp. 163–164)
Davidson goes on to note that this argument, and the one about interpretation, is what best establishes externalism, there being no need for Putnam’s or Burge’s more abstract Twin Earth arguments: “the case can best be made by appeal to obvious facts about language learning and to facts about how we interpret words and languages with which we are unfamiliar.” Interestingly, in addition to these two arguments for “physical externalism,” we can find in Davidson an argument that seems to run in parallel with one of Burge’s: Davidson argues, starting with physical externalism, for what seems to be a very strong version of social externalism. The argument I have in mind is the famous “triangulation” argument. We can begin our more detailed comparison of the two views by considering this pair of arguments with what seem to be very similar premises and conclusions. From Physical to Social “Externalism”: Davidson Both Davidson and Burge are said to subscribe to physical (perceptual, as Davidson calls it) externalism. That Burge does so is apparent from his arguments in (1982a) and elsewhere, and Davidson concurs in several places that, although he does not accept Burge’s argument, he has long held the view itself. From this seemingly common ground, moreover, each of them argues for some sort of dependence of content on social circumstances. The arguments are entirely different, as we will see, since Burge’s depends on the very
When Swampmen Get Arthritis
37
features of his account of the relation between thought and talk that turn out to be quite antithetical to Davidson’s views; and Davidson’s argument relies, predictably, on our acceptance of his thesis that meaning and content can arise only in radical interpretation, a thesis Burge has denied in print many times, both implicitly and explicitly. Furthermore, Davidson denies that he is a social externalist of the Burgean type in the very same papers in which he puts forward the argument in question. In Davidson, the argument is known as the “triangulation” argument and is, unfortunately, somewhat difficult to interpret. The most complete statement of it seems to be in his (1989a), pp. 196–199. Thoughts and meanings are attributed on the basis of causal relations between speaker and object in her environment. This is the thesis Davidson dubs “perceptual externalism.” The thesis is, according to him, a consequence of the basic facts of radical interpretation. We assign contents of thoughts and meanings of sentences to speakers by observing them in their environment and assigning reference to their expressions on the basis of the perceptual relations we observe them to have with objects in the environment. At this point Davidson’s account of interpretation begins to diverge radically from its roots in Quine’s radical translation. Where Quine sees a problem of inscrutability of reference and settles on defining perceptual stimulus as that which “impinges on our sensory surfaces,” Davidson argues that the project of interpretation in effect guarantees a solution on which the stimulus turns out to consist of ordinary, distal objects. There are two separate problems here for Quine. One is that, when the subject is in the perceptual situation in which we would say we are seeing a rabbit, the same event can equally be described as her causal relation to the rabbit, to the rabbit’s surfaces, to some light-waves traveling from the rabbit, to the stimulation of my retina, and so on—this has been called the “depth” problem. The other problem is that, as interpreters, we have no way of differentiating between various conceptual schemes the subject could be using: Famously, we cannot tell whether the referent is a rabbit or an undetached rabbit-part. This is the problem of the “inscrutability of reference.” Davidson’s view is that, because we clearly do assign to sentences and thoughts their ordinary referents in interpretation, the solution to Quine’s problems is to embrace the fact that we, as interpreters, impose on the speaker the only plausible ontology we are familiar with: ours. The objects of thoughts and referents of singular terms we attribute to the subject are, of constitutive necessity, the same ones we, the interpreters, operate with. But, argues Davidson, the reason that we operate with the concepts and ontology that we do is that we have an innate range of perceptual and conceptual abilities that determine our responses to the environment and our propensity to track some similarities over others—in short, our basic conceptual scheme with its central notions of objects and objectivity. Furthermore, the only way we can assign meanings to fellow creatures is on
38
Martin Hahn
the condition that they not only share these innate propensities and responses, but also that there is mutual knowledge of these capacities: The only way of knowing that the second apex, the second creature or person, is reacting to the same object as oneself is to know that the other person has the same object in mind. But then the second person must also know that the first person constitutes an apex of the same triangle another apex of which he, the second person, occupies. For two people to know of each other that they are so related, that their thoughts are so related, requires that they be in communication. Each of them must speak to the other and be understood by the other. (Ibid., p. 199)
Our having any thoughts or meanings at all, then, is conditional on our living in a community of other creatures who share our innate perceptual and conceptual propensities and reactions and with whom we communicate. Furthermore, since it is only given such a community that we can be assigned, and assign to others, thoughts about particular objects, the individuation of contents is dependent on these external, social facts. From Physical to Social “Externalism”: Burge Burge’s seemingly parallel argument from physical to social externalism occurs in his (1989). Burge asks us to consider words that are empirically applicable and nonindexical. These comprise a large class of words, including natural-kind terms that other philosophers have claimed to be indexical. In fact, Burge has argued that the class includes virtually all words that apply to public objects, properties, and events. Burge has also argued against the indexical account of such words (1982a). If we grant that such terms are not indexical, it follows that, if two of them differ in reference, they must differ in meaning. This is just a consequence of the standard view that the meaning of (nonindexical) terms determines their reference. But if two such terms differ in meaning, that indicates that they express different concepts. Physical externalism—more precisely, Burgean anti-individualism—comes in as a premise at this stage. Our cognitive mastery of words such as ‘water’ or ‘elm’ does not fix their referents. In other words, we can operate perfectly well with such concepts without having at our cognitive disposal any way of determining what items they refer to in all (or even most) cases. This is a well-known point found already in the work of Putnam and Kripke. It follows from these facts that the contents of our thoughts involving such terms can differ if their referents differ without there being any difference in the individual, nonintentionally described. In other words, individualism is false because the individuation of our thoughts depends on the identity of their referents, which is not determined by any facts about the individual. But now, asks Burge, how do we individuate, sharpen, and correct the concepts with which we think and the meanings of the words we speak? To do this, we need to be able
When Swampmen Get Arthritis
39
to individuate their referents. But the only way we can do this, given the kind of creatures that we are, is by interacting with other speakers, both because the referents themselves are public objects, and because we often depend on others to link our words to their referents. We rely on each other to refine and correct our usage: sometimes because others have perceptual experiences we lack (this is the case for me and ‘manatee’), sometimes because they are scientific experts (as with ‘glucose’), sometimes because of the sheer wealth of their daily experience (as with ‘dry-rot’). Thus, Burge concludes, the individuation of our mental contents depends, in part, on our social environment. We have the mental contents we have, in part, because we live in a community of creatures with similar perceptual and conceptual abilities.4 The perhaps somewhat surprising conclusion is that the famous ‘arthritis’ case in Burge (1979a) is, in fact, a derivative one. In (1989), Burge argues that what is often called his “social externalism” is, in fact, derived from the fact that the individuation of our thoughts depends on our environment plus some facts about human psychology. Davidson and Burge thus both argue from a version of physical externalism, which the standard story about them claims they agree on, to a version of social externalism, which Davidson is reluctant even to call ‘externalism.’ The two arguments are, to say the least, very different. Each depends on premises the other philosophers takes special care to deny: Davidson’s on his view that meaning and content are constituted by the process of radical interpretation; Burge’s on his view that we can stand corrected in our own idiolectical usage. Since the major premises of physical externalism are thought to be very similar, if not identical, it might well be thought that the perceived difference in the conclusions is explained by this radical difference in arguments. I want to argue, on the contrary, that the difference in arguments explains how the two philosophers reach quite similar conclusions from different premises. Burge’s Anti-individualism Burge’s physical anti-individualism is argued for in his (1982a).5 The argument is the familiar Twin Earth one. Alfred, who is a perfectly competent speaker of English, has many thoughts about aluminum: He believes pots and pans are sometimes made of it, worries about the conflicting evidence on its connection to Alzheimer’s disease, knows that aluminum ore is called ‘bauxite’, named after the lovely southern French village of Baux-en-Provence, and so on. His twin, Twalfred, is physically (and introspectively, if we wish) identical to Alfred. There is, however, no aluminum on Twin Earth. The pots and pans, Les Baux-en-Provence, and so on—it all concerns another metal, qualitatively similar enough to aluminum that whatever differences there might be, they have
40
Martin Hahn
contributed nothing to the Twalfred’s “internal” story to differentiate it from Alfred’s. Now, does Twalfred have any thoughts about aluminum? No, he does not. For where would he get them from? No such metal is even found on Twin Earth. The fact that both twins are occasionally apt to utter the sounds: ‘Aluminum smelters use much electricity’ is an odd linguistic coincidence, much in the way ‘Lepage’s’ is a common brand name of a glue in North America and ‘lepit’ is the Czech verb meaning ‘to glue’. Twalfred’s utterances need to be translated into English: ‘Aluminum smelters use much electricity’ is true in Twenglish iff twaluminum smelters use much electricity. In fact, it means that in Twenglish. Twalfred’s thoughts, like his utterances, involve the concept of twaluminum and not that of aluminum. Thus, the individuation of contents depends, in part, on the referents of the concepts that compose them. The conclusion of the argument is entirely general and largely negative. It is that there are thoughts whose contents are individuated, in part, by reference to the physical environment of the individual, because which concept the thinker is grasping is not determinate without reference to their extensions. The argument is agnostic about the whys and hows of this. Burge has views on this, of course, to be found in his (1977), (1986a,c), (1989) and elsewhere, but the central argument against physical individualism does not depend on these and is consistent with other views. Davidson’s Perceptual Externalism Donald Davidson is unimpressed with the Twin Earth arguments, largely because he thinks the counterfactuals to which we are asked to give our assent are, in fact, impossible to evaluate. This, undoubtedly, is in part because in order for us to be able to say whether Twalfred has or doesn’t have thoughts such as that frying pans are sometimes made of aluminum, indeed in order for him to have or not have them at all, we have to observe his verbal behavior in his environment and interpret it. But we would only assign twaluminum-thoughts to him if we recognized twaluminum for what it is, not aluminum. Otherwise, we would surely assign aluminum-thoughts to him. Now, a very clear difference in Davidson’s and Burge’s views might well be hidden here. If we transported Twalfred to a planet where there is aluminum, it is hard to see how there could be any reason not to attribute aluminum-thoughts to him on Davidson’s account. After all, what would normally cause his ‘aluminum’-responses would be aluminum. But Burge’s view is, it would seem, that a transported Twalfred would not start having aluminum-thoughts until he spent some time here—learning English and interacting with our pots and pans. Given Davidson’s third-person emphasis in his account of content and meaning, Twalfred’s thought-contents seem to depend more on where we, the interpreters, have been spending
When Swampmen Get Arthritis
41
our time than on where he has spent his. Unfortunately, as we shall see shortly, this account of the differences is in sharp tension with one of Davidson’s positive arguments for externalism. Interestingly, Davidson does not seem to think that much of an argument is needed. As we saw, he generally expresses the view that externalism is a thesis that is quite readily apparent if we pay attention to (a) our attribution and interpretation practices, and (b) how we learn the meanings of the first and most basic words. In addition, Davidson gives us his own Gedankenexperiment, the famed Swampman (1987, p. 444): Donald Davidson is standing next to a dead tree in a swamp when lightning strikes him, reducing his body to its elements. By coincidence, a molecule-for-molecule replica of Davidson is, at that very moment, configured out of atoms composing the dead tree. The resulting Swampman proceeds to behave in the way Davidson would have, living in his house, writing articles on radical interpretation. The question isn’t, of course, whether the Swampman is Donald Davidson, for most of us would deny that, but whether his thoughts are the same thoughts Donald Davidson would have had, and his standing beliefs the same beliefs Davidson did have. Davidson, being an externalist, wants to answer in the negative. He says of the Swampman (1986, p. 445): (1) It can’t recognize my friends; it can’t recognize anything, since it never cognized anything in the first place. It can’t know my friend’s names, it can’t remember my house. (2) It can’t mean what I do by the word ‘house’ since the sound ‘house’ it makes was not learned in a context that would give it the right meaning. (3) I don’t see how my replica can be said to mean anything by the sounds it makes, nor have any thoughts.
It is not clear whether these three, rather different, successive claims are supposed to constitute an argument or just stand on their own, as three independent externalist facts about the Swampman. Certainly, philosophers sometimes take them to be the latter, and I don’t know of any plausible way of drawing connections between them. Let us, therefore, consider them one at a time. The first point is undoubtedly true, but has little to do with Burge’s anti-individualism, or any interesting form of externalism, so far as I can discern. The propositional attitudes in question (knowing, remembering, recognizing) are factive: The proposition known must be true, the event remembered must actually have happened, and the person recognized must actually be the one seen before. No one has supposed the individuation of those states to be an entirely individualist or internal matter. Whether or not I know that p depends on not only my mental state, but requires crucially that p be true. And, although it is true that some recent theories of knowledge have been called “externalist,” we have no more than
42
Martin Hahn
a pun here. Generations of traditional philosophers have sought to find internal criteria for recognizing that one had knowledge—for being assured that the external requirement of truth was satisfied. Their failure and the success of scepticism is driven precisely by the fact that truth is required for knowledge. What is “externalist” about these recent views is that they deny that we can give a certain kind of account of what, besides belief and truth, knowledge requires. The status of the second point is more complicated. Trivially, since the creature is not Donald Davidson, it cannot have Donald Davidson’s thoughts or meanings. But can it have the same thought-contents as Donald Davidson, express the same meanings? The jury is out on this one, I would say. Insofar as I find Burge’s arguments starting with his (1977) convincing, I’m inclined to think that for a creature like me to have my thoughts it must have had similar perceptual interactions with objects in the world, including ones with fellow speakers of my language. That leads me to think that a creature with relevantly different previous experiences wouldn’t have my thoughts. But what about a creature with no such experiences? I don’t know how to evaluate this possibility and, without some powerful arguments, I don’t see any reason to decide one way or the other. Davidson himself has arguments available to him to settle the question either way, as I shall argue. Davidson’s point, that I learned the word ‘house’ in a certain kind of context (the causal presence of houses) that gave it the meaning it has, seems to me to be inconclusive. Swampman did not learn ‘house’ at all. The right perceptual experiences, and the right learning circumstances perhaps, are clearly required for creatures like us to have the concepts we do. But Swampman is not like us in the relevant respects, and I am not nearly as sanguine as Davidson is that these are a priori or constitutive requirements; that what it is to have a basic empirical concept is to have learned it in the right way. Below, I will introduce some considerations against the view rooted in Davidson’s own views. Davidson’s third point concerning the Swampman is even stronger than the second. We are told the Swampman not only doesn’t have the same thoughts as Davidson, it has no thoughts at all; its words have no meanings. Now, again, I find this hard to evaluate, even given Burge’s strong arguments for the dependence of our concepts on our perceptual history. The parallel arguments by Davidson are the orthodox and the learning argument set out above. Of the latter, we should note that it seems simply false, on the face of it, that reflecting on the learning situation alone will get us to some externalist thesis. The fact, acknowledged by every empiricist philosopher, and some others, that we learn words, in part, by being confronted with examples of their extensions has notoriously not led people to the desired conclusion. The reason is that prima facie the issue of how concepts or meanings are individuated is independent of how they are learned—indeed, examples like the Swampman might be used precisely to make this point. The point is analogous to some
When Swampmen Get Arthritis
43
other, familiar points about learning, such as that the fact I have to learn the multiplication table by rote from a book does not make arithmetic a posteriori. It is useful to recall at this juncture that anti-individualism is a thesis about the individuation of mental states. It is not the thesis that no one could learn (or, which is much stronger, have) a language without there being an environment to play certain causal roles. It might well be, as most philosophers have believed for hundreds of years, that the only way to learn an individualistic language is in a certain environment. The dependence on the environment that is relevant here is not causal, however, but individuative. Nor does the orthodox argument give us much comfort here. Radical interpretation is a descendant of Quine’s question of how one would translate the utterances of an unknown tribe when we come upon it. I find it hard to get to conclusions about what must have been the case about learning the languages for the creature at hand, based on the exercise of interpreting its current utterances. We come across the Swampman, and it makes utterances in the presence of objects we call ‘house’, ‘tree’, and so on. What choice do we have but to attribute to it thoughts and meanings about houses and trees? That is, if we agree with Davidson’s epistemic points about what is available to the interpreter, his methodological point about what her task is, and his metaphysical conclusions based on these, it is very difficult to see why such radical conclusions about the Swampman ought to follow. And, since I cannot find an argument to suggest that the radical interpretation of the utterances of a creature is impossible if we did not witness it learn the language, it is hard to see how any of the points about learning are relevant. What we are left with, then, appears to beg the question. The Swampman Gedankenexperiment and the things Davidson says about it will be convincing only to someone who already believes the conclusion of what I have called Davidson’s “learning argument”: that, unless someone has learned the words they use in situations like those in which we learned ours, they will not share our thoughts. But if we are not antecedently convinced of this form of externalism, and I have given some reasons for doubting it, we will find nothing convincing about the Swampman. Worse, there is, it seems to me, a certain tension between Davidson’s account of radical interpretation and its constitutive role in the very existence of contents and meanings on the one had, and his favoured learning argument for perceptual externalism on the other. Contents are, according to the radical interpretation story, individuated in part by their normal perceptual causes. But the remarks about learning (and the ones about Twin Earth counterfactuals we noted at the start of this section) tell a different, partially conflicting, story. There is nothing about the Swampman, or Twalfred, that makes them any harder to interpret than any of the rest of us: We meet them, they utter sentences in the presence of objects, we form hypotheses about the perceptual causes and the truths they are likely to be remarking on and, eventually, about the translations of their utterances into our
44
Martin Hahn
language. The only constraint on the causes is that they be “normal”; nothing follows about the constitutive role of learning situations. A first step in an attempt to disentangle this knot of problems is to note that the claim about the Swampman that Davidson needs to make is that he has no thoughts when he is created, not that he cannot be interpreted to have thoughts as he goes about the world. I take it to be central to Davidson’s theory that mental ascriptions apply as a result of actual interpretation; mere interpretability is insufficient. Being interpretable is a dispositional property had, one would assume, by all kinds of nonintentionally describable creatures in virtue of some individualistic facts about them—unless the property is entirely mysterious. The Swampman, you and I, and quite a few other things have constitutions that ground such a dispositional property; while chairs, donkeys and computers lack them. Few philosophers would disagree, but Davidson’s theory is more radical than that. Being interpretable might, for all that has been said, supervene on local nonintentional states. It is actual interpretation that gives rise to mental ascriptions and brings with it externalism both by way of the causes of occurrent beliefs and of the interpreter. But then the Swampman’s causal history is, on this account, immediately relevant to his lack of thoughts because of one feature only: Interpretation has not occurred. What, then, do we make of Davidson’s learning argument, in which he insists that the way the most basic terms of a language were learned (or not) are a relevant feature of Swampman’s causal history? There are two ways causal history in this broader sense might count in assigning contents: directly, or by way of its role in interpretation. The second, more conservative view meshes well with the orthodox argument that the meanings of the most basic terms are assigned with respect to the normal causes of the assertions (and beliefs) being interpreted. In order for her to determine what the normal causes of the assertions in question are, facts about the subject’s past, including perhaps facts about how a language was learned, must be part of the evidence available to the interpreter. But is this plausible even for run-of-the-mill human subjects? If my interpreting the State of the Union address depends on facts about how Bill Clinton learned his particular Southern idiolect, such evidence can at best only be assumed by me on some general grounds, given my lack of acquaintance with or even knowledge by description of the details of that process. In the case of the Swampman, there is no history. The curious effect is that when someone who doesn’t know Swampman’s origin proceeds to interpret him on the very same assumptions I use for Clinton, she gets an interpretation that passes all the empirical tests one would want: It predicts, it explains, and so on—all on a false assumption. But when we turn to the case of an interpreter who does know that Swampman is brand new, we are hard pressed to see how not making that assumption will make the job any harder. Naturally, one would be quite leery at first in attempting to interpret such a creature. After all, what are the chances that something that came out of the
When Swampmen Get Arthritis
45
swamp is interpretable? But as confirmations roll in, one would quickly fall upon the hypothesis that the Swampman is interpretable, that is, that he has thoughts and expresses meanings, and thus that he is substantially like the interpreter. Now what special role ought one to give to the causal-historical hypothesis that Clinton learned language in a situation much like the one I learned it in and the Swampman did not? It is hard to see. But sometimes Davidson writes as though the causal learning history plays a more direct veto role: No creature without the right history could have language or thought, presumably whether the interpreter is aware of the gap or not. This, it would seem, could only mean that such a creature is uninterpretable. Again, we have two choices here, neither of them very attractive, unfortunately. One is that interpretation will be possible only on creatures with a causal history similar to ours. But we are given no argument to support such a view, and it runs counter to the spirit of Davidson’s naturalizing project. Whether or not a creature is interpretable ought to be just a question of whether treating the creature as uttering sentences that are systematically translatable into the interpreter’s language gives us good predictions and explanations of the creature’s behavior. The wild hypothesis that we should use for the Swampman the same theory we developed for Donald Davidson yields as good results as we ever got with Davidson in the past! Hence, in the absence of some powerful argument, the Swampman himself provides a counterexample to the thesis that a particular kind of causal history is necessary for interpretability.6 The only other way we can bring in causal history is as simply an extra condition not on interpretability, but on having thoughts. Being interpreted is not sufficient for having thoughts; the subject’s language must also have been learned in the right way. But this, congenial as it might be to someone like Burge perhaps, seems deeply antithetical to Davidson’s insistence that the third-person interpretive stance is all there is to the attribution of meanings and thoughts: that it is constitutive of them. The conclusion that offers itself is that the learning argument for externalism in fact conflicts with Davidson’s views that actual interpretation is constitutive of meaning and thought. Toward the end of this essay I shall offer a hypothesis as to why Davidson thinks otherwise, but for the present purposes I will concentrate on the more moderate and congenial orthodox argument. A Surfeit of Physical Externalisms It is time to take stock of what I take to be the differences between Davidson’s and Burge’s “physical externalisms.” It will be useful to begin with what each theory might say in response to characteristic Twin Earth and Swampman thought experiments (without too much speculation on what the two philosophers might say).
46
Martin Hahn
According to the orthodox Davidsonian line, Twalfred has no thoughts at all until we interpret him. That will involve either observing him on Twin Earth, or getting him to come here. Let us take the first case. His thoughts are going to be individuated, in part, by reference to the objects that cause them, but whether these will be described as “twaluminum” or “aluminum” in our interpretation is difficult to say. It might be thought depend on whether we know that the stuff he is interacting with is aluminum or the other metal. Presumably, if we take him to be interacting with Twaluminum, we will not use homophonic translation into English. But even if we don’t know that there is no aluminum on Twin Earth, and interpret Twalfred’s thoughts as being about aluminum, it might be possible to give a Davidsonian account of this as a misinterpretation due to simple ignorance, without violating the principle that subject and interpreter must share conceptual schemes. Here is an analogous situation that doesn’t violate Davidson’s strictures against fantastical Twin Earth thought experiments: John was raised in British Columbia. On his arrival in Newfoundland, he finds people talking about cod fishing a great deal. He interprets them homophonically: ‘ “There are few cod left” is true if there are few cod left’, and so on. A bit later, he discovers he made a mistake. The fish the Newfoundlanders are mourning are not at all related to the various spiny bottom-feeders called ‘cod’ on the West Coast.7 If this is right, then on arriving on Twin Earth, however we actually interpret the natives, the correct interpretation does not assign them aluminum-thoughts, but twaluminum ones instead. A source of trouble with this line of Davidson interpretation is Davidson’s antirealism about natural kinds or, as he prefers to call it, his anti-essentialism. According to him, a naturalistic solution to Quine’s problems of indeterminacy of translation and inscrutability of reference cannot include the assumption that our referents fall into kinds in and of themselves. Thus, one aspect of Burge’s anti-individualism is greeted with considerable scepticism (see, e.g., 1991a, p. 200). Burge’s arguments lead to the conclusion that whether or not anyone knows it, whether or not this difference is behaviorally manifested in any way at all, Twin-Earthean sentences like ‘Aluminum is recyclable’ are not correctly translated by English homophones (and attitudes are not correctly attributed to Twin-Eartheans by such that-clauses in English) simply because of the metallurgical facts about the stuff the thinker, his linguistic community, and their ancestors have been interacting with on the two planets—aluminum and twaluminum—are not of the same kind. But Davidson objects to this simple explanation: What determines the content of such basic thoughts [e.g., ‘There’s a cow’] . . . is what has typically caused similar thoughts. But what has caused them? There are many choices. . . . It is we who class cow appearances together, more or less naturally, or with minimal learning. . . . We group together the causes of someone’s responses . . . because we find the responses similar. . . . The identification of the objects of thought rests, then, on a social basis. (Ibid., pp. 200–201)8
When Swampmen Get Arthritis
47
We have here, I think, a very deep division between Burge’s and Davidson’s notions of concept, one that will lead to some fairly startling conclusions below. Whereas Burge is committed to the view that, in employing empirical concepts, we allow the world to determine the details of their application, Davidson’s nominalism precludes such a view. But, since Davidson agrees that such a determination is not “in the head” of the individual speaker, he has to give an account of how such norms arise. The triangulation argument for social externalism is thus, for Davidson, not just a matter of filling in the details of how it is we, as a matter of (even necessary) psychological fact, rely on our linguistic community to sharpen the explications of our concepts. That is the role Burge’s argument in his (1989). In Davidson, by contrast, “The identification of the objects of thought rests . . . on a social basis.” Communication with like-minded others is constitutive of objectivity, classification, and intentionality, according to Davidson. Given this, it is indeed hard to know what to say about the case of an earthly interpreter visiting Twin Earth. The case of Twalfred coming here, on the other hand is a familiar one. Twalfred is, from the orthodox Davidsonian viewpoint, just like Swampman. He behaves like one of us, described nonintentionally, but he has not been interpreted yet. He has no thoughts, speaks no language. How do we interpret him? The only way we know how: as having thoughts about aluminum. After all, he interacts with the stuff we classify as aluminum. The contrast with Burge is quite sharp. Wherever Alfred happens to be, according to Burge, he has twaluminum thoughts. Of course, after a period of dwelling here, his thoughcontents would tend to change, but the contents of his thoughts are not determined by the present relation between him and the metal before him, nor do they depend on the interpreter’s language and thought. In no way is Twalfred ever like the Swampman. Davidson’s learning argument for externalism would seem to put him closer to Burge’s views. After all, don’t they both agree that the causal history of the subject is, in part, what determines the contents of her thoughts? The issue here is subtle. They do agree, along with most even vaguely naturalistic philosophers—internalists and externalists alike—that which concepts a person has is causally determined by her history. Further, a certain kind of causal history is causally necessary for a person (though perhaps not for a Swampcreature) to have any thoughts at all. But, although Burge’s Twin Earth arguments rely on these facts, their conclusion is not that the contents of thoughts are individuated, in part, by the causal history of the subject, but that the referents of her thoughts play this role. Twalfred’s thoughts are twaluminum-thoughts as opposed aluminum-thoughts because he (and his linguistic community and their ancestors) was brought up in a place full of twaluminum and not aluminum. It would have been causally impossible for him to have aluminum thoughts. However, the individuation of the contents of his thoughts depend, in part, on facts about twaluminum—whether or not those individuative facts played any causal role
48
Martin Hahn
in his learning the language. So let us say twaluminum has chemical structure RST, and thus twaluminum thoughts are essentially about stuff of that structure. The fact that what Alfred and his ancestors interacted with was RST, rather than Al, played no causal role in the genesis of Alfred’s beliefs—Alfred’s perceptual system being utterly devoid of sensitivity to such chemical differences. Another way to distinguish Burge’s and the views that follow from Davidson’s learning argument is this: Suppose that Twin Earth and Earth have insufficient chemistry to differentiate the two metals. Following Burge’s line of thought, this makes no difference to the fact that Alfred and Twalfred have, unbeknownst to them and anyone else around them, different concepts. But, for Davidson, the situation is quite different. The twins’ languagelearning histories are identical in the sense that their teachers’/interpreters’ behaviors (nonintentionally described) were identical.9 There was, according to us, a difference in the causes of their various beliefs, of course, since there is a difference between the two metals. But that difference was unavailable to them, and there is absolutely no reason to count it as a difference because, given Davidson’s denial of essentialism, there is no such thing as a difference in the nature of two kinds that is independent of the classifications used by the speakers. In sum, while both Burge and Davidson believe that in order to have the thoughts we have, we must have a history of causal interactions with the referent, Davidson seems to believe that the object of thought (and thus the thought itself ) is individuated by reference to the history, whereas Burge thinks it is individuated by reference to the kind that figures in the history. Davidson rejects the latter view precisely because he believes that the kinds themselves are individuated in virtue of our interactions with the environment. There is a reply available to Davidson, to be sure: Since we make a distinction between aluminum and twaluminum, the only thing we can say correctly is that Alfred’s and Twalfred’s learning histories differ and so their thoughts differ as well. But now we are back to the orthodox Davidsonian view again. We as interpreters don’t need to look to their learning history at all, except perhaps as evidence for our interpretation. That they have different thoughts stands and falls with the fact that we see them interacting with (what we regard as) different substances. The facts of their learning history make no difference to the individuation of Alfred’s and Twalfred’s contents except insofar as they make a causal difference to the states they now have. The only external individuative factors that matter have to do with the linguistic practices (and thus thoughts) of the interpreter. The contrast with Burge’s view is sharp here. Burge’s anti-individualistic arguments are meant to show that the contents of two nonintentionally identical subjects’ de dicto thoughts can differ if the referents of their thoughts differ. The reason for this is that the relevant concepts are empirical in the strong sense that which concepts one uses depends
When Swampmen Get Arthritis
49
in part on the (empirical) natures of the objects they are applied to. When we think with such concepts, we do not necessarily fully grasp them; our thoughts and usage are subject to correction; no set of nonintentional facts about an individual and her behavior suffices to the determine what it is she is thinking about. But, perhaps surprisingly, Davidson turns out to be an individualist, that is, a sort of “internalist” in just this central sense. According to Davidson, what determines the object of thought, in the most basic cases, is whatever causes the belief. But what that is is determined only by the interpreter’s usage. In other words, the concept the subject has is not so much partially individuated by reference to the external facts (the object, the interpreter) as it is identical to the interpreter’s concept. Davidson explicitly rejects Burge’s central thesis that thinking with a concept does not require a complete mastery of it (see, e.g., 1986, p. 448). Unless both interpreter and subject had full mastery of the concepts constituting their thoughts, first-person authority would fail. Further, the nature of the object of thought does not provide part of the content except insofar as the interpreter does. It cannot do that, because this would entail the dark doctrine of essentialism (and the even darker doctrine that normative notions such as concepts have some reality prior to interpretation). But then the only sort of externalism Davidson is committed to seems to be this: Unless interpreted, a subject has no thoughts at all and the individuation of the contents of thoughts depends (and not just “in part,” it would seem) on the linguistic practices and thus the concepts of the interpreter. Indeed, the only conceptual scheme the subject can possibly have is the interpreter’s. Further, in the most basic cases, the interpreter will take the causes of beliefs to be the objects, individuating the objects in the only way she knows how: her way. What the Davidsonian picture guarantees is that subject and interpreter will have the same concepts (more or less).10 But then what of externalism, the doctrine that the contents of thought are individuated, in part, by external factors? The only external factors available for the job are the concepts in the interpreter’s thoughts. But these are a priori guaranteed not to differ from the subject’s. Thus, although contents of thoughts do not, indeed, supervene on nonintentional facts about the individual, there is still something interestingly internalist here: Each individual in the community has available to her—fully grasps—concepts that determine their extension. Which concepts these are is of course determined only by interpretation, but, whatever they are, the concepts we think with do not have the interesting antiindividualistic properties that we rely on the world to fill in their details and on our community to explicate this relationship. The view is not so much that content is nonindividualistic as that there could not be a single individual with content. I conclude that Davidson not only doesn’t propound the same “externalist” thesis as Burge, he seems committed to denying anti-individualism. Nor should Burge wish to
50
Martin Hahn
accept Davidsonian externalism in either the orthodox or the learning version. Burge’s realism about contents precludes his subscribing to the orthodox view: Whatever else contents are, they are not constituted by interpretations. Nor ought he to have too much sympathy with Davidson’s insistence that the crucial causal role played by the circumstances in which we learn to speak ought to be made into a constitutive one. In particular, Burge can quite consistently deny Davidson’s intuitions about the Swampman. There is nothing in the central Twin Earth arguments to lead to a clear conclusion about such creatures. Since meanings and thoughts are not constituted by either interpretation or learning history, the most Burge is committed to saying about the Swampman is that he does not have the causal history necessary for creatures like us to have thoughts. But there are two conclusions one can draw here: either that the Swampman has no thoughts or that he is not a creature (relevantly) like us. The second strategy has something to recommend it, namely that the Swampman patently isn’t a creature much like us. For example, no creature like us could possibly engage in the behaviors the Swampman engages in without a causal history much like ours. But suppose Burge’s conclusion turns out to be that the Swampman doesn’t have any thoughts. Suppose even that the conclusion is reached because, rather than in spite, of the premise that he is not a creature relevantly like us, so that Burge ends up agreeing with Davidson that only creatures like us have thoughts.11 Such a conclusion is still, on the face of it, completely independent of Burgean anti-individualism, the thesis that the thoughts of those creatures that do have thoughts do not have individualistic contents. Davidsonian perceptual externalism and Burgean anti-individualism do not appear to be the same thesis at all. In fact, each philosopher seems committed to denying what other centrally asserts. What, then, of the various agreements documented at the head of this paper? Let us run through them quickly. The first point, Davidson’s citing approvingly Burge’s statement that mental states “cannot be accounted for in terms of physical, phenomenal, causal-functional, computational, or syntactic states or processes that are specified nonintentionally,” turns out to be too general to distinguish between the various theses we’ve been discussing. In particular, ‘accounted for’ means ‘constitutively explained’ to Davidson, ‘individuated’ to Burge, and what is needed in addition are interpretations for the former, objects and their kinds for the latter. Of course, they do agree fully if ‘accounted for’ means ‘causally explained’—our having thoughts is causally explained by the history of our interactions with the world. The second point of agreement cited was that Burge seems to be presenting precisely Davidson’s oft-repeated dictum about the constitutive role of normal causes in the most basic interpretative cases when he says of certain representational states that “[t]heir carrying information, their having objective intentional content, consists partly in their being the normal causal products of objective entities. . . .” In this case, however, the view Burge
When Swampmen Get Arthritis
51
expresses is not general enough to constitute substantive agreement with Davidson. In particular, as the context of the quote makes clear, the states Burge has in mind are perceptual states and the causes low-level perceptible properties such as being a moving edge or a color contrast. Burge does not, as far as I can tell, mean to say what Davidson is fond of repeating: that cow-thoughts are those thoughts normally caused by cows and applethoughts those normally caused by apples. This may be a good time to recapitulate the difference between Davidson and Burge on the issue of the causal history of the speaker. Davidson believes that the particular causal history of the thinker plays a role in the individuation of her mental states. On the orthodox argument, crucial to this history are the normal causes of basic empirical beliefs: Cowbeliefs are the ones normally caused by cows. On the learning argument, what is crucial are the circumstances under which the basic terms were learned (in the normal human case, some of the relevant cases merge: we learn ‘cow’ in the presence of the normal causes of cow-beliefs, viz. cows). Burge believes that what is crucial to the individuation of the relevant mental states is the nature of the kinds in one’s environment. Burge does not believe in magic, however: Without causal interactions, the environment could play no individuative role. But we should distinguish at least three theses here. The one that Burge clearly subscribes to is that our thoughts are, in part, individuated by reference to the nature of the kinds that figure in such interactions. A different thesis, which Davidson’s nonorthodox arguments commit him to, but which is not part of Burge’s view—or so I argue above—is that our thoughts are, in part, individuated by reference to the causal history of interactions themselves—as opposed to the natures of the kinds we interact with. Yet a third thesis is the one at issue in the present case of Davidson’s claim that he and Burge hold the same view: that for some intentional states, “their having objective intentional content, consists partly in their being the normal causal products of objective entities” (Burge 1986a, p. 417) Now, Burge does argue for a thesis like this for basic perceptual beliefs, but he is not committed to it for cow-beliefs. Nor does it seem to be plausible. I, for example, have many manatee-beliefs, but none of them had manatees as causes, and for none of us is it the case that dodos are the normal causes of our dodo-beliefs. Burge’s view, unlike Davidson’s, does not depend on the notion that high-level beliefs have normal causes, much less that unless they had normal causes, they would have no content. This, it seems to me, is a strength of the view, but here is not the place to elaborate. A closer look at the third purported point of agreement will help us bring out another crucial difference between Davidson’s and Burge’s “physical externalisms.” It was noted that Burge and Davidson agree that, unless at least some of a subject’s thoughts were applied to particular objects in our environment, we would not credit her with any thoughts at all. This is, I believe, the point at which the two philosophers are closest to agreement.
52
Martin Hahn
Note, however, that Burge articulates this thesis several years before he begins to defend anti-individualism and, although it does play some role in explaining why contents are not individualistically individuated, it is pointedly not that very thesis. That it isn’t is both important and quite obvious in Burge’s work. Anti-individualism is a thesis about our de dicto propositional attitudes, and Burge goes to some considerable trouble to persuade us that thoughts involving, say, the concept of water are not somehow implicitly de re courtesy of a hidden indexical. The thesis Davidson and Burge seem to agree on is that there are some propositional attitudes that are irreducibly de re and that without them we would have no thought at all. But a crucial difference between Burge and Davidson turns out to be that, whereas for Burge this thesis is of necessity different from his anti-individualism, for Davidson it forms the very crux of externalism. There are, thus, three separate theses in Burge that Davidson regards as one: One is that, for certain attitudes, their normal causes are constitutive of their content; the second is that some of our attitudes are de re, and the third is that the contents of our mental states cannot be individuated without reference to objects outside of the individual, nonintentionally described. In every discussion of externalism I can find in Davidson, he identifies the thesis that some attitudes are irreducibly de re as an externalist thesis, in fact, as his externalist thesis (at more explicit moments, he couches the point in terms of scepticism concerning the de re / de dicto distinction). There is good reason for Davidson’s conflation of Burge’s theses. When Burge says that the nature of the objects around us determines the contents of our thoughts, he means to say that “[t]hought-kinds are individuated in a way that depends on relations one bears to kinds in one’s physical environment.” But Davidson follows Quine in denying that there are any kinds in our environment independently of how think about them. Thus, when he says that the content of our most basic thoughts is determined by their causes, he must per force mean that such thoughts are de re—there are no causes except particulars (particular events, actually). Kinds only come into being with interpretation. The last point of purported agreement we started with, the relevance of causal history to mental contents, has already been discussed at some length. Although Davidson and Burge agree that the causal history of a person is causally crucial to her having the thoughts she has, neither philosopher seems committed to the view that history is individuatively crucial to those attitudes. There is nothing in Burge to suggest this, and while Davidson asserts it with what I have called the “learning argument,” it doesn’t seem compatible with the orthodox Davidsonian view that actual interpretation constitutes meanings. The Unexpected Connection: Davidson and Burge on Social Externalism The relationship between Davidson and Burge on the issue of physical or perceptual externalism is, then, fairly complex. Burge has an argument to the conclusion that it is in the
When Swampmen Get Arthritis
53
very nature of some of our attitudes that their contents cannot be fully individuated without reference to the nature of their extensions, and I have argued that this is a thesis Davidson rejects. But both Davidson and Burge assert, very roughly, that our conceptual schemes can be what they are only because we have certain perceptual relations with objects around us. This thesis comes to, in Burge, a positive account of why anti-individualism is true of creatures like us, but it is a separate thesis and does not constitute the main argument for it. If it were false, anti-individualism, and Burge’s arguments for it, would not necessarily fail. Davidson, on the other hand, takes these considerations about perception and de re attitudes to be constitutive of externalism. From these rather different premises, and other assumptions that are deeply antithetical to the other’s theories, Burge and Davidson construct arguments to a conclusion they largely agree on: social externalism. This is the view that in order to have the thoughts and linguistic meanings we have, we need other people whose perceptual and conceptforming abilities and dispositions are innately substantially similar to ours. It is not clear why this agreement is usually denied, in particular, why it is denied so strongly by Davidson—often in the same papers in which he presents an essentially Burgean social externalist view.12 I have, however, come across one reason for denying it that is based on a confusion. A short digression to clear this misunderstanding up is in order at this point. It is sometimes thought that social externalism concerning linguistic meaning has as a consequence, or even is identical to, the thesis that languages are essentially social in quite a different sense. For philosophers, at least, it is an interesting question whether languages are essentially conventional and, if they are, whether that means that the unit of study for semanticists and syntacticians of natural languages is the language spoken by a certain community (along, perhaps, with their other social practices). There have been many philosophers who, in one way or another, have subscribed to such a view. There is, for example, a whole tradition in philosophy of language that takes convention to be central (Grice, Lewis, Bennett), another group that takes a certain sort of social embedding to be essential to language and draws conclusions about the proper subject of study from it (Wittgenstein, Charles Taylor), and yet a third group that is impressed by its redrawing of the boundary between semantics and pragmatics and counts, at least partially, on social factors to determine the referents of some of its terms, taking the meaning of proper names, for example, to be essentially social (Kaplan, Kripke, Donnellan). The three groups have virtually disjoint views except for their agreement on the proper object of study for philosophers and linguists: social, conventional, natural languages such as English (or perhaps more localized dialects of them). Many linguists, on the other hand, agree with Chomsky that the only theoretically respectable object of study is the idiolect. Not only are the boundaries of any other object of study constantly shifting and ultimately arbitrary (how many speakers, what sorts of social circumstances, what relevant similarities, etc. does it take to constitute a dialect?),
54
Martin Hahn
but the essential questions linguists need to ask about the relationship between competence and performance can make sense only when asked of an individual speaker. Competence is necessarily a matter of individual psychology. We may speak of equivalence-, or rather similarity-classes of idiolects for the purposes of, for example, the diachronic study of phonetics or syntax, but this does not change the fact that a language is, first and foremost, an idiolect. There are many subtle moves and distinctions to be made here, but one thing that is quite clear is that Burge and Davidson are in essential agreement on this issue. It simply is not the case, as is often thought, that whereas Davidson falls in with the linguists and philosophers who believe that languages are idiolects, Burge’s social externalism shows him to be opposed to this view. In “Wherein Is Language Social?” which appears in a volume of essays on Noam Chomsky, to whom it is addressed, Burge takes great pains to spell out his substantive agreement with Chomsky on this very issue. Semantics does need to take idiolects as its objects of study; competence is a matter of individual psychology. The thesis is fully compatible with the view that the contents of the individual’s thoughts do not supervene on the local, nonintentional, facts about her. In addition to this broad agreement between Davidson and Burge, then, we also find them in agreement on a somewhat more surprising thesis: In order to have thoughts or language we need other people whose perceptual and concept-forming abilities and dispositions are innately substantially similar to ours. Since the individuation of our meanings depends on this fact, we have a version of social externalism. Davidson believes this because he believes that interpretation, and in particular the assignment of objects as causally relevant in perceptions, requires it. Burge’s argument is that other, substantially similar people are required in order for creatures like us to be able to use concepts whose individuation depends, in part, on their extensions. But the conclusions are quite similar, more similar, I think, than the only very rough agreement concerning causes of perceptions. Still, there is a large difference here. Davidson is an externalist because he thinks that without interpretation, there is no content or meaning. The notions are essentially interpersonal, since interpretation is constitutive of contents and meanings. Burge is a realist about content and considers its conceptual components in creatures like us to be fixed in the first place by perceptions of objects and the nature of the objects perceived. For Burge, it seems that there could be a private language, and certainly there might be private thoughts (and animals might have thoughts). But as a matter of psychologically necessary fact about critters like us, there aren’t any such thoughts, and human language is necessarily social. This essential difference between the two philosophers is nicely illustrated by the following pair of quotations:
When Swampmen Get Arthritis
55
The problem is not, I should stress, one of verifying what objects or events a creature responds to; the problem is that without a second creature interacting with the first, there can be on answer to the question. And if there can be no answer to the question what a creature means, wants, believes or intends, it makes no sense to hold that the creature has thoughts. (Davidson 1989, p. 198) It is metaphysically possible for an individual to learn his idiolect in isolation from the community. But it is no accident that one obtains insight into an idiolect by considering the usage and attitudes of others. In learning words, individuals normally look to others to set standards for determining the range of legitimate examples and the sort of background information used in explicating a word or concept. I believe this is a psychological necessity for human beings. (Burge 1989, p. 186)
Davidson’s version of social “externalism” is then a priori, based on an a priori argument grounded in the methodological strictures on meaning and interpretation of the Davidsonian program. Burge’s, on the other hand, is the result of philosophical arguments concerning how we actually come by, use, individuate, and correct our usage. The conclusion is meant to be empirical, thus revisable. I have argued that the standard account of the relationship between Davidson’s externalism and Burge’s anti-individualism is mistaken. According to this account, the two philosophers agree on “physical externalism” and disagree on social. We have found that the only straightforward agreement between them is on the importance on perceptual interactions with objects around us for having beliefs at all. On theses about individuation, little agreement can be found. Burge asserts, and Davidson denies, that our concepts are in part individuated by the kinds that they refer to and that we do not fully grasp the contents that we think with: We stand to be corrected by the facts of the world as well as by others. Davidsonian externalism, on the other hand, asserts that a subject’s thoughts are individuated by reference to what the interpreter takes to be the normal causes of assertions and beliefs. The triangulation argument articulates this thesis further: Unless a subject is a member of a community of speakers who have, and mutually known to have, substantially similar concepts, she has no thoughts. Insofar as physical externalism is true, for Davidson, it is because of the constitutive role of the community. That is, insofar as my concept of tiger is dependent on the nature of tigers, it is because to count as a tiger is to be classed as such by a community of like-minded speakers (a redundant description, since there could be no other kind). For Burge, the converse is true. Insofar as my thoughts are individuated, in part, by the community of like-minded speakers, it is because I rely on them to help me individuate the kinds that my concepts refer to. The result is that not only does Davidson not deny something quite close to Burge’s “social externalism” while asserting the “physical” kind; insofar as he believes the latter, it is because he subscribes to the former.
56
Martin Hahn
Notes 1. It is not clear what Davidson means by the ‘object of a belief’ in the last quote. It doesn’t seem that it straightforwardly means the ‘referent of the belief’. While looking at a cow, I utter the inevitable ‘Lo, a cow!’ The referent of that belief is, indeed, the cause of my cow-perception and thus of my perceptual belief that there is a cow in front of me. But this, while no doubt true, is not any kind of externalist thesis: no one (well, no nonidealist, at any rate) has ever thought that the referents of my thoughts about medium-sized dry goods were internal. Thus, when he says that the objects of a belief are its causes, he is using ‘object’ in a different sense from the one in the assertion that the contents of beliefs are partly constituted by the objects that cause them. What that sense is, I will not try to spell out here. 2. We may note a curious inversion in these two quotes. Burge, who is a nonreductionist realist about psychological states if there ever was one, writes as though the issue were a matter of “having adequate grounds,” when what he means is that without some applications of her thoughts to particular objects in her environments, a subject would have no intentional contents. Davidson, on the other hand, writes as though such interactions between thinker and object were necessary so that we might “discover” what she thinks, when it is in fact central to the views that interpretation is precisely not the discovery of some independently existing thoughts and meanings. Rather, meanings and thoughts are constituted by the process of interpretation. 3. Because an unambiguous statement of interpretationism is so hard to find, and because this is a thorny topic in Davidson interpretation, I have some misgivings about attributing to him the view that actual interpretation is constitutive of meaning and attitude-content. Nonetheless, this seems to me to be the most natural reading of Davidson, and so I will proceed on the assumption that it is, at least in its essentials, not far wrong. I will not address the interesting question of how much of the argument of the present paper would survive, mutatis mutandis, on a less interpretationist reading of Davidson. 4. It is not, however, necessary that members of the linguistic community actually share concepts. Burge’s argument is meant to show that language is social in at least one sense, even on the Chomskian assumption that we eschew the notion of a shared community language. Even if, in our idiolects, we all used somewhat different concepts, we would still need others to help us get access to their common referents. 5. Interestingly, Davidson seems to think this argument is against social individualism, while, for example, Burge (1986c) is concerned with the causal or perceptual thesis. How he comes to this somewhat startling conclusion is not explained. Nor is it clear why he ignores what seems to be Burge’s most congenial work to his own thesis: Burge (1977). See “Epistemology Externalized,” p. 197. 6. The Swampman has suffered an analogous fate in his application to another form of “externalism,” the biological account of intentionality defended in various forms since Millikan’s seminal work. While some philosophers, notably Millikan and Dretske, consider the Swampman to be just the sort of creature that illustrates their thesis, others (for example Pappineau and Neander) have taken him to provide a potential counterexample; one that needs to be countered by a powerful argument. Of course, it is precisely such an argument for a dependence of intentional content on evolutionary history that the story about intentionality and proper function is meant to supply. It is no “mere reflection on how biological function arises and what we mean by intentionality” that will get us the result. 7. Real-life examples like these seem to me to put some pressure on Davidson’s denial of the possibility of conceptual relativism. It seems straightforwardly true that John’s concept of cod is different from the concept his Newfoundland friends employ—whether we think this is because Atlantic cod are different species from their Pacific namesakes, or we chalk the difference up to some highly distributed, holistic facts about the two linguistic communities. But a strong reading of Davidson’s strictures on interpretation makes it necessary for West Coast interpreters to translate Atlantic-English utterances as expressing Pacific-cod contents since, after all, that is the concept of cod they have. That is, the very arguments used to deny the possibility of conceptual relativism seem also to deny commonplace conceptual differences between actual linguistic communities. A subtle defender of Davidson’s views can, no doubt, find relevant differences. But there is work to be done here. 8. Contrary to what I have suggested in an earlier note, it would seem that Davidson does, at least sometimes, mean by the ‘object of thought’ whose individuation is externalist simply the referent of the thought. The passage
When Swampmen Get Arthritis
57
strongly suggests that it is the causes of our beliefs—cows, or perhaps the kind cow—that are individuated socially. 9. One might object that there is no way to describe behavior nonintentionally. Indeed, Burge is committed to this view, I believe. But if the learning history is supposed to be the evidential basis for, or even constitutive of, intentional ascriptions, it had better be described nonintentionally, on pain of circularity or infinite regress. 10. The triangulation argument can then be seen as reinforcing this point: Meaning exists only if interpretation occurs, but the latter can occur only if subject and interpreter have substantially similar ways of carving the world up into similarity-classes and know this about each other. In other words, meaning can arise only in a community of substantially similar communicators. 11. I don’t believe that there is anything in Burge’s published works to commit him to such a Davidsonian view. We would be hard put to understand how a creature like a Swampman could have intentional content, given its lack of history of perceptual interaction with objects in the world, but that need not make us conclude that it is impossible. Even the arguments of Burge (1977) taken at their strongest have only the conclusion that no creature without de re attitudes could have intentional content. But the second the Swampman opens his eyes, he is interacting with objects in the world. Perhaps the conclusion is that the Swampman cannot dream before having some visual experience. 12. When he does articulate his reasons (1989a, pp. 447–449), he says that Burge’s “social externalism” commits him to denying first-person authority, because it commits him to the thesis that we have thoughts whose contents we do not fully grasp. This is, indeed, a central thesis of Burge’s—arguably the central thesis of his antiindividualism. But it is one that characterizes his “physical externalism” equally well. Furthermore, Burge’s own account of why this thesis does not conflict with first-person authority is strikingly similar (mutatis mutandis irreconcilable differences about the role of interpretation) to Davidson’s own account of externalism and firstperson authority in the very paper where his rejection of Burge’s “social externalism” is presented.
References Burge, Tyler. 1977. Belief De Re. Journal of Philosophy 74(6): 338–362. ———. 1979a. Individualism and the Mental. Midwest Studies in Philosophy 4: 73–121. ———. 1979b. Sinning against Frege. Philosophical Review 88(3): 398–432. ———. 1982a. Other Bodies. In Thought and Object, Andrew Woodfield (ed.), 97–120. Oxford: Oxford University Press. ———. 1982b. Two Thought Experiments Reviewed. Notre Dame Journal of Formal Logic 23(3): 284–293. ———. 1986a. Cartesian Error and the Objectivity of Perception. In Subject, Thought, and Context, John McDowell and Philip Pettit (eds.), 117–136. Oxford: Oxford University Press. ———. 1986b. Individualism and Psychology. Philosophical Review 95(1): 3–45. ———. 1986c. Intellectual Norms and Foundations of Mind. Journal of Philosophy 83(12): 697–720. ———. 1988. Individualism and Self-Knowledge. Journal of Philosophy 85(11): 649–663. ———. 1989. Wherein Is Language Social? In Reflections on Chomsky, Alexander George (ed.). Oxford: Blackwell. ———. 1992. Philosophy of Language and Mind: 1950–1990. Philosophical Review 101: 3–51. ———. 1993. Mind-Body Causation and Explanatory Practice. In Mental Causation, John Heil and Alfred Mele (eds.), 97–120. Oxford: Clarendon Press. ———. 1996. Our Entitlement to Self-Knowledge. Proceedings of the Aristotelian Socicety 96: 91–116. Davidson, Donald. 1975. Thought and Talk. In Mind and Language, S. Guttenplan, Oxford: Oxford University Press. ———. 1982. Rational Animals. Dialectica 36: 317–328.
58
Martin Hahn
———. 1983. A Coherence Theory of Truth and Knowledge. In Kant oder Hegel? D. Henrich (ed.). Suttgart. Also in 1986. Truth and Interpretation, Lepore (ed.). Oxford: Blackwell. ———. 1984. First Person Authority. Dialectica 38: 101–112. ———. 1986. A Nice Derangement of Epitaphs. In Truth and Interpretation: Perspectives on the Philosophy of Donald Davidson, E. Lepore (ed.). Oxford: Blackwell. ———. 1987. Knowing One’s Own Mind. Proceedings and Addresses of the American Philosophical Association 60: 441–458. ———. 1988. Reply to Burge. Journal of Philosoply 85: 664–666. ———. 1989a. The Conditions of Thought. In The Mind of Donald Davidson. Amsterdam: Rodopi. ———. 1989b. The Myth of the Subjective. In Relativism, Mickael Krausz (ed.), 159–172. Notre Dame: University of Notre Dame Press. ———. 1991a. Epistemology Externalized. Dialectica 45: 191–202. ———. 1991b. Three Varieties of Knowledge. Philosophy 30(supp): 153–166. ———. 1991c. What Is Present to the Mind? In Philosophical Issues VI: Consciousness. Enrique Villanueva (ed.). Ridgeview: Atascadero. Putnam, Hilary. 1975. The Meaning of ‘Meaning’. In Language, Mind and Knowledge, K. Gundorson (ed.). Mimeapolis: University of Minnelote Press.
4
Burge Thought Experiments
Keith Donnellan I My theme is that Tyler Burge has developed thought experiments quite different from Hilary Putnam’s Twin Earth examples. To set the stage, let me quote from a recent paper of Burge’s, “Mind-Body Causation and Explanation” (Burge 1993, p. 104). In that paper he is concerned with the problem of whether or not “we can reasonably believe that mental properties are causally efficacious.” That will not be a problem I will directly address here, but, in the course of an argument where he is taking up the “token-identity” theory of mental states and brain events, he says: The argument against this sort of token-identity theory is partly based on Twin Earth thought experiments that I shall presume are familiar. According to these thought experiments, it is possible for a person’s body, considered in isolation from its relations to the environment, to be physiologically and molecularly the same even if the person were to think thoughts that have different intentional content. The difference in content depends on differences in the individual’s historical relations to his or her environment. (Burge 1993, p. 104)
Now in a footnote to the first sentence of this passage, Burge not only cites earlier articles, but says that “the thought experiments use the methodology set out in Putnam.” He then cites Putnam’s “The Meaning of ‘Meaning’” (Putnam 1975),1 the original source for Putnam’s Twin Earth thought experiments. Here, as in other places,2 Burge seems to me to conflate the two sorts of thought experiments. In doing so, he does himself an injustice. He has developed, I think, a new kind of thought experiment based on different principles from those of, as I will call them, Putnam thought experiments. Perhaps Burge was not just being modest, but thought that it did not much matter whether there were central differences between his and Putnam’s thought experiments— he was after bigger game, an attack on what he calls individualism, and all that entails. Perhaps both kinds of thought experiments are equally good for this purpose—both might show that the content of mental processes such as thought and belief is not fixed by what goes on in one’s head (or within the confines of one’s body) but also depend on the external circumstances in which one is placed. In the cases of Putnam, it would be the underlying physical properties of what a person has been calling ‘water’ or ‘tiger’ or ‘gold’ or ‘arthritis’; in the cases of Burge, it would be the surrounding linguistic community that uses those terms as well as terms such as ‘sofa’ and ‘contract’. The content of one’s thoughts, beliefs, and so on in either case may then seem to be not a matter of what materials are in the head but as well what goes on in the surrounding environment. So, it is plausible to think that both sorts of thought experiments can be used to show that
60
Keith Donnellan
individualism is wrong, or to argue, for example, that there is something wrong with Descartes’s problem from the Meditations that seems at once to assume that whether or not there is an external world, an environment in which we live and operate, our thoughts and beliefs are in place and the only question is whether or not they have an external world to refer to. If the content of our thoughts and beliefs depends on what is the case in an external world, then something seems wrong about the Cartesian setup. Perhaps then, for these reasons, Burge saw no reason to make heavy weather of the differences that exist between his and Putnam’s thought experiments. Still, there is a difference, I think, and a profound one at that. So, what are these differences between Burge and Putnam thought experiments? I will be bold and list them: First, the environment that makes the difference in Burge thought experiments is linguistic, the surrounding linguistic community in which one lives. In contrast, the Putnam environment is the physical world of the underlying structure of the stuff, animals, or whatever one has thoughts and beliefs about. Sofas and contracts, Burge examples, will not have a place here, although arthritis may. Second, experts or authorities are connected with the two thought experiments in quite different senses. In Putnam’s, the “experts” are experts about the physical world, not about what words mean: in Burge’s, deference is given to those in a speech community who are deemed in a position to know the correct use of words. Thus, a Putnam thought experiment with arthritis as subject might involve medical scientists, but only insofar as they are thought to have the truth about the nature of arthritis; in fact, for Putnam experiments, it is only the truth about the nature of subject that need be brought in—experts are only a convenient means of referring to it. A Burge thought experiment might involve doctors as arbiters of the correctness of use of a medical term or phrase. Third, Putnam thought experiments can be thought of as showing a semantical rule that might in fact be a semantical rule that lies behind each of the sorts of examples he gives. Burge’s thought experiments, on the other hand, do not seem to have this property. This is an odd result. Putnam has often been taken (for all I know this was his view) to be telling us that “meanings are not in the head.” But in fact what he gives us is something that could be “in the head.” Not, to be sure, the sort of rule that Locke or Hume would have given us—but nevertheless one that would in fact determine the extension of a term. Burge thought experiments, however, do not seem to yield such a rule—except in a very general sense. At this juncture, Burge thought experiments seem more likely to be antiindividualistic than do Putnam’s. I will discuss this important point later. Fourth, and finally, Putnam thought experiments involve a deep indexical character. I say “deep” because I agree with Burge that terms such as ‘water’ and ‘gold’ are not on the surface indexical. But underneath there is an indexical element, an indexical reference to the local paradigms of what we locally call ‘water’, ‘gold’, and so on. This indexical
Burge Thought Experiments
61
feature makes Putnam thought experiments work in the Twin Earth cases. It is not, to be sure, necessary for his theory of natural kinds, but it is necessary for his thought experiments. (Whether or not there is such an indexical element in our real life use of ‘water’, ‘gold’, etc. I am not concerned with here.) I do not see any such indexical element in Burge thought experiments. This is not a criticism, but only a note of an important difference. II Let me try to justify finding these differences between Burge and Putnam thought experiments by looking first at Putnam and then at Burge. About Putnam I have the feeling that he has been misunderstood and that he himself did not get it all straight.3 A well-known consequence Putnam draws from this view of the nature of a large number of terms for kinds is that the extension of such terms is not determined by the psychological states of users of them. If we choose a time prior to the discovery of the molecular structure of the stuff in the lakes or oceans of Earth and Twin Earth and/or pick a person from Earth and his or her counterpart on Twin Earth who are scientifically naive, the word ‘water’ as each uses it will have a different extension even though their psychological states may be identical. (As we will see later, we need to be somewhat careful, however, how we construe ‘psychological state’ and what it is for such a state to ‘determine’ the extension.) In his papers on the topic, Putnam has emphasized what he takes to be the central role of indexicality in obtaining this result. He very early on, perhaps incautiously, said flatly that “Water is indexical,” (Putnam 1974, p. 451) and in “The Meaning of ‘Meaning’,” less incautiously I think, he tells us, “Our theory can be summarized as saying that words like ‘water’ have an unnoticed indexical component: ‘water’ is stuff that bears a certain similarity relation to the water around here” (Putnam 1975, p. 234). It is possible that Putnam thought of the claimed indexicality of words like ‘water’ as explaining how the psychological states of an Earthling and a Twin Earthling could be the same although the extension of the term ‘water’ for each is different. At least this is a reasonable impression one can get from reading “The Meaning of ‘Meaning’.” If only it were true, it would be a tidy explanation. For indexical words, such as the first- and secondperson pronouns and words such as ‘here’ and ‘now’ in their “pure” indexical uses, seem to have this feature. In an important sense two people could be in identical psychological states and yet their use of ‘I’, for example, would obviously have a different referent, a different extension.4 A quick and, I believe, correct response to this reading of Putnam is that the word ‘water’ and other examples of the same sort simply are not indexical in the required sense. This is Burge’s answer in “Other Bodies.” To be an indexical, is, to quote Putnam, “. . . to have
62
Keith Donnellan
an extension which varies from context to context or token to token” (1975, p. 234). And Burge says, “I think it is clear that ‘water’, interpreted as it is in English, or as we English speakers standardly interpret it, does not shift from context to context” (1982). About this I think he is absolutely right. Nathan Salmon in Reference and Essence makes a similar point (1981, part IV). Salmon goes on as well to suggest, correctly, I believe, that Putnam recognized that ‘water’ and other such terms are not themselves straightforwardly indexical. And I think Salmon is also correct when he points out that the importance of indexicality in the Putnam thought experiments really shows up at a deeper level, at the level of “explaining the meaning of natural kind terms” (p. 106).5 Salmon was more concerned with the modal consequences to be drawn from them, and the point is worth exploring again in connection with our concerns here. Where, then, might indexicality come into our use of a term such as ‘water’? Let me try to reconstruct an argument that puts off indexicality until the last possible moment, the argument that I think is really involved in the derivation of the moral. The first step is perhaps the most revolutionary of all in Putnam’s use of Twin Earth examples. It is a new way of thinking about how a term for a kind might determine its extension.6 Let us think of this as a new kind of semantic rule. The semantic rule for ‘water’, for example, is either just the one Putnam gives in “The Meaning of ‘Meaning’ ” or some close relative. I am inclined to suggest, for the term ‘water’, for example, a somewhat modified version of Putnam’s various formulations: something like, “Water is the stuff that has the important underlying physical characteristics of the stuff we (or I) have been used to calling ‘water’”—this is close to the way I put it in “Kripke and Putnam on Natural Kinds” (1983). The idea is that what “we” have been calling ‘water’ loosely gives us the paradigms of water, and that we have been calling bits of stuff ‘water’ on the basis of certain surface properties, for example, color, liquidity at ordinary temperatures, potability, and so on (it is the latter fact that makes the rule noncircular). The indexicality involved is not an immediate semantic property of a word such as ‘water’, but comes in as an internal part of the rule for applying the term. In my version of the rule I suppose it comes in the use of ‘we’ or ‘I’ in the phrase “the stuff we (or I) have been used to calling ‘water’.” However one specifies the Putnam rule, the indexicality involved will come in via a reference of some sort to local paradigms. This is what allows him to generate his Twin Earth examples: two cultures speaking an identical language except for one term, psychophysically identical but—if we agree with Putnam’s intuitions—using the term with either a different meaning or with a different extension for the same meaning. At this point I may as well show my hand about Putnam. Suppose that what Putnam has given us is a new kind of semantic rule, a rule that governs or can govern certain of the common nouns and predicates of a language. There is no reason they should not be in
Burge Thought Experiments
63
the head. In fact, insofar as we have a strong inclination to believe that what is called ‘water’ on Twin Earth is not water because it does not have the molecular structure we now think we know to underlie ‘water’ as we call it on Earth, we seem to be attributing to Twin-Earthlings the same semantic rule. Crudely, “call it ‘water’ if it has the same underlying structure of (indexical!) paradigms of what we call ‘water’.” It seems that Putnam’s thought experiments depend on a commonality of meaning “in the heads” of Earthlings and Twin-Earthlings. The meaning of such common nouns is nothing like the classical view, of course; it is a rule that directs us to the nature of things, rather than giving us directly the properties to use when deciding whether a certain term does or does not apply. There is, however, no reason to think that new and novel as it might be, such a rule cannot be “in the head.” What is not in the head are straightforward properties used to determine whether, for example, this before me is water or a tiger. The rule in the head only directs us as to how to find such properties. Nevertheless it is a rule that if followed will ultimately yield the answer if nature is cooperative. In that sense it determines the extension of the term one uses. If all of this is correct, Putnam thought experiments, it would appear, cannot be used for an anti-individualist point. There is no reason, for example, Descartes cannot have a thought about, say, water, and, Cartesian that he is, skeptically wonder whether there is anything external to his mind corresponding to his thought. We successfully used the word ‘water’ to talk about a certain substance before we discovered that water is H2O. So too Descartes could use a word to formulate a thought about it even if there is no such substance because, although the Putnam rule directs him to an external world, as we might say, in order to discover what makes something water, it gives him the use of a term for water before he has followed the directions. In any event the important feature of the rule will be that the extension of the term is determined by the underlying nature of some sets of paradigms, and the paradigms are fixed by the rule itself. But what the paradigms, underlying nature consist in, which in the end determines the extension of the term, is not. That is left for empirical research. This is, of course, quite opposed to the classical view, the one we find in Locke, that all kind terms are determined by rules directly accessible to the mind (Locke’s “nominal essence”) that directly specify necessary and sufficient conditions for being a member of the extension of the term. In speaking of a semantical rule, I see no bar to viewing it as “in us” at least in the sense in which linguists think of the often complex rules of grammar as being in us (without the further view of Chomsky’s that they might be innately in us). Most of us feel that the fact that only a stroke of inspiration has allowed us to formulate the rule would not be an objection. The nature of the semantic rules for demonstratives and indexicals analogously has not been easy to discover, but surely in some sense we follow such rules.
64
Keith Donnellan
That in some important sense the rules are followed by us in the case of natural kinds is perhaps attested to by the force of the intuitions about what we would say in this or that imagined case, intuitions appealed to, for example, by Putnam in defense of the conclusions he draws from the Twin Earth examples. Even if it is only with great difficulty that the semantic rules for natural kind terms such as ‘water’ and ‘gold’ are uncovered, it may be that in principle they are accessible to us apriori, although certainly not in the continually immediate manner in which Locke’s nominal essence, the content of ‘abstract ideas’, was supposed to be. But where does indexicality enter for the startling result we are interested in here, the possibility of two people in the same psychological state with words in their vocabulary having different extensions? If it comes in at all, it must be, of course, in the specification in the semantic rule of the paradigms. Perhaps (roughly), ‘water’ is whatever has the underlying nature of the stuff in our lakes and oceans. Even here, however, indexicality is not strictly essential. A rule such as we are contemplating might involve no indexicals at this point, but only a purely descriptive definite description denoting the paradigms.7 I think that regarding the paradigms as indexically specified becomes important, however, when we want to generate the Twin Earth examples. To reach the conclusions to be drawn from these examples, we want the semantical rule for ‘water’ on Twin Earth to be indistinguishable, except for the context in which it is applied, from the semantical rule on Earth. We want this in part at least to ensure that in saying that the psychological states of counterparts on the two planets can be the same, we are not immediately in trouble because the semantical rules for the two words are different. This is precisely because the semantical rules may plausibly be thought to be “in the head” of the users of the word, as I suggested above. If it is even an open possibility that the semantical rule for ‘water’ is in the heads of Earth and Twin Earth users of the word, then the rule had better be the same if we are to keep their psychological states the same. At the same time we want the paradigms associated with the word ‘water’ on Earth to be different from the paradigms associated with the word on Twin Earth. Now if the paradigms were specified by each of the rules in purely nonindexical, qualitative terms, it seems highly unlikely that the rules could be the same. Twin Earth and Earth are, by hypothesis, identical except for the underlying nature of the stuff called ‘water’. Without some indexicality showing up in the rules’ specifying paradigms, it is difficult to see how there could be different paradigms, if the rules are to be the same. Now this question of indexicality, which arises prominently as it does in Putnam Twin Earth examples, is intimately connected to my view that Burge thought experiments are different. As I have said, I see Putnam Twin Earth examples as giving us a semantic rule for the use of the term, albeit not anything like the classical view of Locke, Hume, or the
Burge Thought Experiments
65
positivists. Indexicality, although not necessary for a Putnam-like semantic rule, is nevertheless necessary to generate his Twin Earth thought experiments. With Burge thought experiments, the case seems to me somewhat different. The first question is whether they involve a semantic rule at all. With Putnam-like examples there is no reason to suppose that the semantic rules are not in the head and that they determine, if only in an indirect way, the extension of the term they are associated with. Indeed, despite what Putnam may be thought by many to have said, his account of how terms such as ‘water’, ‘gold’, and ‘tiger’ function almost demands that there be a semantic rule that the user of these words in some sense is in command of. I do not think this to be true of Burge thought experiments. Let me give a quick summary of Burge’s famous ‘arthritis’ thought experiment. We are to imagine an ordinary English-speaking person who has some command of the English word, ‘arthritis’, although not a complete mastery. Having a pain in his thigh, he believes, and we are to intuit that we would agree that he so believes, that he has arthritis there. (Of course, in agreeing that he has this belief, we are in no way committed to the truth of nor even the coherence of the belief.) Told by a doctor that one can only have arthritis in a joint, he sees that his previous belief was false; he defers in this way to a somewhat special member of his speech community. But, counterfactually, we can imagine the same person (or his counterpart) having the same history and (in an important sense) psychological states—up to the point of talking to the doctor—in a community where the word ‘arthritis’ is given a wider use, a use that allows that arthritis can occur in soft tissue. In such circumstances, although the subject has not changed, by hypothesis, he now may well have, or so our intuitions should tell us, a true belief. The beliefs of the actual person and his counterfactual counterpart must then be different beliefs, although everything but the surrounding speech community is the same. We should be sure, I think, that we are clear what deep point Burge is trying to make. It is not, for example, that the meaning of one’s words is necessarily dependent on the surrounding speech community. The essential principle involved in the thought experiment is, I believe, the one Burge emphasizes in his paper when he says, “the thought experiment does appear to depend upon the possibility of someone having a propositional attitude [belief, for example] despite an incomplete mastery of some notion in its content” (1979). This, I think, is a very important principle in its own right, and Burge correctly spends time to point out that it represents an apparently pervasive phenomenon, not restricted to examples with any peculiar properties. In regard to its role in the Burge thought experiments, we can view this principle as the first of three premises that together yield the results of the experiments. The second premise we need will tell us something about the standard whereby complete (or at least
66
Keith Donnellan
greater) mastery of a notion is to be judged. In the counterfactual situation of the thought experiment, the subject of the experiment is thought of as being in a community where the word for the concept involved as used by those with a complete or greater mastery differs from the use made by their counterparts in his actual community. In this sense the speech community provides the standard whereby a speaker can be said to have only incomplete mastery of a notion. A final premise needed asserts that when a person can be said to have incomplete mastery of a concept judged against one standard for mastery, that person unchanged in respect to all relevant traits of behavior, history, and so on may be said to have an incomplete mastery of a different concept when judged against a different standard of mastery. In the Burge examples, the speech community changes in the counterfactual situation, and the subject, unchanged, is credited with an incomplete mastery of the different concept designated by the same word in the new speech community. Once we have these premises, the conclusion we are to draw from the Burge thought experiments seems to drop out immediately. It is of some interest to see that the second premise, the one specifying the standard for mastery, is at least theoretically independent of the other two and may not involve a speech community at all. A somewhat fanciful analogue argument may help to show this. Plato, I suppose, could be thought of as holding the first premise. Insofar as we can be said to have beliefs about justice or piety, we are said to have these beliefs in spite of the fact that, unless we are completely enlightened philosophers, we have only incomplete knowledge of the Platonic forms. And knowledge of the relevant form provides the standard for complete mastery of a concept, and such knowledge would be what is mentioned in a Platonic version of the second premise. Thus Plato would agree with the first premise, it seems, but provide a different version of the second. (The historical Plato would have trouble, I think, with what I count as the third premise. There is just one form of justice for him, and ordinary people have only a partial mastery of the concept of justice because they lack knowledge of the form itself. This makes it hard to see how two people in the same relevant states could have a partial mastery of different concepts. But perhaps we could imagine a fanciful relativized version of Platonism according to which the gods have provided different forms of justice for different nations of people. Then two counterparts in the two nations might have partial mastery of different concepts.) I am inclined to think that this shows that it is the first premise, the principle that we attribute beliefs to persons who have only a partial mastery of concepts involved in the beliefs, which lies at the heart of the Burge thought experiments. The community use of a term as a standard for complete mastery is a very plausible and possibly true version of the second premise. But one can imagine other, not absolutely implausible versions of the second premise that would, if true, yield the same results.
Burge Thought Experiments
67
III Are the Burge and Putnam thought experiments merely variants of the same fundamental phenomenon? I think not—not if we are thinking of Putnam’s Twin Earth examples and his treatment of natural kind terms. But Putnam in “The Meaning of ‘Meaning’” introduces a principle, what he calls the “principle of the division of linguistic labor,” which may well have a similarity to the use of the surrounding speech community as a standard form of complete mastery of concepts in Burge examples. The confusing thing is that Putnam introduces this principle as if it were something like a generalized version of what we are to conclude form the Twin Earth examples. Yet, as I view it, it is an entirely different matter. The division of linguistic labor has to do with the fact that for many items in a person’s vocabulary, not at all restricted to terms for possible natural kinds, he or she is prepared to defer to the use of the word as represented by other people in the speech community. And there is, I think, the suggestion in Putnam, that in such cases the person may be said nevertheless, in a quite ordinary sense, to know the meaning of the word. Thus I have the word ‘carburetor’ in my vocabulary, but if you ask me whether the fairly new fuel injection systems are carburetors, I may be uncertain. This might be because, for example, I am not sure whether something is a carburetor just in case it delivers a mixture of gasoline and air to the cylinders or whether it is one only if it does so in a certain way. A good mechanic can probably enlighten me as to how to use the word in this respect. The principle of the division of linguistic labor may conflict with the classical view of what a word in a person’s vocabulary means and according to which what I have in my head at any moment determines what I mean by ‘carburetor’ and in which no room is provided for the possibility that I tie my words in this way to the usage of others. And since I think that my uncertainty about a particular application of the word ‘carburetor’ does not prevent me from having thoughts and beliefs that are correctly reported as, for example, ‘Donnellan believes that his carburetor needs adjusting’, the principle seems to give us a start toward the conclusion Burge reaches by supplying the perhaps most plausible version of what I have distinguished as the second premise. In fact, however, the principle is not one involved in the Twin Earth examples and their employment in connection with natural kind terms. The essential difference, I believe, is that what the Twin Earth examples show is the possibility of a new kind of semantic rule for a term utterly different from the classical model, the model in which the extension of the term is determined by a set of necessary and sufficient conditions completely at the command of either the user of the term or at any rate of the speech community in which the term is employed. The revolutionary idea is that of a semantic rule that employs paradigms and their underlying nature, a nature that may not even be known to any users of
68
Keith Donnellan
the term. Nothing in the principle of the division of linguistic labor yields such a result. Nor, it is important to add, do the Burge thought experiments. One thing that can obscure this fundamental difference between what is going on in the Twin Earth examples and what is involved in the principle of the division of linguistic labor is that about both there can be and has been a loose use of the notion of “experts.” My mechanic is an expert about the use of the word ‘carburetor’ in a way in which I am not and I am prepared to defer to him. Scientists, of course, may well be experts about the underlying nature of water and I may also defer to them when I agree that because of their conclusions, nothing can be water, whatever I may have thought before, unless it is H2O. But the use of these “experts” in determining the extension of my terms is utterly different. In the case of ‘water’, it is the nature of the paradigms that determines the extension. There is nothing in the new notion of the semantical rule introduced by Putnam Twin Earth examples that has to do with experts. Scientists come in only insofar as they happen to have opinions about or knowledge of the nature of the paradigms. Scientists are “experts” not about the use of a word or expression, but as conveyers of the truth about the nature of the paradigms that the semantic rule associated with my use of the word ‘water’ alludes to. It is their experiments, not they, that are crucial. (They need not be in fact a part of my speech community at all.) What I defer to, one might say, are the experiments and observations of scientists. In the case of water what I defer to has nothing to do with members of my speech community and everything to do with the results of putting electrodes into representative samples of water and getting two parts by volume of hydrogen at the cathode and one part oxygen at the anode.9 Another way of seeing the independence of these theses is to look at how they confront a classical theory of semantics such as that of Locke. Locke held that the extension, as we would say, of a term is determined by the “nominal essence,” a set of necessary and sufficient conditions for the correct application of the term that he also held were the property of each user of the term, that user’s term being governed by the nominal essence in his or her mind. The principle of the division of linguistic labor conflicts with Locke’s view only on this last point—the nominal essence may be the property of the speech community and each individual’s term may have its extension governed by the speech community. But the Twin Earth examples rebut Locke on the first point; they introduce the possibility of a semantic rule completely antithetical to Locke’s picture of a set of necessary and sufficient conditions that require no investigation of the nature of the things themselves.10 Insofar as we compare Putnam and Burge, it is then the principle of the division of linguistic labor in Putnam that forms a possible bond. The Twin Earth examples lead us to accept a new kind of semantical rule; the Burge examples do not have this outcome. This is not to denigrate the latter, only to say that there is a very important difference. The Burge examples, as we have seen and as he emphasizes, depend on the idea of a partial mastery of a notion or concept. The Twin Earth results of Putnam, it is important to note,
Burge Thought Experiments
69
do not. We do not need at all to imagine people who have a less perfect grasp of a notion than their confreres. In John Locke’s day, no one knew that water is H2O or that gold has atomic number 79, but for all that they could all have had the very same semantic rule governing those terms in their vocabulary that we do today. The point is that at a similar stage on Twin Earth with the same semantic rule, the extension of the terms would have been different because the indexical specification of the paradigms by the rule would lead to different underlying natures. IV In his attacks on the view he has called “individualism,” Burge has argued that both his own and Twin Earth thought experiments show how the content of our propositional attitudes is dependent on the external world, that what is wholly within our minds, as one might put it, is not sufficient to determine, for example, what we believe. I want to suggest that the distinctions drawn in this paper between the two kinds of thought experiments suggest that the Twin Earth examples have more radical consequences in this regard than do Burge’s examples. Because the underlying principle behind Putnam’s views involves a semantic principle and Burge’s involves, rather, principles about the attribution of beliefs, it is easy to construct a Putnam-like principle that has nothing to do with belief attribution, social phenomena, and, yes, even natural kinds. If this is all that is needed to defeat “individualism,” then we can do it without any assumptions about social practices or the intuitive notion of natural kinds. Nothing so grand as that is involved at all. Burge examples, on the other hand, involve us in much more difficult matters. Here is an example of the semantic operation of Putnam’s principle that seems at once to have little to do with “natural kinds” but at the same time to determine a kind by how things are in the “external world.”11 It seems to show how easy it is to generate the Putnam mechanism (which is not to say that it was easy to propose the mechanism either as Putnam does or in the more generic way Kripke does). You and I have before us a copy of the Shorter Oxford English Dictionary. We agree to use the word ‘widget’ in the following way: You will open the dictionary at random three times. Each time, we will look for the first entry designating a property physical objects can have (i.e., we will not count “even” in the sense of divisible by two). A ‘widget’ is any physical object that has all three properties. It is obvious, I think, that we could develop a full-blown Putnam principle for ‘widget’, making it clear that the properties that are essential to widgets are those discovered from the OED we consult in the actual world, that if our copy of the OED is strange and has no such entries what the fall-back position will be about our use of ‘widget’, and so on.
70
Keith Donnellan
In this very simple way we can at once get the modal results—that, supposing the three entries are ‘round’, ‘red’, and ‘cubical’, a widget is necessarily round, red, and cubical. And we can get the epistemic result that this cannot be known apriori—we have to look at the dictionary to ascertain the properties. (Depending on your view of Kripke’s suggested possibility of the contingent apriori, we could also possibly derive the result that it is contingently apriori that a widget is anything that has the properties uncovered by the method of looking at the OED.) If the semantic features of Putnam’s ‘water’ case and others are sufficient, if correct, to defeat anti-individualism, then this simple example does it also. In the same sense in which you and I will not know the specific essential properties for widgethood until you have undertaken the dictionary search, we do not know, on Putnam’s account, the essential properties of ‘water’ until we have undertaken the investigation into the underlying structure, if any, of water. But I doubt that ‘widget’, with its quite artificial rule, would be thought to have much to do with any intuitive notion of a natural kind. I am inclined to conclude from the fact that the semantic features of Putnam’s examples can be reproduced in this way, first, that the semantic features are not sufficient (though still possibly necessary) for a term to be a term for a natural kind; and second, that if one thinks that Putnam has shown that the meaning of a term cannot be both in the head and determine the extension of the term, then one must also think that my example shows the same thing. In some sense the meaning of ‘widget’ is in the head—we have a rule, you and I, for determining the extension. What is true is that the direct, if I may put it that way, properties that make something a widget require us to follow out the rule—in this case to go through the dictionary search. Unlike the classical, Lockean view, the essential properties are not at the outset in our heads. We could, of course, use ‘widget’ in a perfectly respectable way even before you have looked at the dictionary. We could speculate about whether the cup on the table next to the dictionary is or is not a widget. Were I to have a strong hunch about what pages of a dictionary you might “randomly” select, I might even feel justified in my own conjecture. Nevertheless, in my parallel case, the rule for the use of ‘widget’ is clear even though it does not directly designate essential properties of the ordinary sort. V I would now like briefly to turn to Burge’s use of his thought experiments to shed light on the nature of propositional attitudes. In discussing this matter I will not assume much about the nature of belief and other propositional attitudes save that whatever the correct account is it will involve some relationship between the believer and the proposition
Burge Thought Experiments
71
believed. This relationship may be a simple two-place relationship or it may be more complex, involving, say, a mode of presentation. Burge speaks of an “incomplete mastery of concepts,” and up to this point I have not challenged the aptness of this as a description of the phenomenon I believe he has in mind. I now want to do so. All of us, probably, have many words and expressions in our vocabulary that we can use in assertions, requests, and even, it seems, in expressing our propositional attitudes, and yet for which we lack important details regarding their proper use. The phenomenon can occur with words that stand for natural kinds. But, as I have tried to emphasize, this is a matter entirely separate from the phenomenon pointed to by Putnam Twin Earth examples. To keep things as straight as possible, I want to use concepts that are, as far as we know, not concepts of natural kinds. And I want to consider at the beginning the sort of case in which our partial grasp is extremely limited. Imagine, then, the following sort of situation. John has been eating in a ritzy American “continental” restaurant whose conceit is to list its dishes in French. While he can make a reasonable stab at pronouncing them, John does not understand what the French words on the menu mean. Having enjoyed his first meal at the restaurant, he decides to sample more of the menu. He visits the restaurant a second time with Betty. Not wanting to appear ignorant, he does not ask the waiter to explain the menu for him, but simply orders new items from the menu. At one point he says to Betty, “I believe the champignons sautés à la crème should be good.” We can imagine that John is sincere: He believes from experience that anything served at this restaurant will be good. The situation is not outré, and John’s mode of expressing himself is neither unusual nor, I think, incorrect. John, however, is certainly deficient in some respect. It does not, however, seem appropriate to describe his deficiency as having only a “partial mastery of the concepts” the French words express. As far as the concepts go, he may in fact have as perfect a mastery as anyone of the concepts expressed by the English words ‘mushrooms sautéed in a cream sauce.’ What he does not have, of course, is a mastery of the meaning of the French words. (He may know something about their meaning, that, for example, ‘champignons sautés à la crème’ describes an edible dish.) Does John simply and straightforwardly believe that champignons sautés à la crème will be good? Certainly, the way in which he expresses his belief and the way in which we might ordinarily describe his belief suggest this. But I think both in a philosophical context are misleading. Speaking entirely in English, we may ask whether he believes that the mushrooms in cream sauce will be good. (For that is what the menu promises to serve.) It seems to me that he does not have this belief. I am strongly inclined to say that in the imagined circumstances, the belief he has is really something like this, that the dish called champignons sautés à la crème will be good, and that this is the belief he can be said to sincerely express even though he does not express it in this fashion.
72
Keith Donnellan
I might put my inclination by saying that the proposition that the mushrooms in cream sauce will be good is not itself any part of the content of John’s belief. Russell once spoke of knowing a proposition, not directly, but only by description.12 In a way, that is what I want to say about John. He believes that the proposition expressed by the words he used is true, but he does not have a direct belief in that proposition itself. It might be asked how it can be correct for John to express his belief in a seemingly straightforward way if in fact the belief he has is strictly speaking as I have claimed. I think the situation is not unlike one concerning the use of proper names that has had some recent discussion. Someone calls and leaves the message on my answering machine: “This is Mary. Have Catherine call me.” Hearing this but not knowing in any sense who called I nevertheless tell Catherine, “Mary called and wants you to call her back.” This is not, I think, incorrect, and in some sense expresses my sincere belief. But many, I think, would be inclined to say that what I really believe is that someone named ‘Mary’ called, and so on. Even one such as myself who believes that proper names have no Fregean sense or descriptive content, but refer via historical connections, may nevertheless think that in such a situation one is not in a position to apprehend the proposition that Catherine may apprehend on hearing the message I relay. What I said to Catherine in the circumstances I described does not directly express a belief of mine. There is a way in which I pass along to Catherine a proposition that I myself do not apprehend by saying the words that I am confident in fact express it. I could have directly passed on what I do apprehend and believe by saying something like “Someone named ‘Mary’ called,” and so on. If we can accept my inclinations in these cases for the sake of argument at least, then I wonder whether similar things should not be said about the Burge thought experiments. When the person expresses his belief to the doctor by saying “I think I have arthritis in my thigh,” thus displaying a less than perfect grasp of the concept expressed by the word ‘arthritis’, is it beyond doubt that the proposition expressed by the words ‘I have arthritis in my thigh’, as said by someone in complete command of the use of the word ‘arthritis’, is part of the content of the person who has only a partial grasp? Is it not that what he really believes is only that he has in his thigh the condition called ‘arthritis’, or something like that? Continuing to follow my inclination here, were this the correct account it would no longer be possible to argue from the Burge thought experiment in which an exact counterpart is imagined in a society in which ‘arthritis’ is used differently that different belief contents are involved. Or at least this would not be an immediate conclusion one could draw. In the example as detailed by Burge, both the person in the actual world and the counterpart would believe that they have in the thigh a condition properly called ‘arthritis’. And this identical belief would be true in the first situation and false in the second. Their external circumstances would not determine the content of their belief, but only the truth-value.
Burge Thought Experiments
73
Not everyone may agree with my inclinations in this matter, and I have not presented, I believe, more than a plausible possibility. Let me then draw only a modest conclusion. To pursue the course my inclinations take me along is at least a possibility for one who holds a view such as Locke’s. No such possibility presents itself when we confront the Putnam Twin Earth examples. For expressions that operate in accordance with the sort of semantic rule those examples imply, it is built into the very semantics that the identity conditions for the kind we have beliefs about is to be determined by investigation of the external world. It is not a matter of having an incomplete grasp of the semantics; the most complete grasp still leaves the job of investigating the nature of our paradigms to be done. And whether two people have beliefs about the same kind or not—whether in the fleshedout examples, they have the same or different beliefs—will depend not on what is in their minds alone, but on whether their paradigms have the same or different fundamental natures. If one admits that some of our kind words function in the way the Twin Earth examples seem to imply they do, the classical view represented by that of Locke must be abandoned. With the Burge thought experiments there seems, at any rate, a way to make at least a temporary stand. Acknowledgments The present paper was read at the conference on Tyler Burge’s work sponsored by Simon Fraser University in 1993. A somewhat different version was already commissioned and was published in Philosophical Perspectives 7, 1993. The present version is published here with the kind permission of the editor. Variants of the first sections were read at the University of Notre Dame and in a symposium with Professor Pierre Jacob at the Pacific Division 1989 meetings of the American Philosophical Association. A later version was read at the Central Division 1990 meeting with Professors Genoveva Marti and John Perry as commentators and subsequently at the University of California, Riverside. I have greatly benefited by the comments of the participants and the audiences in these meetings. I am also very grateful to Professor Charles Crittendon and Catherine Donnellan for helpful comments on an earlier draft. Extensive changes from the earlier versions have resulted, but obviously I am responsible for the remaining defects. Notes 1. Putnam acknowledges, correctly, a debt to Saul Kripke’s views as given in Kripke (1980). I believe, however, that Putnam’s views add important details and that he is incorrect in supposing that Kripke’s notion of a “rigid designator” is the locus of his debt. 2. I believe Burge first constructs his experiments in “Individualism and the Mental” (1979). In “Other Bodies” (1982) he argues that they are a variant on Putnam’s.
74
Keith Donnellan
3. In regard to my treatment of Putnam’s view on kind terms, I should warn the reader that I am here assuming, for the present purposes, that it is generally correct and that my stance is thus sympathetic. That something like Putnam’s view represents a possibility for language, I am convinced of. I do, in the end, have some reservations about Putnam’s actual formulation, some of which are contained in Donnellan (1983). 4. The point is elegantly made by David Kaplan in his example of the Castor and Pollux twins in Kaplan (1977). 5. In what follows I believe that I am probably neither adding to nor disagreeing with any deep points of Salmon’s extensive analysis of the role of indexicality in the Twin Earth examples. 6. The way of thinking about how the extension of a term is determined is new and revolutionary in its use of empirical paradigms and the resulting need for empirical research. As Charles Crittendon (in correspondence) pointed out to me, that the nature of something external to the mind may determine extension would seem, for example, to be a feature of the Platonic theory of forms. (I am inclined to think, of course, that the Twin Earth examples point to a viable way in which reference is fixed externally and that Platonism does not.) It needs to be pointed out as well that Putnam’s rules can be viewed as a special form of Saul Kripke’s device of using “reference-fixing” definite descriptions. 7. To make the semantic rule for the use of ‘water’ reflect the fact that the paradigms are available and to derive this result from the rule itself, we probably have to build in something like a reference to actuality, to the actual world—and this may involve indexicality. But that is not the startling result we are interested in here. 8. This is not to say that the division of linguistic labor cannot have an application to a natural kind term. The world ‘elm’ may be a term for a natural kind. I am prepared to defer to a knowledgeable gardener about whether the tree in my yard is an elm or not, although the gardener may know nothing about the structure of the DNA, if that is what the underlying nature of trees has to do with. 9. The principle of the division of linguistic labor has to do with deferring to the use of a term by other members of my speech community. I may thus defer for a variety of reasons—to be able to use the term as a certain favored group uses it; to use it in a way which for practical purposes will make it more useful. It may be that ‘carburetor’ would be a much less useful term if an artificial distinction were made between what is to be called a carburetor and what is called a ‘fuel injection system’. And my mechanic may have good reasons for using the term to cover the old-fashioned kind of device as well as the newer. These reasons may involve some expert knowledge of how these devices work. But it is not that there are paradigm carburetors and what I need in the end is the truth about their underlying nature. 10. One is reminded here of the situation with regard to the classical description theory of proper names, which often took the form of putting descriptions identifying the referent of a name in the head of each user. A move away from this was to suggest that the descriptions be obtained from the users, in the plural, of the name, from the community of users. This is analogous to the principle of the division of linguistic labor. It was a quite different attack on the classical theory of names when it was argued by Kripke, myself, and others, that the referent of a name might not be determined by a set of identifying descriptions at all. 11. The idea and the example are mine, but the idea arises from the problems about natural kind terms raised by Dominik Sklenar in discussion. 12. See, for example, Russell (1959), p. 57.
References Burge, Tyler. 1979. Individualism and the Mental. Midwest Studies in Philosophy 4: 73–121. ———. 1982. Other Bodies. In Thought and Object, A. Woodfield (ed.). Oxford: Oxford University Press. ———. 1993. Mind-Body Causation and Explanatory Practice. In Mental Causation, John Heil and Alfred Mele (eds.), 97–120. Oxford: Clarendon Press. Donnellan, Keith. 1983. Putnam and Kripke on Natural Kind Terms. In Knowledge and Mind, Genet and Shoemaker (eds.). Oxford: Oxford University Press. Kaplan, David. 1977. Demonstratives. In Themes from Kaplan, J. Almog, J. Perry and H. Wettstein (eds.). Oxford: Oxford University Press, 1989.
Burge Thought Experiments
75
Kripke, Saul. 1980. Naming and Necessity. Cambridge, MA: Harvard University Press. Putnam, Hilary. 1974. Comment on Wilfrid Sellars. Synthese 27: 451. ———. 1975. The Meaning of ‘Meaning’. In Mind, Language, and Reality: Philosophical Papers, volume II. Cambridge: Cambridge University Press. Russell, Bertrand. 1959. The Problems of Philosophy. Oxford: Oxford University Press. Salmon, Nathan. 1981. Reference and Essence. Princeton: Princeton University Press.
5
Anti-Individualism, Indexicality, and Character
Joseph Owens In 1979 Tyler Burge offered the first of a number of arguments for the startling claim that physically identical individuals can differ in propositional attitudes. This antiindividualistic conception of the mental did not sit well with many—it threatened a variety of deeply entrenched metaphysical and epistemological intuitions—and there have been numerous attempts to undermine or at least diminish the force of these arguments. In this essay I want to examine one prominent line of response, the response that the problems posed by these twins can be resolved and the individualistic character of the mental retained once we realize that the concepts they use in expressing their beliefs have an indexical element. In an early paper, Burge argued against this line of response, but the temptation to suppose that the answer lies in an appeal to some kind of indexicality persists to this day (Burge 1982). I want to explore this strategy, to see exactly why it continues to have the appeal it has, and to evaluate its prospects. I will argue that Burge’s initial assessment is correct; there is no way to meet the challenge posed by Twin Earth examples. I In this section I explore the arguments for the claim that the concepts employed by the twins have an indexical element, and indicate how this is supposed to resolve the problems posed by the twins. First, however, a few remarks on the anti-individualist’s arguments and the problems they pose. Burge has offered arguments of two very different kinds for his anti-individualistic model of the mental, and the differences between the two are important in assessing the indexical strategy. The argument in “Other Bodies” is a direct descendant of the argument advanced by Hilary Putnam (1975) in “The Meaning of ‘Meaning’ ”; in this argument the twins are ensconced in different physical environments—the Earthling inhabits a waterworld, while his twin inhabits a twater-world—and it is this difference in the physical environments that justifies us in saying that one of them believes that water is clear, while the other believes that twater is clear. The argument of “Individualism and the Mental” is quite different; it does not turn on there being some relevant difference in the twin’s physical environments, in the kinds of stuff out there, but on the following assumptions: (i) the twins are members of communities with different linguistic conventions—in the one arthritis is by definition a disease of the joints, while in the other ‘arthritis’ is used to denote a wider range of diseases, including some that can occur outside the joints; and (ii) one of the twins, the Earthling, can be truly characterized as believing that arthritis is such and such, even though she has a faulty understanding of the concept arthritis; she thinks
78
Joseph Owens
that arthritis can occur in the thigh. The other twin, of course, lacks this belief; she believes, correctly, that tharthritis can occur in the thigh. Both examples appear to show that psychophysical supervenience is false, and that physically identical individuals can differ in their propositional attitudes. This anti-individualistic conception of the mental gives rise to a number of metaphysical and epistemological worries. On the metaphysical front, it poses a direct challenge to contemporary materialistic approaches to the metaphysics of the mental: It tells against all reductionistic theories of the mental, and it even seems to tell against identifying token mental and physical events, thereby undercutting all contemporary materialistic approaches to the mental. In short, the Burgean arguments seem to show that current materialistic understanding is at odds with our intuitive conception of the mental.1 Then there are a couple of epistemological concerns. On the one hand, a number of theorists have argued that anti-individualism undermines self-knowledge, the intuitive idea that one has direct and authoritative knowledge of one’s own occurent thoughts. The argument is a simple one: On the anti-individualistic model, the contents of a subject’s thoughts vary as we vary her physical and social environment, and this variation in content is not introspectively detectable. Hence, one must know that one’s environment has these features in order to know the contents of one’s own thoughts; to know that one is thinking the thought that water is clear (or that arthritis is painful) as opposed to thinking the thought that twater is clear (or that tharthritis is painful). [The subject] has to be able to exclude the possibility that his thought involved the concept of arthritis rather than the concept tharthritis, before he can be said to know what his thought is. But this means that he has to reason his way to a conclusion about his thought; and reason to it moreover from evidence about his external environment. . . . (Boghossian 1989)2
On the other hand, Hilary Putnam (1981) has argued that not only is anti-individualism compatible with self-knowledge, but the two ideas can be combined to fashion a new antiskeptical argument. Here, too, the strategy is the obvious one: (i) We have introspective knowledge of the contents of our own thoughts—I know, for example, that I am now thinking that water is clear (part of our intuitive conception); (ii) we now know (from the anti-individualistic arguments) that our thoughts can have the contents they have only if the world we inhabit is of a certain kind (e.g., a water-world); hence (iii) our introspective knowledge of our thoughts affords us knowledge of the external world. Putnam, for example, has used this strategy to argue that I can know that I am not a brain in a vat: I know that I am now entertaining the thought that I am not a brain in a vat, and I know I could not do this if I really were a brain in a vat (I would lack the right linkage to vats, etc.); hence, contrary to the Cartesian sceptic, I do know that I am not a brain in a vat.
Anti-Individualism, Indexicality, and Character
79
The status of these epistemological problems is very different from that of the metaphysical problems. The anti-individualistic metaphysics is at odds only with prevailing philosophical theory, but these alleged epistemological consequences run counter to deeply entrenched, commonplace intuitions, and thus they seem to tell against the antiindividualistic model itself. Descartes’s vision of introspection may have been inflated, but few if any want to make self-knowledge, the knowledge one has of one’s occurrent attitudes, depend on one’s knowing that the external world is of a certain sort. Putnam’s contrary view, that knowledge of the contents of our own minds enables us to infer that the world is of a certain kind—that I am in a water-world or that I am not a brain in a vat—is equally problematic. The anti-individualistic model would be in trouble if it entailed that I could determine I was in a water-world without checking the world; this kind of information is not to be gleaned from a mere examination of the contents of my mind. More generally, classical Cartesian sceptical concerns seem to be rooted in our conception of knowledge; they don’t seem to be the product of some faulty theory of mind. Anti-individualism is in trouble if it either undermines self-knowledge or provides for an implausible route from the mind to the external world. Proponents of the indexical strategy argue that these metaphysical and epistemological concerns can be minimized or eliminated by a simple recognition of the fact that each of the problematic thought contents contains an indexical element. The idea that the crucial notions might be indexical was first broached by Putnam in the very paper in which he introduced us to Twin Earth. In this paper, Putnam was not arguing for anti-individualism, for the claim that the twins differ in belief or other psychological states (Putnam 1975, p. 220). Indeed he made an explicit assumption of methodological solipsism, “the assumption that no psychological state, properly so called, presupposes the existence of any individual other than the subject to whom the state is ascribed,” and this assumption ensures that the identical twins don’t differ in their “real psychological states.” Putnam’s central concern in this paper was with meaning, not mental content; his concern was to argue that classical semantic theory—the theory of Frege, Carnap, Church—was committed to two inconsistent assumptions: (i) Knowing the meaning of a term is just a matter of being in a certain psychological state; and (ii) the meaning of a term—its intension—determines its extension. Given that (a) the twins both know the meaning of their respective ‘water’ terms, (b) these terms have different extensions, and (c) the twins don’t differ in psychological state (given the assumption of methodological solipsism), we have to abandon (i) or (ii). When the thought experiment is viewed in this light, as putting pressure on (i) or (ii), rather than casting doubt on the individualistic conception of the mental, it may seem to suggest some kind of indexical reading of the twins’ assertions. For it is characteristic of indexicals that intension or meaning does not serve to determine extension; different
80
Joseph Owens
utterances of the first-person pronoun ‘I’ plausibly agree in meaning or intension, but differ in extension. And this is just the strategy Putnam opted for. Our theory can be summarized as saying that words like ‘water’ have an unnoticed indexical component: ‘water’ is stuff that bears a certain similarity relation to the water around here. Water at another time or in another place or even in another possible world has to bear the relation same to our ‘water’ in order to be water. Thus the theory that (1) words have ‘intensions’, which are something like concepts associated with the words by speakers; and that (2) intension determines extension—cannot be true of natural-kind words like ‘water’ for the same reason the theory cannot be true of obviously indexical words like ‘I’. (Putnam 1975, p. 234)
Burge, of course, offered a very different construal of the Twin Earth thought experiment—seeing it as telling against (c), against methodological solipsism or individualism, rather than as against (i) or (ii), and, in arguing for this interpretation, he rejected Putnam’s argument for the indexical interpretation of the natural kind terms. He concluded: “The fact that Twin-Earthians apply ‘water’ to XYZ is not a reflection of a shift in extension of an indexical expression with a fixed linguistic (English) meaning, but of a shift in meaning between one language, and linguistic community, and another. Any expression, indexical or not, can undergo such ‘shifts’ as a mere consequence of the conventionality of language. . . . If Indians applied ‘bachelor’ to all and only male hogs, it would not follow that ‘bachelor’ as it is used in English is indexical” (Burge 1982, p. 105). Burge’s response is surely correct, indeed obviously so—the considerations adduced by Putnam lend no support to the indexical interpretation—and yet the indexical reading persists to this day. The persistence of this interpretation has, I suggest, little or nothing to do with the kinds of considerations Putnam offered, or indeed with any untutored intuitions about indexicals or natural-kind terms. This approach remains tempting because of one particular theory of indexicals, that of David Kaplan. In a series of papers, Kaplan developed an intuitive and highly original theory of demonstratives and other indexicals, and elements in that theory seemed to provide just what was needed to accommodate our intuitions that there is a sense in which the twins think and say different things, while at the same time minimizing the metaphysical and epistemological impact of the examples.3 My primary concern in this essay is with the question: Can this theory of indexicals serve such a deflationary role, can it seriously alter our understanding of the Twin Earth examples? For our purposes, it is useful to break Kaplan’s account into three components: (1) his account of the logic and semantics of expressions that are explicitly indexical in character; (2) his claim that important elements in this account generalize and apply to expressions, for example, kind terms, which are not explicitly indexical in character; and (3) his metaphysical and epistemological commentary on the semantics and the distinctions drawn therein. In each case I will be very brief, confining my attention to those elements of the account that bear on the issues of special concern here.
Anti-Individualism, Indexicality, and Character
81
1. The distinguishing feature of indexical expressions, of course, is the fact that their extensions are determined by contextual factors, and traditional treatments of indexicals focused on accommodating this feature. Kaplan, however, argues that this is only half the story: Certainly, extension is a function of contextual factors, but, he argues, so too is the intension, the proposition expressed by an indexical sentence. You and I express different contents (propositions) by our utterances of the one sentence ‘I am ill’. Nevertheless there is a clear sense in which different utterances of ‘I am ill’ agree in linguistic meaning, and indeed it is this common meaning that determines what specific content or intension is expressed in a given context of utterance. We now have three semantic factors to play with. We have two factors that vary from context to context: (i) the referent or extension of an expression; (ii) the variable intensional element, which he dubbed content; and then we have (iii) that other intensional entity, the constant linguistic meaning, which Kaplan dubbed character. The following gives us a good sense of the interplay between context, character, and content. Let us call this second kind of meaning, character. The character of an expression is set by linguistic conventions and, in turn, determines the content of the expression in every context. Because character is what is set by linguistic conventions, it is natural to think of it as meaning in the sense of what is known by the competent language user. Just as it was convenient to represent contents by functions from possible circumstances to extensions (Carnap’s intensions), so it is convenient to represent characters by functions from possible contexts to contents. . . . This gives us the following picture: Character: Contexts fi Contents Contents: Circumstances [of evaluation] fi Extensions or, in more familiar language, Meaning + Context fi Intension Intension + Possible world fi Extension (Kaplan 1977, pp. 505–506)
For our purposes, the central point is this: The theorist cannot rest with the traditional intensional/extensional distinction if she is to do justice to the logic of indexical sentences. She has to distinguish between two kinds of content: a kind of content that varies from one context of utterance to another (content, akin to a proposition), and a kind of content (character), akin to linguistic meaning, which is determined by the conventions of the language and remains constant across these different contexts of utterance. This intuitive distinction enables Kaplan to do justice to elements in the logic of indexicals that had been neglected in earlier treatments. In particular, it enables him to account for the failure of the familiar principle of necessitation—if |=Ø then |=Ø, where Ø is
82
Joseph Owens
indexical: the sentence ‘I am here now’ is a priori and logically true in that it cannot be uttered falsely, but, unfortunately, the proposition expressed is anything but necessary.4 Kaplan accounts for this feature of indexical sentences by drawing a sharp line between logical truth and necessity, arguing that logical truth is properly viewed as a property of character, whereas necessity and contingency are properties of the proposition expressed by the sentence in a given context. 2. Although this theory was developed to account for the peculiarities of indexical expressions, there are, Kaplan argues, good reasons for not restricting this apparatus to indexical expressions. True, in the case of nonindexicals, there is not the same obvious distinction between character and content. Unlike indexical sentences, the character of a nonindexical sentence is ‘stable’—generating the same content in every context of use. However, the fact that the character is stable should not blind us to the fact that the distinction can still be drawn. He argues that it can and should be drawn, even when the sentence in question is not explicitly indexical. Here too, he argues, we must distinguish between those sentences that are logically true (can’t be uttered falsely) and those that express necessary truths. On the one hand, we have sentences such as ‘$x Exists x’, sentences whose meaning is such that they can’t be asserted falsely, but don’t express necessary propositions (assuming that some possible worlds are empty). On the other hand, there are the familiar examples of a posteriori necessary truths, sentences that serve to express necessary propositions, but the necessity seems little to do with the meaning of the sentence and everything to do with what kinds of world are possible; here the nature of our world, not the meaning of our expressions, determines what is and what is not necessary.5 Kaplan holds out the hope that we can deepen our understanding of such phenomena if we insist on the general distinction between content and character, linking necessity to the first and validity to the second. In his recent remarks on this issue, he writes: I find it useful to think of validity and necessity as never applying to the same entity. Keeping in mind that an actual world is simply a circumstance of a context of use, consider the distinction between: (V) No matter what the context were, R would express a truth in the circumstances of that context, and: (N) The content that R expresses in a given context would be true no matter what the circumstances were. The former states a property of sentences (or perhaps characters), validity; the latter states a property of a content of a sentence (a proposition), necessity. (Kaplan 1989, p. 596)
Now we can see the beginnings of an answer. In the case of indexicals, we had to distinguish between two kinds of contents, one that varies from one context of utterance to another, and one that is common across these different contexts of utterance. In addition,
Anti-Individualism, Indexicality, and Character
83
we are now advised that there are good reasons for extending this distinction to the language at large, to distinguish generally between two kinds of content, one constant and one variable across contexts of utterance. This seems to open the way to an intuitive understanding of the Twin Earth examples: The twins utter the same sentence, and this sentence serves to express different contents or propositions. But, we should not think of this difference as a difference in psychological state: The twins agree in character—the propositional differences have to do with the external world—and this agreement in character is enough for sameness in psychological state. 3. These expectations, moreover, are supported by Kaplan’s explicit comments on the metaphysics and epistemology of belief. He writes: What we must do is disentangle two epistemological notions: the object of thought (what Frege called “Thoughts”) and the cognitive significance of an object of thought. . . . a character may be likened to a manner of presentation of a content. This suggests that we identify objects of thoughts with contents and the cognitive significance of such objects with characters. . . . To break the link between cognitive significance and universal Fregean senses and at the same time forge the link between cognitive significance and character we must come to see the context sensitivity (dare I call it ego-orientation?) of cognitive states. (Kaplan 1977, pp. 530–531)
The suggestion is that we think of psychological states as being cognitively equivalent if they agree in character, even if they involve different propositions, and this seems to provide the individualist with what he needs. Indeed Kaplan comes close to explicitly committing himself to a kind of individualism in the following remarks on a Putnam-type example: We raise two identical twins, Castor and Pollux, under qualitatively identical conditions, qualitatively identical stimuli, etc. . . . Have we not been successful in achieving the same cognitive (i.e. psychological) state? Of course we have, what more could one ask! But wait, they believe different things. Each sincerely says, My Brother was born before I was And the beliefs they thereby express conflict. . . . This does not reflect on the identity of their cognitive states. (Ibid., p. 531, my italics)
The lesson to be drawn is that although the twins express different propositions by their respective utterances, this difference does not provide for a difference in psychological or cognitive content. In addition, character seems to provide for a solution to the epistemic problems. Recall that there were two worries: Some have argued that externalism is counterintuitive in that it seems to undercut the possibility of direct and authoritative knowledge of the contents of my own occurrent beliefs (since their content is not determined in part by context, and
84
Joseph Owens
this difference in content does not make for any “introspective difference”). Others have worried that externalism is problematic because it appears to provide for a suspect route from knowledge of my mind to knowledge of my world: Given externalism and selfknowledge, we can simply infer that the world is of a certain kind, for example, a water-world. Kaplan’s account seems to provide just what is needed to alleviate both concerns: Character is one thing, content is another, and we have different kinds of epistemic access to each. Ordinary linguistic competence is enough for knowing the character of what we say and think, but much more may be needed to know the content of what we say and think (the proposition we express): “Although lack of knowledge about the context . . . may cause one to mistake the content of a given utterance, the character of each wellformed expression is determined by the rules of the language . . . which are presumably known to all the competent speakers” (Kaplan 1977, p. 548). Since the external differences make for a difference in proposition, not in character, we don’t kneed to know that the world is such and such to know the character of what it is we are thinking, and knowledge of character, the constant element that corresponds to meaning, is all that is necessary for self-knowledge. Moreover, since knowledge of character does not directly yield knowledge of the proposition entertained by means of that character, I cannot introspectively determine whether the proposition I entertain when I think ‘water is clear’ is a water-proposition or a twater-proposition, and so I can’t infer that the world is water-world of a twater-world; skepticism remains a real possibility. Other philosophers have welcomed Kaplan’s suggestions, arguing that an adequate understanding of the Twin Earth examples is possible only within a framework such as that advanced by Kaplan, and furthermore, when they are construed in this way, they pose no special metaphysical or epistemological problems. Two examples will suffice. In an early discussion of the Twin Earth examples, Joseph Almog has argued that these examples cry out for a Kaplan-type semantics; he endorses Putnam’s proposal that we treat kind terms as indexical, but he complains that Putnam’s discussion is conceptually “flawed” because it is developed within an “inadequate logical framework”—the dichotomous framework derived from Frege-Church (Almog 1981). To do full justice to the Twin Earth examples we have to employ the more complex trichotomous framework developed by Kaplan. [Putnam’s] informal ideas and basic conceptions are ones with which I agree, with some minor differences. YET, it seems to me that Putnam’s work suffers from a lack of proper logical and conceptual frameworks. The lack of a conceptual framework is the result of the lack of an adequate logical framework. I will argue that in using the main ingredients of classical Churchian intensional logic and consequently the conceptual framework of Frege-Carnap-Church, Putnam’s insightful intuitions may be jeopardized. (Almog 1951, p. 347)
Anti-Individualism, Indexicality, and Character
85
So let me sketch how the Putnamian picture fits Kaplan’s logic of Demonstratives. At the conceptual level the following should be said: just as with other indexicals (say ‘I’), there is no point in talking about the reference of ‘water’ without a specification of a context of use. So we have a threelevel process: First there is the character of ‘water’ which is a rule which tells one that in a given context of use, one dubs something as ‘water’ if that object has the phenomenal appearance specified as follows: liquid, transparent, fills lakes and rivers, etc. Then, given a specific context, the character determines a specific Content (intension). This is a function which, given a possible world, assigns an appropriate body of liquid. (Ibid., pp. 352–353)
While Almog’s main claim is that we can do justice to these Twin Earth examples only if we employ something like the apparatus proposed by Kaplan, others have focused on specific ways in which Kaplan’s analysis helps to alleviate the metaphysical and epistemological problems posed by the examples. Graeme Forbes (1995), for example, has argued that a Kaplan-style interpretation of these examples exposes the flaws in Putnam’s problematic antiskeptical argument. Putnam’s argument, remember, was a simple one: (i) I know that I am now entertaining the thought that I am not a brain in a vat. (ii) I know that I could not entertain this thought if I were a brain in a vat (the antiindividualistic premise). Therefore, (iii) I know that I am not a brain in a vat. One might be tempted to counter this line of argument by rejecting the first premise, insisting that all I really know, when I genuinely open myself to skeptical doubt, is that I am entertaining the thought I express by the sentence ‘I am a brain in a vat’, whatever thought that is. Forbes rightly rejects this easy response. There is a clear and recognizable difference between simply assenting to the sentence ‘I am a brain in a vat’ and using this sentence as a sentence one understands to frame a specific thought, the thought that I am a brain in a vat. And when I use this sentence in the manner envisaged by Putnam, I use it as a sentence I understand; I use it to frame the thought that I am not a brain in a vat. Forbes thus grants Putnam the intuition that we know what we are thinking when we think the thought ‘I am not a brain in a vat’, and he sees the real challenge, then, as one of providing an account of self-knowledge, such that (i) I may be truly said to know that I am now entertaining the thought that I am not a brain in a vat (or that water is clear), and (ii) this knowledge of my thought content should not (contra Putnam) enable me to infer that I am not a brain in a vat (using the anti-individualistic intuition that a brain in a vat could not entertain this thought). He answers this challenge by developing a picture of self-knowledge within the Kaplan framework and arguing that it satisfies conditions (i) and (ii):
86
Joseph Owens
For Putnam . . . , there are two languages in play, the language of the normally embodied agent and the language of the envatted brain: the word sequence ‘I am not a brain in a vat’ is meaningful in both languages, but means something different in each. From the viewpoint of Kaplan’s logic, on the other hand, there is only a single language, in which words like ‘brain’ and ‘vat’ are indexicals: they have a character that generates one type of context for normally embodied contexts and another for envatted contexts. This contrast between two languages on the one hand, versus indexicals in a single language, on the other, is in principle inconsequential, but the ‘one language with indexicals’ version has the advantage of suggesting, via comparison with ordinary indexicals, a mechanism that allows us to get beyond the mere metalinguistic conclusion. The issue is how the agent of the context can understand his words if he does not have discriminating knowledge of their referents, that is, if he cannot tell if he is thinking about brains, rather than brains-in-the-image [the referent of the envatted term ‘brain’], and so on. . . . My proposal is that the agent does have discriminating knowledge, in the only sense that is appropriate for understanding the indexicals ‘brain’ and ‘vat’. Consider an analogous case involving the ordinary indexical ‘here’. Suppose that E is an explorer who is lost in a featureless sand desert. After wandering aimlessly for a while, E notices a viper at her feet and thinks to herself there is danger here. It is clear that E understands her own words. . . . Similarly, E knows such propositions as she expresses with ‘By “here” I am referring to here’. How is this possible if she is lost? (Forbes 1995, pp. 212–213)
The solution is strikingly simple: The one lost in the desert certainly knows the thought she entertains when she entertains the thought that there is danger there, but this does not mean that she is in a position to distinguish between entertaining this thought and the very different thought (proposition) she would entertain were she to entertain the thought ‘there is danger there’ at a different (but qualitatively identical) location. So too with the subject who entertains the thought that she is not a brain in a vat: she knows what she is thinking, but only under the character ‘I am not a brain in a vat’. The normal and the envatted subject both employ this same character to express different propositions; we would say, in English, that the one entertained the thought that she was not a brain in a vat while the other entertained some ‘vat-in-the-image’ thought. Just like the individual in the desert, this subject knows what she is thinking, but she is not in a position to say which of these two thoughts (propositions) she is entertaining; the knowledge she has of her thought does not enable her to say whether it is a ‘real vat’ thought or a mere ‘vat-in-theimage’ thought. This is an appealing strategy: Kaplan has given us independent reasons for thinking that the character/content distinction should not be confined to explicit indexical expressions, and this seems to provide for a very natural and different understanding of the Twin Earth examples. It seemingly enables us to grant Burge his central claim that the twins entertain different propositions (and so warrant different belief characterizations), while denying that this difference in propositions has any interesting or controversial metaphysical or epistemological consequences. This initial optimism, however, is misplaced.
Anti-Individualism, Indexicality, and Character
87
In the next sections, I argue that Kaplan’s theoretical framework simply cannot be used to defang the Twin Earth thought experiments. II In arguing that a Kaplan-style semantics cannot be used to undermine the force of the twin earth thought experiments, I will grant the central elements of his account. In particular, I will assume that the character/content distinction is general and applies to sentences such as: (S)
Water is clear.
Of course, this in itself is of little comfort to the individualist unless she also assumes that the character of a sentence such as (S) is not stable, and so far as I can tell there is no good reason to think that this is true; certainly it does not seem to be needed to accommodate the a priori/necessity distinction (and Kaplan clearly does not endorse it). However, I don’t wish to argue this specific issue either, since there are more fundamental problems with this strategy. First, it is important to recognize that the question as to whether the character of (S) is stable or not seems to be simply irrelevant to the problems posed by the twins. To suppose that (S) has a nonstable character is to suppose that this sentence serves to express different propositions in relevantly different contexts. It is to suppose that the rules of the language, English, are such that this English sentence may be used to express one proposition in one context and this same English sentence may be used to express a different proposition in a different context. We are, by assumption, looking at one expression of one specific language, English, an expression whose character is determined by the rules of this language, and so we cannot treat the linguistic community of the speaker—the community of rules—as just one of the factors to be varied at will. In arguing that the character of (S) is not stable, we do not commit ourselves to what if anything is expressed by sentences in languages other than English, even if they happen to sound just like English sentence (S). Kaplan envisions an adequate semantics as assigning character as well as content and extension to linguistic entities, but these entities are not individuated simply in terms of their being composed of this or that string of characters, or in terms of their satisfying this or that phonetic description. Character is assigned to instances of sentences, individuated as instances of sentences in a specific language. Of course, in formulating the semantic rules of some language E, one assigns characters to syntactically characterized entities, but in doing so one assumes that this syntactic categorization is only over entities drawn from E. The fact that expressions in other languages may merit the same syntactic, nonsemantic characterizations is simply irrelevant; the rules have nothing to say about such entities.6
88
Joseph Owens
Unfortunately, many presentations of the twin examples tend to be fashioned in such a way as to obscure this point. In presenting these examples it is convenient to present them as examples of two individuals, Alf and Alf*, uttering the same sentence ‘water is clear’; they utter the same sentence, but because of differences in their environment they express different propositions by their respective utterances. This is one way to tell the story, and when it is told like this it is easy to conclude that this is just one more case in which the same sentence is used to assert different propositions when uttered in relevantly different contexts. This is how both Almog and Forbes read the examples. And when they are read in this way, it is natural to conclude that the Twin Earth examples serve merely to reveal the pervasive context-sensitivity of language, to conclude that they serve to illustrate the implicit indexicality of sentences such as ‘water is clear’. These examples are examples in which the term ‘water’ is used to designate H2O in this context and to designate something very different, XYZ, in this different context. The examples are interesting insofar as they illustrate the implicitly indexical character of these kind terms, but they pose no special epistemological or metaphysical problems so long as we follow Kaplan’s suggestion and think of the cognitive element as corresponding to character rather than content. But this is an outright misreading of the examples. These are examples in which individuals speak different languages that happen to coincide in interesting ways: The inhabitants of Twin Earth use a term that sounds like our term ‘water’, and they use it when confronted with stuff that has surface similarities to water—but that’s as far as it goes. This is the important point that is obscured in speaking of them as uttering ‘the same sentence’. Once again, the point here is the obvious one that there is nothing in Kaplan’s theory to support the idea that same sounding sentences should agree in character, when these sentences are drawn from different linguistic communities, with different rules, even if these same sounding sentences are used in the same way (nonintentionally characterized).7 There is simply no reason to think that this theory and the character/content distinction throw any light on the problems posed by the twins. III In this final section I argue for the stronger claim, that appeals to character can’t help the individualist’s program for the simple reason that character itself does not supervene on nonintentionally individuated states. I look first, in (a), at sentences and thought that are explicitly indexical in form, and then, in (b), I look at sentences (and thoughts) expressed with the aid of kind terms, which, for the sake of argument, I will assume are implicitly indexical in character. In each case, I argue, we should think of the corresponding character as failing to supervene.
Anti-Individualism, Indexicality, and Character
(a)
89
Explicit Indexical Expressions
Consider the simple English sentence: (F)
Ralph has had nothing to eat today.
When uttered on a specific day Dn, F asserts that Ralph has had nothing to eat on day Dn nothing to eat from midnight on Dn-1 to the time of utterance. Twinglish, we assume, is very like English; sentences in this language sound like English sentences, they contain indexicals, and so on. There is, however, one difference of special interest. Like us, speakers of Twinglish think of a day as a period of twenty-four hours, but unlike us, they count the day as beginning not at 12 (‘midnight’) but only after that hour has expired. They mark the new day as beginning at ‘the end of the twelfth hour’, as beginning when the clock strikes 1. The character (rule) encoded by their term ‘today’ is different form that encoded by our term ‘today’. If we take a speaker from each community, give them a common temporal location (the only relevant index), and have each of them assert the string ‘Ralph has had nothing to eat today’, they will assert different propositions. This is even more evident when fashioned in the counterfactual idiom preferred by Burge: The proposition Alf would express on Dn if he were speaking Twinglish is different from the one he does express (in English). Since these same-sounding sentences serve to express different contents or propositions when uttered in the (relevantly) same content, they differ in character (represented as a function from contexts of utterance to contents/propositions). An adequate semantics, then, will assign different characters to the English and Twinglish versions of ‘Ralph has had nothing to eat today’. Alf, let us assume, is a generally competent speaker of English, even though he appears to have some unusual views. He knows that days are periods of twenty-four hours and he knows that the new day begins at the end of the (second) twelfth hour. He insists, ‘The new day begins at the end of the twelfth hour, when the clock strikes 1’. Alf, it seems, believes that the day begins at 1. This is a slightly peculiar belief, but that is all; there is nothing impossible in the supposition that one might have such a belief; it is simply a mistaken belief about when the day begins. This same Alf now asserts, ‘Ralph has had nothing to eat today’. Surely, in this case, just as in the original Twin Earth examples, we have every reason to say that Alf asserted that Ralph has had nothing to eat today; indeed we have every reason to say that he thinks that Ralph has had nothing to eat today. (I am assuming here that this indirect reporting of what Alf says and thinks is itself carried out on Dn.) We recognize that, given Alf’s mistaken beliefs about when the day begins, it might be misleading to simply say that Alf believes that Ralph has had nothing to eat today; it might encourage the hearer to incorrectly attribute a variety of other beliefs to Alf. These pragmatic considerations are
90
Joseph Owens
important, but they don’t seem to tell against the accuracy of the simple homophonic interpretation. A variety of philosophical considerations might, of course, tempt one to reject this intuitive, homophonic reading of Alf, to argue that we must incorporate his mistaken belief about when the day begins into our reporting of what he says and thinks; to argue that we cannot literally interpret him as making the same claim we make when we say ‘Ralph has had nothing to eat today’. There are numerous variations on this general ‘reinterpretation’ strategy, but they are uniformly implausible. Burge discussed these strategies at length in his original treatment of the arthritis example, arguing persuasively that they are without merit; they are the product of misplaced philosophical intuitions and they distort our practice.8 And these same arguments apply, with little alteration, to the suggestion that we should reinterpret Alf’s claim, ‘Ralph has had nothing to eat today’. The arguments are now familiar and I shall not recount them here. I conclude then, that (a) it is possible for Alf to have the mistaken beliefs he appears to have, and (b) this does not undermine the homophonic interpretation of Alf; we should interpret him as making the claim we would were we in his situation. What then of the character of Alf’s token ‘Ralph has had nothing to eat today’? I see no reason whatsoever for supposing that it differs in character from our tokening of the same sentence. We use the same indexical sentence, and, as I have just argued, we use it to express the same proposition in the same contexts of utterance; the linguistic rule that serves to determine the various propositions Alf expresses by his utterances of ‘Ralph has had nothing to eat today’ also serves to determine the propositions we express by our utterances of that same sentence. Alf’s utterances of ‘Ralph has had nothing to eat today’ has the same character as our corresponding utterances; his utterances and ours share the same character, the character assigned the English sentence ‘Ralph has had nothing to eat today’.9 But to grant this much opens the way to a new series of Twin Earth examples, examples in which the physically identical twins differ in character. The Twinglish speaker Alf* asserts the same-sounding sentence ‘Ralph has had nothing to eat today’, and, as we have seen, this sentence has a different character—the rule governing ‘today’ in Twinglish is not the same as the rule governing ‘today’ in English. Alf and Alf* utter same-sounding sentences, but these corresponding sentences differ in character, and so serve to express thoughts with different characters (given that we are supposing thoughts to have character—indeed to be individuated by character). This is so even if Alf* is a physical replica of Alf. Hence, these twins differ not only in the propositions asserted and entertained; they utter sentences and entertain thoughts with different characters. Character it seems is of no help to the individualist’s program. (b)
Implicit Indexicals
This lesson is even more evident when we look to natural-kind terms and other expressions that are not explicitly indexical. The individualist insists that here too we must dis-
Anti-Individualism, Indexicality, and Character
91
tinguish between content and character, that even though the corresponding thoughts and utterances of the twins differ in content they don’t differ in character, and this sameness in character is all we need for sameness of psychological state. This strategy is mistaken, for contrary to the assumptions of the individualist, the character of natural-kind terms, to the extent that we think of them as having character, is itself nonindividualistic. This point can be clearly seen if we consider an example that combines elements from both the water and the arthritis examples: 1. Assume that Alf has acquired the terms ‘fish’ and ‘whale’ in a fairly routine fashion: He has, we may suppose, seen photographs of both; he has heard the usual things about the gigantic size of whales, and so on. 2. He has one deviant belief: He sincerely thinks to himself: ‘These whales sure are much larger than any other fish, they must be the biggest fish in the world’. Alf apparently believes that whales are fish. 3. Alf* is physically just like Alf: He hears the same things (nonintentionally characterized); he is disposed to produce the same sounds, and so on. He is, we assume, a physical replica of Alf, down to the last molecule. However, in Alf*’s world, they speak Twinglish, not English, and the things they call ‘whales’, though many of them look like our whales, are not whales, but a kind of fish. In this case, as in the last, there really are only two real issues of concern: (1) Should an adequate semantics assign the same character to the term ‘whale’ as it occurs in English and Twinglish? (Call this the cross-language question.) (2) In assigning a character to Alf’s token sentence ‘whales are fish’, should we assign it the character assigned that sentence in English? Should the token Alf produces be assigned the same character as tokens of this sentence produced by speakers who have no specially deviant views of whales—speakers who recognize that whales are mammals? It seems to me that the answer to the first question is a clear no, and the answer to the second is an equally clear yes. Consider the first question, the cross-language question. We have, by stipulation, two linguistic communities, using a same-sounding term ‘whale’, but they apply it to very different creatures: We, members of the English community, apply it to a kind of mammal, and they apply it to a kind of fish, some of which happen to look like our whales. These different communities, we may assume, have very different theories of these creatures, different mythologies, and so on; indeed, so long as we assume that Alf and Alf* are illiterates who have little knowledge of the written word, we may even suppose that the two communities spell it differently (‘whale’, ‘whail’). I can’t think of any good reason to think that these terms, drawn from these different communities,
92
Joseph Owens
should receive the same character. Clearly the fact that they sound the same (or at least utterances of them in the neighborhood of Alf and Alf* sound the same), is no reason, and neither is the fact that some of the creatures bear a surface similarity to each other. We don’t have to suppose that the creatures they refer to as ‘whales’ on Twin Earth are all similar to whales. Both individuals are acquainted only with a sampling of the kind, and we can assume that there are clear and striking differences outside this common sample; many whales don’t look anything like some of the things they label ‘whale’ (or ‘whail’) on Twin Earth (I shall assume they use the term ‘whail’ from here on out). If we think of the character of ‘whale’ as being at all like the character of an explicit indexical, such as ‘I’ (and what other model do we have?), as a meaning-rule that dictates how the term is to be employed, then this rule must somehow include the stipulation that ‘whale’ properly applies only to whales, to mammals of a certain sort (certainly, it won’t do to say that the word is used in such a way that it properly applies to anything of a certain appearance whether or not it is a mammal). And this stipulation has no place in the rule for the Twinglish term ‘whail’. The lesson is clear: Though the terms sound the same, they are governed by different rules in the two communities; they differ in character. Once again, the only avenue open to the individualist is to respond negatively to the second question, the internal English question; she must insist that Alf’s token ‘Whales are fish’ differs in character from tokens of that sentence uttered by speakers who don’t share Alf’s mistaken beliefs about whales. But this path too goes nowhere. In interpreting Alf’s whale utterances, we intuitively interpret them homophonically, we interpret him as saying that there is a whale there, that whales are large; indeed, we ordinarily characterize his mistake as one of thinking and claiming that whales are large fish. But this is to say that we interpret him as using the word ‘whale’ in a variety of contexts to assert the proposition that would be asserted by a nondeviant speaker in that same context; we get the same function from context to proposition, the same character. So far as I can tell, no one has given us the slightest reason to reject this intuition: Alf really does entertain the thought (as we do, if only to deny it) that whales are fish, and he expresses this thought with the sentence ‘Whales are fish’. He can use this sentence to express this thought because it has the character it has, namely, the very character it has as a sentence of English; the proposition he gets to express when he uses this sentence, intending to abide by the rules, is the proposition dictated by the rules of English—the thought or proposition that whales are fish. Defenders of individualism have not provided any serious argument for a positive response to the cross-language question or for a negative response to the internal English question. In appealing to character, they have simply ignored issues relating to Alf and Alf*’s participation in their respective linguistic communities; they look at the twins in isolation, see them as using their corresponding terms according to the same ‘rule’, and
Anti-Individualism, Indexicality, and Character
93
then they take this ‘rule’ as the shared character. Almog, for example, writes: “First there is the character of ‘water’ which is a rule that tells one that in a given context of use, one dubs something as ‘water’ iff that object has the phenomenal appearance specified as follows: liquid, transparent, fills lakes and rivers, etc.” (Almog 1981). Certainly, one can think of the twins as using their respective ‘water’ terms in accordance with some such rule, but this provides no support for either the cross-language thesis (same character) or the internal thesis (different character)—it provides no support for the claim that these terms agree in character. The fact that twins both rely on such a rough and ready rule in their use of their terms would support the cross-language thesis only if, among other things, the rule corresponded to the character of the English term ‘water’. There is simply no reason to think that this is true. This is not the rule that governs our use of the term ‘water’—the rule that dictates how an English speaker should use it— unless we read it as entailing that something is truly designated (‘dubbed’) ‘water’ only if it is the same as the stuff in these lakes, rivers etc.—that is, H2O. Of course, we do use the term ‘water’ to designate the stuff in the lakes etc., and our practice rests heavily on the fact that we can readily distinguish water by its familiar phenomenal features. But to suppose this much is not to suppose that we use the term in such a way that it is said truly to designate something if it merely looks like water (H2O), if it flows in counterpart rivers, etc. No, the stuff must be water, it must be the same kind of stuff as flows in the river Liffey, etc. But when we read it in this strong fashion, as requiring more than mere phenomenal similarity, we are forced to distinguish between the character ‘water’ has in English, and the character it has in Twinglish. The inadequacy of this line of thought is even more evident in the whale example. In this example we stipulated not only that the corresponding terms designate different kinds, but also that members of the kinds (not encountered by the twins) differ markedly in their phenomenal features. True, the twins rely on similar phenomenal features in their actual use of the corresponding terms, but others are more familiar with full range, and we know that we cannot fashion rules for the proper use of ‘whale’ and ‘whail’ in terms of this restricted set of phenomenal features; the rules that govern these two terms are different, and they cannot be specified in terms of the impoverished set of phenomenal properties that twins rely on. We cannot argue from the fact that there is a certain limited similarity between Alf’s use of ‘whale’ and Alf*’s use of ‘whail’ to the cross-language thesis that the English word ‘whale’ agrees in character with the Twinglish word ‘whail’. Likewise, such similarity in use between Alf and Alf* clearly provides no argument for the internal English claim that we must distinguish between the character of Alf’s token ‘whale’ and the character that term is assigned as a term of English. The individualist offers no arguments for either of the essential claims (the cross-language and internal theses); she simply assumes that the kind of similarity in use between Alf and Alf* is sufficient to
94
Joseph Owens
ensure sameness of character; she assumes that the character of a speaker’s expression is determined by her own idiosyncratic usage—by the description or criteria she relies on in identifying instances, ignoring the speaker’s embeddedness in a linguistic community. In doing so, she fails to recognize that the intuitions that gave rise to anti-individualistic model of the mental also support an anti-individualistic reading of character. Such uncritical appeals to character don’t constitute a serious response to the anti-individualistic model; they simply ignore the fundamental motivations of anti-individualism. A brief glance at Burge’s original arthritis example serves to highlight this point. In this example, the relevant difference between the twins cannot be construed as simply a difference in the things they are talking and thinking about; both suffer the same bodily infirmities. The twins are said to say and think different things, not because of some difference in their physical environment, but simply because their corresponding ‘arthritis’ terms differ in meaning; the terms are used, understood and defined differently in the two linguistic communities. It is transparently clear that this difference cannot be construed on the indexical/character model, as a difference that is the product of a change of context rather than change of meaning or character. If we are going to assign characters to terms such as ‘water’ and ‘arthritis’, then this arthritis example can be read as directly illustrating the nonindividualistic nature of character. This is an example in which the corresponding terms are stipulated to differ in meaning/character, and nothing else matters very much to the story. This much is clear: Anyone who thinks of character as something like linguistic meaning (as Kaplan did) cannot assume that character supervenes on the nonintentional, unless they can independently disarm the arthritis example—the example that seems directly to demonstrate the anti-individualistic nature of character. In assuming an individualistic understanding of character, this theorist not only fails to confront the intuitions of the anti-individualist, she opts for an account saddled with the kinds of problems that plague “descriptional” theories of names. On the descriptional model, when a subject S uses a name N, she does so with a certain description in mind, and this description serves to express what she means by N and it determines what she refers to by N. Kripke (1980) rightly rejected this individualistic theory of names: The associated description, which reflects S’s idiosyncratic usage, cannot in general be supposed to give the meaning or the reference of the name N. The individualists’ mistake is of a similar nature: They assume that the character expressed by a speaker’s use of a sentence containing a natural-kind term such as ‘water’ is determined in part by the description that the speaker happens to associate with ‘water’—the conditions she actually relies on in identifying water. As in the case of names, this individualistic picture of character gives us the wrong results; it distorts the identity conditions for propositions, leading us to identify propositions that should be kept distinct and to distinguish propositions that should be identified. The issues here are subtle and deserve extended treatment, but a
Anti-Individualism, Indexicality, and Character
95
sketch of one example gives a sense of the problem. From the individualists’ perspective, the way in which the subject actually uses the sentences is all-important, and so when we interpret S’s utterances, we should assign the same character to sentences N and N* if he uses them interchangeably (according to the ‘same rule’).10 So if S happens to use the sentences (N) ‘water is necessary for life’ and (N*) ‘the clear liquid stuff that is found in the river, lakes, etc., is necessary for life’ interchangeably, then we should assign them the same character (in his idiolect). But if they have the same character, then when he uses one, say N, in a given context to express a certain proposition P, he is using it to express the proposition he could have expressed using the other sentence, N* (in that same context). This is a mistake. When S uses the name ‘Gödel’ he is not thinking or talking about Schmidt (even if he fits the description S associates with ‘Gödel’) and, likewise, when S thinks ‘water is necessary for life’ S is thinking that water is necessary for life, not that the stuff that flows in rivers etc. is necessary for life. This is so even if S happens to think these sentences say pretty much the same thing, even if S tends to use ‘water’ and ‘the stuff in the river’ interchangeably. If, for example, it should turn out that, because of a strange cosmic twist, we could not survive without river-stuff, lake-stuff, etc., but could survive without water, then S’s thought that water is necessary for life is false, but his thought that the river-stuff is necessary for life is true. When we individualize character to correspond to individual usage we individualize propositions, and this we know we can’t do. I have said little or nothing about epistemological issues, focusing instead on the metaphysics of character. However, it is, I think, clear that this nonindividualized character is not going to help resolve the epistemological problems associated with antiindividualism. Once more, a sketch of an argument will have to suffice. Consider again Alf who has the somewhat odd belief about when one day ends and another begins—he thinks that the break comes at 1 a.m. He thinks to himself ‘Ralph has had nothing to eat today’, and in doing so he thinks that Ralph has had nothing to eat today. Alf is not in a position to introspectively distinguish between being in this situation—using this sentence to fashion this belief—and being in the situation of his Twinglish counterpart, who uses this word-form to express a very different thought. The corresponding sentences and thoughts of these twins differ in character (section III (a)), but Alf is not in a position to introspectively recognize this difference—he is not in a position to say that the thought he is entertaining has character E (the character it in fact has as a sentence of English) or character E* (the character the corresponding sentence has in Twinglish). Indeed, given the story we have told, he tends to think (falsely) that it has the character E*. Clearly Alf’s epistemic position is no better than it was in the original examples: Competence in the language does not bring with it some kind of automatic, introspective access to sameness and difference in character.11
96
Joseph Owens
It is true, of course, that some of Kaplan’s comments on his own theory fostered the individualistic reading. He distinguished between the object of thought (the proposition believed) and the cognitive significance of an object of thought, suggesting that we identify character with the latter; he conceived of the subject as having much more direct access to character than she has to (propositional) content; and he seemed to think that the qualitative similarity between the twins, Castor and Pollux, guaranteed sameness of cognitive state (character). However, we now have good reason to suppose that this picture is at odds with core elements in his theory—indeed with the very central ideal of character as a semantic notion akin to the linguistic meaning of an indexical such as T—the meaning that is common to the many utterances of this expression. The theory after all is a logical theory, a theory that appeals to character in an attempt to provide for a better understanding of why it is, for example, that the sentence ‘I am here now’ cannot be uttered falsely. The notion of character that is relevant to this project has everything to do with the rules governing ‘I’ in English, and virtually nothing to do with the eccentricities of the individual speaker—with the ways in which she conceptualizes herself.12 When a speaker employs a token expression from a language such as English, then the character of that specific token is determined, at least in part, by the rules governing that term in the language; this is why we can say that utterances of the sentence ‘I am here now’ are true, regardless as to who the subject is, how she conceives of herself, her location, and so on. If we want character to play this kind of semantic role, then we cannot think of it as determined by individual usage, as being like the images, pictures, or descriptions that we happen to associate with our linguistic terms. Character is not something the speaker ordinarily creates and imposes on his expressions; it is, rather, something they inherit as expressions of a specific language. Indeed this has been recognized by Kaplan himself. In a recent discussion he abandons his original individualistic construal of character. He rejects the internalist’s epistemological intuition that a competent speaker is always in a position to know the character of her own utterances: “This is contrary to my claim in Demonstratives that the character of pure indexicals is known to every competent speaker. There I claimed that Character = Linguistic Meaning. I still believe that Character captures an important sense of Linguistic Meaning, but I have become more sceptical about the competence of competent speakers and about our access to what our words mean” (Kaplan 1989, pp. 277–278 n. 26). He also now rejects the identification of character with cognitive significance—recognizing that a number of expressions (e.g., coreferential proper names) might differ in cognitive significance for an individual even though they agree in character: In Demonstratives I tried to get at cognitive value through the notion of character. . . . I found it attractive to follow Frege in using a strictly semantical concept (character), needed for other semantical purposes, to try to capture his idea of cognitive value.
Anti-Individualism, Indexicality, and Character
97
As in the case of content, the possible-worlds style of formal semantics in Demonstratives represents character as a function. . . . The characters of two distinct proper names of the same individual would be represented by the same constant function, and, thus, under the functional interpretation, coreferential names would not differ in character. Since it is indisputable that distinct proper names have distinct cognitive values, the project of discriminating cognitive values of proper names by character is immediately defeated. The danger of trying to find characterological differences in distinct proper names is that the notion of character either will slip over from semantics to metasemantics or will become an ad hoc pastiche. In either case the dignified reality of character as the fundamental semantical value for indexicals would be seriously diluted. (Kaplan 1989, pp. 597–598, my italics)
This, of course, is just what I have been arguing: You cannot employ character as a semantic concept to be used in the analysis of indexical expressions and at the same time construe the character of an individual’s expressions as being a function of how he happens to use them. You cannot use character to capture the “cognitive similarities” between the twins. Conclusion Indexicals are strange brutes; they generate hordes of interesting questions, they challenge elements in the traditional (Fregean) picture, and the character/content distinction— between a nonpropositional content that remains constant across contexts (of utterances) and propositional content that varies from context to context—is clearly important. But, contrary to Kaplan’s own original intuitions, none of this is of any assistance to the individualist. Burge was on target when he rejected Putnam’s reading of the Twin Earth examples as illustrating the indexicality of kind terms; he was right in claiming that indexicality was irrelevant to the problems posed by the Twin Earth examples. And the same is true of its theoretical counterpart, character, and, the apparatus employed by Kaplan in Demonstratives. The Twin Earth examples are distorted when construed as examples of individuals using the same sentence (as one must to argue the character interpretation). Worse: Even if we suppose that the character/content distinction applies to kind terms, this is of no help whatsoever. Like the more familiar content, character fails to supervene on the nonintentional, and introspection, no matter how rational, may not be enough to determine sameness and difference in character. Unfortunately for the individualist, character exhibits all the problematic characteristics of propositional content: It is just another nonsupervening semantic entity. Notes 1. In addition, some theorists see this metaphysics as threatening the causal construal of psychological explanation. See, for example, Jerry Fodor (1987), Psychosemantics.
98
Joseph Owens
2. See also Woodfield (1982). 3. The most complete and the most influential account of Kaplan’s theory is to be found in “Demonstratives,” which circulated as an unpublished manuscript for more than twenty years and has now been published (Kaplan 1989). Page references are to the published version. This collection of essays also contains Kaplan’s recent reflections on these same issues: “Afterthoughts,” pp. 565–614. 4. See Kaplan (1977), pp. 508ff. For Kaplan, logical truth or validity is truth in all contexts (of utterance) as opposed to truth in all possible worlds (all possible circumstances of evaluation). See Kaplan (1989). 5. He writes: The logic of demonstratives determines a sublogic of those formulas of LD which contain no demonstratives. These formulas (and their equivalents which contain inessential occurrences of demonstratives) are exactly the formulas with a Stable Character. The logic of demonstratives brings a new perspective even to formulas such as these. The sublogic of LD which concerns only formulas of stable character is not identical with traditional logic. Even for such formulas, the familiar Principle of Necessitation (if |= Ø then |= Ø) fails. And so does its tense logic counterpart: if |= Ø, then |= (ÿPÿØ & ÿ|=ÿØ & Ø). From the perspective of LD, validity is truth in every possible context. For traditional logic, validity is truth in every possible circumstance. Each possible context determines a possible circumstance, but it is not the case that each possible circumstance is part of a possible context. In particular, the fact that each possible context has an agent implies that any possible circumstance in which no individuals exist will not form a part of possible context. (Kaplan 1977, pp. 548–549) 6. My comments here on the irrelevance of character mirror Burge’s original comments on Putnam’s indexical reading. 7. Indeed, the corresponding terms and sentences in Twin Earth examples can differ syntactically as well as semantically. See the ‘whale’/‘whail’ example in III (b). Forbes’s claim: “This contrast between two languages, on the one hand, versus indexicals in a single language, on the other, is in principle inconsequential . . .” (p. 213) is unfounded and begs the central questions at issue. 8. See “Individualism and the Mental,” section III. 9. Here, and elsewhere, I assume that there is no corresponding concern about the character of the counterpart’s utterances; the counterpart is not guilty of any conceptual error and his token sentences have the character they are assigned when viewed simply as sentences of Twinglish. 10. The story is more complex than this (because of differences in sentential structure, etc.), but I will ignore such complexities here. 11. The epistemic problems posed by the twins can be resolved; we can retain anti-individualism and preserve direct knowledge of the contents of our own thoughts without providing for an implausible inference from such knowledge to the nature of the external world. We can do all of this without any appeal to character. See, for example, Burge (1988) and Falvey and Owens (1994). 12. This point is emphasized by Kaplan himself, in his discussion of Frege on Dr. Lauben, on the way in which each individual is ‘presented to himself’. Frege writes: Now everybody is presented to himself in a particular and primitive way, in which he is presented to no one else. So when Dr. Lauben thinks that he has been wounded, he will probably take as a basis this primitive way in which he is presented to himself. And only Dr. Lauben himself can grasp thoughts determined this way. . . . He cannot communicate a thought which he alone can grasp. Therefore, if he now says ‘I have been wounded’, he must use the ‘I’ in a sense that can be grasped by others, perhaps in the sense of ‘he who is speaking to you at this moment’. . . . (Frege, “The Thought: A Logical Inquiry,” p 298. Quoted in Kaplan 1977) Kaplan comments: What is the particular and primitive way in which Dr. Lauben is presented to himself? What cognitive content presents Dr. Lauben to himself, but presents him to nobody else? Thoughts determined this way
Anti-Individualism, Indexicality, and Character
99
can be grasped by Dr. Lauben, but no one else can grasp that thought, determined in that way. The answer, I believe, is simply that Dr. Lauben is presented to himself under the character of ‘I’. (Kaplan 1977, p. 533, my italics) Here it is clear that the doctor’s images of himself do not determine the character, but rather the character associated with ‘I’ determines how he is thinking of himself. The individualists have it backwards.
References Almog, J. 1981. Dthis and Dthat: Indexicality Goes beyond That. Philosophical Studies 39: 347–381. Boghossian, Paul Artin. 1989. Content and Self-Knowledge. Philosophical Topics 17: 5–26. Burge, Tyler. 1979. Individualism and the Mental. Midwest Studies in Philosophy 4: 73–121. ———. 1982. Other Bodies. In Thought and Object, A. Woodfield (ed.). Oxford: Oxford University Press. ———. 1988. Individualism and Self-Knowledge. Journal of Philosophy 85(11): 649–663. Falvey, Kevin and Joseph Owens. 1994. Externalism, Self-Knowledge, and Skepticism. Philosophical Review 103(1): 107–137. Fodor, Jerry. 1987. Psychosemantics: The Problem of Meaning in the Philosophy of Mind. Cambridge, MA: MIT Press. Forbes, Graeme. 1995. Realism and Skepticism: Brains in a Vat Revisited. Journal of Philosophy 92: 205–222. Kaplan, David. 1977. Demonstratives. In Themes from Kaplan, J. Almog, J. Perry, and H. Wettstein (eds.). Oxford: Oxford University Press, 1989. ———. 1989. Afterthoughts. In Themes from Kaplan, J. Almog, J. Perry, and H. Wettstein (eds.). Oxford: Oxford University Press. Kripke, Saul. 1980. Naming and Necessity. Cambridge, MA: Harvard University Press. Putnam, Hilary. 1975. The Meaning of ‘Meaning’. Mind, Language, and Reality: Philosophical Papers, volume II. Cambridge: Cambridge University Press. ———. 1981. Reason, Truth, and History. Cambridge: Cambridge University Press. Woodfield, Andrew (ed.). 1982. Thought and Object: Essays on Intentionality. Oxford: Oxford University Press.
6
Competence with Demonstratives
James Higginbotham 1
Introduction
Some of the simplest uses of language involve two distinct phases. We set things up, by actions or by words, or simply take advantage of the local scene and the hearer’s expectations of us; and having set them up we make, in Wittgenstein’s words, our “moves in the language game.” What we say, in the most intuitive sense of that term, is constituted by the move we make, the setup being given; but what we let on that we believe, and intend others to believe that we believe, encompasses much more. (For this reason, it is easy to be misleading without lying.) Because the setting-up phase generally involves a perspective, individual or shared, of one or another party on an object, or on the interpretation of a predicate or other referential device, and because the same linguistic or other devices set up the reference or extension in ways that exploit the particular features of that setting, we cannot in general, having moved to a different setting, preserve reference and perspective simultaneously. There is a tension, then, when we attempt to express, from where we are, what a person says or thinks or wonders about from where she is or was or will be. Her statements or thoughts are structured in terms of her perspective, which yielded a reference according to linguistic and perhaps other communicative principles open and common to all. In our different setting, all crucial features of reference must be preserved, for the way we put her statements of thoughts must at least admit the same evaluation for truth. But then something of her perspective will be lost. Suppose that Joan thinks, “Should I go home for Thanksgiving?” Then she wonders whether she should go home for Thanksgiving. The transition I just made, from quotation to indirect reporting, loses the perspective that Joan had on herself when she wondered about herself in the first person. Not all of her perspective is lost, however. The word ‘home’, when used as a bare noun and with specific reference always refers to the home of a particular object of the discourse. Mary’s thought showed cross-reference, for she was wondering whether to go to her home. This cross-reference is preserved in my indirect report. But that does not help with expression of the first-person perspective. That the first-person perspective on herself was the perspective that Joan actually had might be indicated parenthetically. I could say that Joan wonders whether she (thinking of herself in the first person) should go home for Thanksgiving. The parenthetical, however, is not a constituent part of what Joan wonders; the information it contains belongs to the setting-up phase, and not to her thought. The above remarks are impressionistic, and are meant to be. I will argue, nevertheless, that the impression of conflict between the aim of preserving perspective and the aim of preserving reference, induced by the distinction between those aspects of language use
102
James Higginbotham
that belong to the setting-up phase and those that belong to the phase of saying things, survives under scrutiny and is even inevitable given the principles that govern our languages. What I am calling loss of perspective is seen not only in respect of context-dependent expressions, but also in respect of all elements that are contextually anchored, though not expressed overtly. For the special case of ellipsis, we can supply the words that might have been added but were not, as when ‘It’s raining’ is evaluated with respect to the local scene, or as a response to the question, ‘What’s the weather in Milan?’ In the latter case the speaker surely thought, and intended to be understood as asserting, that it was raining in Milan, so that the missing element can be supplied as a constituent of what was said; and in the former the speaker surely thought that it was raining (as she would put it) here, so that the intended location that needs to be supplied for evaluation is recovered through an indexical, reducing the case to the previous problem of context-dependent expressions. However, many cases where missing elements are contextually supplied cannot be construed as elliptical, if only because there need be no particular form of words that what is missing is elliptical for. A simple example is that of the incomplete definite description or quantificational noun phrase, where the range of the nominal is restricted, say, to objects on the local scene, as in (1). (1)
Each table is covered with books.
The speaker may have intended no particular restriction, but the range of the quantifier is confined to the local tables just the same. When those same tables are not part of the local scene, something more than the statement itself must be supplied to indicate which were the tables said to be covered with books. Whatever that is, it is foreign to the speaker’s original content. Ellipses aside, then, the blank spaces whose completion is essential to providing truth conditions, hence content, are as problematic as indexical expressions themselves: One cannot in general preserve both their reference and the salient contextual features on which it originally depended. Of the circumstances in which reference and perspective come apart, I spoke above in terms of tension and conflict. I believe that these terms are justified, and not only from a theoretical point of view. A substantial amount of philosophical thought and ingenuity have gone into relieving the conflict, for instance by packing into indirect discourse, and reports of propositional attitudes and epistemic states, parameters that, taken together, will allow the simultaneous linguistic expression of the now-remote perspective and the reference it there, or then, or under such and such circumstances secured; or alternatively into the construction of a means outside common speech for getting at what we say, or anyway let on, that people say and think.
Competence with Demonstratives
103
I will argue here that the conflict is unavoidable, and that nothing even in principle can resolve it. But ordinary linguistic practice, and therefore our linguistic competence in virtue of which we are able to participate in the practice, takes the conflict into account and gives us the means to live with it. We do convey perspective through the use of language, but we do not, and cannot, do it by saying what it is. Thus we have no difficulty in grasping what Joan wondered when she wondered whether she should go home for Thanksgiving; and we do this because we know, and it is common knowledge that, there is a first-person perspective that one has on oneself when using the word ‘I’ in accordance with the rules of the language, and we take it that Joan has such a means to refer to herself, and that she did so in wondering whether she should go home for Thanksgiving. We may think of the field of reference as the points in a space, and the perspectives as demonstratively established, and often egocentric, coordinate systems for the points. Reference to the same point from different perspectives is not a one-shot affair, but involves a general transformation of coordinate systems, mapping one entire set of perspectives into another. Moreover, the family of all sets of perspectives may be so organized that any one can be transformed into any other by a general routine, known to the speakers of a language, and known by them to be known to other speakers. Such mutual knowledge is part of linguistic competence, or so I will maintain. In the remainder of this essay I will endeavor to substantiate the views just advanced, through a consideration of linguistic competence with demonstrative and indexical constructions, and the semantics of complement clauses, and besides analyzing a well-known typical case I will try to support the metaphor of transformations of coordinate systems. If these views are on the right track, then the notion of semantic competence—knowledge of meaning—should be seen as including the principles governing demonstrative and indexical expressions. 2
Normal Forms for Demonstrative Reference and Truth
In 1974, Tyler Burge gave a powerful and persuasive account of the place of demonstrative constructions in theories of truth for historically given natural languages.1 I begin with two essential features of this account. Consider a simple demonstrative utterance, say (2), uttered while conspicuously pointing to a dog: (2)
This is a dog.
A theory of truth for our language should establish about (2) the truth conditions common to all utterances of its type, independently of how the reference of ‘this’ is secured. The type of the utterance is a syntactic structure S, and so it should establish (3):
104
(3)
James Higginbotham
If u is an utterance of S, then u is true iff a is a dog
where ‘a’ holds a place for the reference of the demonstrative. Furthermore, the account should say what is common to all the demonstrative uses of ‘this’. The latter is accomplished by relativizing the truth condition to the speaker’s act of reference (accomplished by pointing in the example given, but of course not confined to that). The result is the universal closure with respect to u, x, and s of (4): (4) If u is an utterance of S, and the speaker s of u refers with s’s utterance of ‘this’ therein to x and nothing else, then u is true iff x is a dog. The first feature of the account just sketched, and adapted from Burge’s original article, is that truth conditions as in the biconditional consequent of (3) show places on the right, marked there by ‘a’, that do not appear on the left. The format is therefore radically different not only from that for languages without indexical expressions, where the right side of the biconditional is always a closed sentence, but also from that advanced in the truthat-an-index accounts of truth conditions for sentences with demonstratives, which if put in terms of utterances would have been approximately as in (5): (5)
If u is an utterance of S, then u is true at the index a iff a is a dog.
Truth-at-an-index accounts generally stem from Montague (1968) and were pursued by a number of authors, including especially Lewis (1970). In these accounts the truth predicate is exactly as relative as there are places to be filled by demonstrative referents in the statement of truth conditions. The development in Kaplan (1977), an advance in several respects, is not essentially different in this one: Truth is relativized to possible worlds w and contexts c, and in the case of (2) we will have (6): (6) S is true in the context of c in w iff the reference of ‘this’ in c is a dog in w2. The second feature of Burge’s account, exemplified in the passage from (3) to (4), is the construal of the position marked by ‘a’ as occupied by a bound variable in the full statement of truth conditions. The strategy generalizes across demonstratives and indexical expressions, and what is more, across the nonelliptical elements that serve to restrict the range of a predicate. In the case of (1), for example, we will have (7): (7) If u is an utterance of the syntactic structure for (1) by s, and s confines the range of the utterance of ‘table’ therein to tables x such that X(x), and refers with the present tense as used therein to the time t(u) of u itself, then u is true iff for each x such that X(x), x is covered with books at t(u).
Competence with Demonstratives
105
I will call a statement like (7) a normal form for semantic theory.3 The general structure is as in (8): (8) If u is an utterance of S by s, and F (u, S, s, x, X), then u is true iff (u, S, s, x, X). The construction of normal forms for any substantial piece of an actual human language is a theoretical matter. More than this, however, even the successful presentation of normal forms for individual sentences, presupposing the lexicon of nonindexical words, does not of itself answer the obvious question of how the semantics of indexical elements is to be states. Borrowing terminology from David Kaplan (from his unpublished Gareth Evans Memorial Lecture, Oxford, 1995), let us broadly divide the linguistic pieces of an utterance into those with meanings, and those with rules of use. Ordinary nouns and verbs have meanings (which may nevertheless be contextually variable, for example along dimensions of strictness—recall J. L. Austin’s question of whether France is hexagonal). Words like ‘this’ do hot have meanings but have rules of use that are as much common coin as meanings are, which specify how they are to be employed in setting up an utterance to be evaluated for truth or otherwise appraised. I shall suppose that rules of use for words, like rules of use for tools and home appliances, are stated in imperatival form, as for example in (9) and (10): (9)
‘this N’ is to be used to refer to proximate, salient objects satisfying N.
(10) The periphrastic future ‘will’ is to be used to restrict times to those of some interval following the time if u. There is of course a distinction between expressions that are truly demonstrative as in (9) and those that are merely indexical, as in (10): The speaker has some latitude about what to refer to with ‘this N’, but none whatever about the periphrastic future, apart from limiting its extent. But the distinction follows from the form of the rule of use itself, together with the fact that it is common knowledge that it is the rule of use; for insofar as the rule fixes the contribution of the expression in all contexts, it will be taken in any given context as making as determinate a contribution as the rule prescribes. We know that there may be many proximate, salient objects satisfying N, even in a given context, and equally that there is just one stretch of time following any u. Hence ‘this N’ is demonstrative, but the future tense more nearly indexical. The first-person is purely indexical, at least on the assumption that an utterance has just one speaker. So the rule of use for it is elementary: (11) ‘I’ is to be used to refer to the speaker of u (and only in nominative case positions).
106
James Higginbotham
The use of a word under a rule of use is grasped as signifying that a certain act is taking place, because referring to a thing with a word is an act. But of course the utterance of the word cannot signify that the act is taking place. There cannot be a word signifying that the object referred to is in fact proximate and salient, or that the speaker believes it to be, or believes himself to have referred to something proximate and salient. Donald Davidson pointed out in his (1979) that a hypothetical assertion-sign, signifying that the speaker actually is, or believes himself to be, making an assertion, would once admitted immediately be used on stage, where it signifies no such thing. Similarly, a sign that one was actually or anyway believed oneself to be referring to a proximate and salient object would leave no room for the actor to say, “Is that my old friend Mary?” while gazing offstage left at nothing in particular.4 3
Complement Causes
Having assembled the elements of Burge’s account of demonstrative reference and somewhat generalized it, we may take the consequences that substantiate the impressionistic distinction between the setting-up phase and the performance phase of a typical utterance. According to the account, all demonstrative and indexical reference, including the unstated restrictions on a predicate, flows from rules of use and belongs entirely to the setting-up phase, so that the content of the rules does not enter the truth conditions of what is said. This consequence agrees with Kaplan’s (1977) discussion of demonstratives and indexicals as referring rigidly across counterfactual situations. That demonstrative reference belongs to the setting-up phase is not a reflection of rigidity, or an equivalent formulation of it, but rather explains it. It follows also, given the split of the lexicon into items with meanings and items with rules of use, that we are generally bound to lose the perspectives of those whose sayings or thoughts we report indirectly. Once (2) is embedded, as in (12), and we put its truth conditions into normal form, nothing is left of the demonstrative but its reference: (12)
Mary wonders whether this is a dog.
If circumstances are favorable, I may conspicuously indicate to my hearer that the perspective on the basis of which I call her attention to the object of which I am saying that Mary wonders whether it is a dog is the very perspective that (I believe) Mary had on it when she thought to herself, “Is that a dog, I wonder?” (“Look at it from over here, and you’ll see why Mary wonders whether this is a dog”). In these cases, we may think of the act of reference as carried out in the setting-up phase, not of the whole of (12), but rather within the complement clause. But to examine the possibilities more closely, it will be necessary to formulate the truth conditions for such clauses.
Competence with Demonstratives
107
I will assume here without argument an account of the reference of complement clauses that I have defended elsewhere (Higginbotham 1986, 1991): They refer to their own syntactic structures, understood as if the speaker uttered them.5 Let S be the syntactic structure for ‘Snow is white’. Then ‘Bill Clinton believes that snow is white’ is true just in case Bill Clinton believes S, understood as it would be were it an utterance of mine, where what it is for it to be so understood is for the conditions for its truth to be known as I know them.6 Syntactic structures can be understood as having objects other than words as constituents, so that quantification into the standard contexts is not problematic. What we say are words, but we admit syntactic structures with objects in some positions, as in (13): (13)
[[Bill Clinton] [‘is F’]]
containing the person Bill Clinton in the subject position, and we stipulate that the truth conditions as apprehended by the competent speaker are that Bill Clinton be among the objects of which ‘is F’ is true; and similarly for other categories of expressions. Suppose that we conceive the act of reference to the dog (call him ‘Rex’) by ‘this’ in (12) to be part of the setting-up of the whole of (12). Then the wonderment attributed to Mary concerns (14): (14)
[[Rex] [‘is a dog’]]
understood as if the speaker had said it. So to understand (14) is to know of the dog Rex that (14) is true if and only if he is a dog. Thus the target statements for utterances of syntactic structures with demonstratives will contain objects, and elements in the range of second-order variables, in positions where names and predicates would go in structures made up only of words, and understanding these statements will be knowing about those objects and the second-order referents that the structures are true just in case they satisfy some target condition. But now it follows that whether the act of reference were part of the setting-up of the whole of (12) or only of its complement clause, no difference would ensue; the perspective that the speaker uses and the hearer grasps will not except by accident be any part of what Mary is said to be wondering about. The liberty granted by the account of demonstrative reference in terms of normal forms as discussed above, to construe the act of reference as taking place with respect to a whole utterance or merely with respect to its complement clause, therefore makes no difference in respect of the content contributed by that clause. This consequence may underscore the point that Kaplan’s generalizations follow from the theory of the truth conditions of utterances of sentences with demonstrative elements, as given by the rules of use for those elements and the normal forms for the theory of truth and linguistic competence.
108
4
James Higginbotham
Coordinate Transformations
Consider the egocentric coordinate system for days: . . . 2 days ago, yesterday, today, tomorrow, 2 days hence, . . . It is the whole-number sequence whose zero point, today, changes with the day. We regularly use elements from one point in the sequence to refer to days that those to whom we report would refer to with another point, as in (15), as said by me on Sunday: (15)
Mary said on Friday that she would leave yesterday, but she didn’t leave then.
As I have noted elsewhere (Higginbotham 1994), Gareth Evans in response to John Perry’s thesis that indexicals supply only a reference and no Fregean sense, suggested that words like ‘yesterday’ are understood with what Evans called a “completing sense”, and Perry in turn has responded to Evans that the completing sense of ‘yesterday’ in (15) diverges from that of the corresponding word, ‘tomorrow’, that was no doubt used by Mary, so that the account is false to the Fregean semantics of indirect discourse and attitude reports (Evans 1981; Perry 1993). But I argued also that neither a completing sense in Evans’s sense, nor a view of indexicals as lacking sense, was warranted by the facts; rather, we count the embedded clause in (15) as faithful to Mary’s speech when she said on Friday, “I will leave tomorrow,” because we presuppose that all parties to the discourse know and use the egocentric coordinate system. Evans and Perry are both right in a way; but neither has distinguished between senses and rules of use, known to the speaker. Given the distinction, we have that what is required to understand ‘yesterday’ is not a completing Fregean sense as in Evans, but rather a rule of use that not only locates that day as prior to the day of thinking or speaking, but also embeds it as part of the above coordinate system to which it belongs. Conversely, Perry’s response to Evans fails to note that my perspective on Saturday when I said (15), although not the same as Mary’s when she spoke, is transformable into that perspective by sliding ‘today’ back two notches, thus superimposing her ‘today’ onto my ‘two days ago’, and her ‘tomorrow’ onto my ‘yesterday’. The example (15) carries another moral, I believe: It is by no means necessary to suppose that my ‘yesterday’ goes proxy for, or carries in addition to its reference, any imagined replacement such as in (16): (16) Mary said on Friday that she would leave the day after that, but she didn’t leave then. For one thing, the content of my ‘then’, which is anaphoric to the embedded ‘yesterday’, is just what it would be if I had said ‘yesterday’ itself. For another, we should miss the point that when Mary said ‘tomorrow’ it is not as if she said ‘the day after this one’ (indi-
Competence with Demonstratives
109
cating a day), but rather as if she said ‘the day after today’, using a purely indexical expression; so the replacement gets us no closer to Mary’s own perspective. I conclude that the transformation of coordinates exemplified by this simple example is the crucial element involved in the conveyance of perspective. But other cases are not so simple and may have to be configured as they arise. 5
Puzzles of Perspective
We have been considering complement clauses as they are used to classify acts of speech and states of mind, broadly speaking, thus including speech acts and epistemic states where the truth or falsehood of the complement is presupposed. The states and acts themselves are assumed to have whatever content they have, and our questions have concerned the extent to which our classifications can be faithful to that content, given that the perspective on the reference in the state or original speech may under the situation in which the classification is to be made yield a different reference, or none, and that the reference figuring in the content may allow indication only by means foreign to the original. I will concentrate on just one of the many puzzles and conundra that have figured in discussion, due to Mark Richard (1983). A is talking on the telephone to B, whom A knows well. A is in a telephone booth on one side of Elm street, and he sees on the other side a woman in another telephone booth. A sees also that a steamroller is bearing down on her, and is of course horror-stricken to see that she is in such danger. A does not know it, but the person in the booth is B. Describing the situation over the telephone, A may well say (17): (17)
I believe she is in danger.
B, who can see the other side of Elm Street and the man in the telephone booth there, observes his panicky behavior and guesses, correctly, that he thinks she is in danger. If B were to say (18)
The man watching me thinks I’m in danger,
she would speak truly. Therefore if A were to say, addressing B (19)
The man watching you believes you are in danger,
A would also speak truly. But as a matter of fact if A were to say (20)
I am the man watching you,
he would also speak truly. Putting equals for equals in (19), we have that (21) is true as said by A:
110
(21)
James Higginbotham
I believe you are in danger.
But if B were to ask A, “Do you think I’m in danger?” then A would respond, correctly, “No, I don’t think you’re in danger.” Contradiction. The response to the puzzle suggested by Richard (1990, pp. 190ff.) and by Mark Crimmins (1992, pp. 29–32, 164) is that more should be put into the content of what is said and would be understood than is given by the words. (Views of this type can be traced at least to Schiffer 1977; see also Schiffer 1987.) There are a number of technical points connected with Crimmins’s and Richard’s views, and somewhat different treatments are suggested in Salmon (1986), Higginbotham (1991), and others. But all of these have in common the idea that we should understand (17) and (21) relative to different perspectives (or “modes of presentation”), taking them up in the semantics as if the perspectives occurred where the indexicals sit.7 Abstracting form the details of the various presentations, observe that A would speak falsely were he to say (22): (22)
I believe that [the woman I am talking to] is in danger,
where what appears in the square brackets—the parenthetical of my discussion above— has replaced the simple ‘you’. But A would speak truly were he to say either (23) or (24): (23)
I believe that [the woman across the street from me] is in danger.
(24) The man watching you believes that [the woman across the street from him] is in danger. If we understand the passage from (19) to (21) as shifting the context so that the secondperson pronoun is a proxy for, or invokes in addition to its reference, the square-bracketed constituents, then we can see why the substitution licensed by (20) should alter our judgment. However exactly the technical details go, all is well under the rewriting envisaged: No contradiction with intuition ensues. There is a further point to be noted in connection with Richard’s ingenious example, which will be taken up further below. Every instance of (25) must be true in A’s speech: (25)
The man watching you believes that p ´ I believe that p.
Suppose that B actually says (18) to A, and A suspends judgment about whether B really is in danger, or the man across the street from her is panicking unnecessarily. Then (24) is as true as before, but (26) is false, or perhaps anomalous: (26)
I believe that [the woman across the street from him] is in danger.
Competence with Demonstratives
111
The reason is that the pronoun ‘him’ is functioning as a bound variable in (24), but not in (26). There is no contradiction with (25), since the principle holds only for closed sentences, not open sentences. Preserving cross-reference gives the true (23). Obliterating it by a further substitution taken from the perspective A has from B gives (27): (27) I believe that [the woman across the street from the man watching you] is in danger, which is false. Preservation of cross-reference is therefore essential, and we must distinguish For x = a, x believes that . . . x . . . From a believes that . . . a . . . even where ‘a’ is the indexical ‘I’.8 Returning to the main theme, we have observed that transforming the clauses by substituting perspective for indexical removes the counterintuitive or paradoxical appearances. I refer to this as the substitution strategy for removing the puzzles.9 The strategy, if correct, must reflect what we do and intend with language, clarifying the rational basis for our intuitive judgments. An alternative strategy, due to Stalnaker (1981), is to evaluate A’s and B’s actual or potential utterances as they would be evaluated if their beliefs were true. It is part of the story above that A believes he is not the man watching B and that B is not the woman in the telephone booth across the street from him. In a possible world in which these beliefs are true, there are four persons and two sets of street-opposite telephone booths (and two steamrollers). Let B¢ π B be the woman across the street from A, let A¢ π A be the man across the street from B, and let us preserve the steamroller scenario as before. Then A’s ‘she’ refers to B¢, who is in danger, his ‘you’ refers to B, ‘the man watching you’ to A¢, and ‘I’ to himself. Then (17) and (19) are true as said by A; (18) is true as said by B; and (20) and (21) are false as said by A. I repeat the examples here: (17)
I believe she is in danger.
(18)
The man watching me thinks I’m in danger.
(19)
The man watching you believes you are in danger.
(20)
I am the man watching you.
(21)
I believe you are in danger.
112
James Higginbotham
The semantics takes ‘believes’ (or more strictly ‘a believes’, for a an individual) as a modal operator, so that for a proposition to be believed is for it to be true in every possible world doxastically accessible to a; but now Stalnaker proposes the twist, that the English indexical expression in the complement clauses are to be evaluated in a possible world, not with the reference that they actually have, but with the reference that they would have as used there. These diagonal propositions, as he calls them, are what is expressed (or up for evaluation) in the complements of (17) through (19), and (21). There are important differences between the substitution strategy and the diagonal strategy just illustrated. But it is clear that the reconstruction of the perspective taken by the person whose words or thoughts are reported is the common aim of both. For many purposes, the results of the two strategies will coincide. Given the diagonal strategy, we can generally construct the proper substitution by considering what the reference of the demonstrative would have been, and what perspective would have been taken upon it, in an appropriate counterfactual environment; and conversely, given a substitution that we can recognize as correct, we reach the crucial possible world on the diagonal by endowing the counterfactual environment with properties that make the demonstrative take on the appropriate reference.10 6
Perspective and Truth
Does the Richard puzzle call for a revision of the basic conception underlying semantic theory? Some devices that have been suggested in the literature (including attempts by the present author) would effectively answer this question affirmatively. There is no space to outline these devices here, which in any case are rather different from one proposal to another. Generally speaking, they have in common the introduction, or reintroduction, of a “Fregean” element, a mode of presentation or possibly simply a notation, which will sit alongside or in place of the reference that the demonstrative secures. In Stalnaker’s procedure, the Fregean element appears indirectly, in the form of the specification of the counterfactual situations against which the indexical sentence is evaluated on the diagonal. Consider again Richard’s first example, (17): (17)
I believe she is in danger,
said by A. The semantics sketched in section 2 above gives as what the competent speaker of English knows about A’s utterance exactly (28) (ignoring the tense): (28) If A refers to himself with his utterance of ‘I’, and to x with his utterance of ‘she’, and x is female, then his utterance of (17) is true iff A believes the syntactic structure [[x] [‘is in danger’]], understood as A said it.
Competence with Demonstratives
113
We know what is going on in the story, enough to know that the consequent of (28) can be detached, with B as value of ‘x’. For the content of the belief we are left, not with the robust (29), but with the emaciated (30): (29)
[[‘the woman across the street from’ A] [‘is in danger’]]
(30)
[[B] [‘is in danger’]]
But it is (29) that we must appreciate, not (30), if we are to grasp the import of A’s speech. Precisely this task, however, is one that we can perform, just because the demonstrative contributes nothing but its reference. The point generalizes, as follows: When X reports the thoughts or speech of Y, X’s demonstrative does not contribute Y’s perspective (which is generally dead and gone anyway), and X’s own perspective is foreign to Y’s thought, although X does not (and if a demonstrative mode of reference is to be employed, cannot) express it. Since X and Y as competent speakers both know this, Y can fathom X’s intention. Communication then takes place without residue, or guesswork. If this line of thought is even roughly correct, then Fregean elements need not be imported into indirect discourse and similar contexts, or rather they import themselves. The Fregean elements are not senses, however: They may involve objects intrinsically, and they may be perceptual rather than linguistic (it was for this reason that I used the neutral term of art “perspective”). They are not, and sometimes cannot be, said. But the rules of use for demonstratives make way for them to be intended, just because the rules deliver only objects. 7
Concluding Remarks
A number of semantic problems have turned on the fact that indexical and demonstrative reference, made through a perspective such as a rule of use or by other contextual means, including especially perceptual salience, typically takes place in a setting where the perspective gets through to the referent at the particular point where the act of reference from that perspective takes place, and not at other points. It follows that we cannot in indirect speech be faithful to both. We nevertheless convey both, because we can exploit our common knowledge of the general system of demonstrative and indexical reference, and of particular circumstances as they arise. For simple cases such as indexical time reference, we succeed because we can correlate egocentric coordinate systems by a simple transformation. For complex cases such as Richard’s puzzle, we grasp from the surrounding story what is going on with A and B, and we do this by grasping how the demonstrative reference was secured. A semantics for our language—that is, a semantics that aims to express explicitly what we know when we know our language—should
114
James Higginbotham
incorporate rules of use, and explain how we make use of the surrounding information to arrive at the now remote perspective from which the demonstrative reference was made. The relativization of truth to acts of reference, suggested by Burge, is a first step, and the systematic transformations of coordinates perhaps a second. Further reconstructive projects can be envisaged, although questions of theoretical tractability arise here as in other domains. But we may hope as these steps are taken to arrive at a fuller understanding of our linguistic competence. Notes 1. An important part of this account is due independently to Scott Weinstein (1974); but Weinstein’s note was not, and was not intended to be, more than a demonstration that demonstrative reference could be accommodated within a standard view of a theory of truth. 2. Kaplan’s formulation speaks of sentences in contexts rather than utterances, but contexts will be discriminated sufficiently finely that the utterance (if there is one) will be determined from the sentence and the context. 3. In this I continue the terminology used in Higginbotham (1988). 4. Kaplan’s distinction suggests that the imperative mood, like an indexical, is associated with a rule of use, rather than a meaning. If so, then imperative sentences would not have truth conditions although they may have (much more complicated) compliance conditions, depending on what the form was being used to do. 5. This way of putting the proviso about how they are to be understood is due to Burge (1978). 6. More strictly, I assume that the speaker knows, and expects the hearer to know, a target statement of truth conditions reflecting semantic competence. (Such statements are not in general trivial or “disquotational.”) In general, the hearer understands S if he knows that it is true if and only if p, where that statement is the target. To understand this attribution of knowledge one must understand the target for ‘S is true if and only if p’, say that it is true if and only if g, but having the knowledge and understanding the attribution of it are different things. 7. Similarly, for many discussions including Richard and Crimmins, for comparable issues involving proper names. 8. I believe that this distinction is at work in the intuitions underlying what David Lewis has called de se interpretations of the first person (Lewis 1979). The point of view defended here is incompatible with there being such an interpretation, since it is incompatible with there being any more to the first person than the elementary rule of use. 9. It is not as if the person reported on might as well have used what the substitution strategy delivers. The substituend is longer than the original, so in the interest of saving breath one would not expect it to be used unless necessary. There is also the fact that, the system of indexical reference being in place, the substitutions if actually carried out would give the wrong impression—if I were to substitute ‘the person I am talking to’ for ‘you’, as though being coy about the identity of my addressee, then I would, to put it mildly, be the regular recipient of funny looks. In Richard’s example, A ought to say ‘you’, even where the intended understanding is ‘the woman across the street from him’, with ‘him’ functioning as a bound variable, as we have seen. 10. Crimmins (1992) uses the substitutional method, but retains the “singular proposition” (i.e., the truth conditions with respect to the reference) as part of the report.
References Burge, Tyler. 1974. Demonstrative Constructions, Reference, and Truth. Journal of Philosophy 71: 205–223. ———. 1978. Self-Reference and Translation. In Meaning and Translation, F. Guenthner and M. GuenthnerReutter (eds.), pp. 137–153. New York: NYU Press.
Competence with Demonstratives
115
Crimmins, Mark. 1992. Talk about Beliefs. Cambridge MA: MIT Press. Davidson, Donald. 1979. Moods and Performances. In Meaning and Use, A. Margalit (ed.). Dordrecht: D. Reidel. Reprinted in Davidson, 1984, Inquiries into Truth and Interpretation. Oxford: Oxford University Press. Evans, Gareth. 1981. Understanding Demonstratives. In Meaning and Understanding, H. Parret and J. Bouveresse (eds.). Berlin: de Gruyter. Reprinted in Evans (1985), pp. 291–321. Higginbotham, James. 1986. Linguistic Theory and Davidson’s Program in Semantics. In Truth and Interpretation: Perspectives on the Philosophy of Donald Davidson, E. LePore (ed.), pp. 29–48. Oxford: Blackwell. ———. 1988. Contexts, Models, and Meanings: A Note on the Data of Semantics. In Mental Representations: The Interface between Language and Reality, R. Kempson (ed.), pp. 29–48. Cambridge: Cambridge University Press. ———. 1991. Belief and Logical Form. Mind and Language 6: 344–369. ———. 1994. Priorities in the Philosophy of Thought. Proceedings of the Aristotelian Society 68 (suppl.): 85–106. Kaplan, David. 1977. Demonstratives. In Themes from Kaplan. J. Almog, J. Perry and H. Wettstein (eds.). Oxford: Oxford University Press. ———. 1995. Meaning and Use. Gareth Evans Memorial Lecture, University of Oxford. Unpublished. Lewis, David. 1970. General Semantics. Synthese 22. Reprinted in Lewis (1983), Philosophical Papers, vol. I. Oxford: Oxford University Press. ———. 1979. Attitudes de dicto and de se. Philosophical Review 88: 513–543. Reprinted in Lewis 1983, Philosophical Papers, vol. I. Oxford: Oxford University Press. Montague, R. 1968. Pragmatics. Reprinted in Formal Philosophy, R. Thomason (ed.). New Haven: Yale University Press. Perry, J. 1993. Postscript to “Frege on Demonstratives.” In The Problem of the Essential Indexical and Other Essays, pp. 26–32. Oxford: Oxford University Press. Richard, M. 1983. Direct Reference and Ascriptions of Belief. Journal of Philosophical Logic 12: 425–452. ———. 1990. Propositional Attitudes: An Essay on Thoughts and How We Ascribe Them. Cambridge: Cambridge University Press. Salmon, Nathan. 1986. Frege’s Puzzle. Cambridge, MA: MIT Press. Schiffer, Stephen. 1977. Naming and Knowing. In Contemporary Perspectives in the Philosophy of Language, P. French, T. Uehling Jr., and H. Wettstein (eds.), pp. 61–74. Minneapolis: University of Minnesota Press. ———. 1987. The ‘Fido’-Fido Theory of Belief. In Philosophical Perspectives 1: Metaphysics, J. Tomberlin (ed.), pp. 457–480. Atascadero: Ridgeview. Stalnaker, Robert. 1981. Indexical Belief. Synthese 57: 129–152. Weinstein, S. 1974. Truth and Demonstratives. Noûs 8: 179–184.
7
Implicit Conceptions, Understanding, and Rationality
Christopher Peacocke I will be advocating the importance of what I will call implicit conceptions in the theories of linguistic understanding, of concepts and of rationality. I will try to say something about why we need implicit conceptions; how we can discover them; what they explain; and what they are. Implicit conceptions also seem to me capable of furthering our understanding of some classical and recent issues in the theory of meaning and knowledge. Several forces pulled me toward this position on implicit conceptions. One was a growing dissatisfaction with the treatment of primitive axioms and rules given in A Study of Concepts (1992). Another was reflection on what is involved in rational acceptance of new principles that do not follow from those a thinker already accepts. A third force was my attraction to the conception of sense expounded and developed in Tyler Burge’s paper “Frege on Sense and Linguistic Meaning” (1990). Indeed there is more than one link between the position I will present and Burge’s many illuminating works on the notion of sense. I hope the present paper will support the suspicion that there is, in a Festschrift, a way of honoring an author’s thought other than attempting to give it a thorough refutation. 1
Implicit Conceptions
Consider someone who is introduced to a primitive logical axiom, or to a primitive logical rule. This person might be yourself, when you were first taught logic at around the age of eighteen. Your introduction might be to an axiom A Æ (A or B), or it might be to the inference rule ‘From A, the conclusion A or B can be inferred’. There is such a phenomenon as a thinker in your situation reflecting, drawing on his understanding of the expressions in the rule, and coming to appreciate the axiom or rule is valid. What is going on when such reflection takes place? The example is specified as one in which the axiom or principle is a primitive one: it is not something that is derivable from other axioms or rules. So the movement of thought in which our rational, reflective thinker is engaged cannot be one of straightforward inference. Nor is it a matter of accepting a stipulation involving some newly introduced symbol. The axiom or rule is appreciated, on reflection, as correct when taken as involving the very same words, such as ‘or’, which an eighteen-year-old learner of logic, for instance, will have understood for more than fifteen years. Nor is it plausible that our thinker has to draw on memories of his own previous uses of the word ‘or’ on particular occasions. The logical principle is not about his use of the word. In any case, if he is like me, he will not remember any particular occasions as ones on which he used that very word. All the
118
Christopher Peacocke
same, he can still reflect, drawing on his understanding of the word, and come to appreciate that the axiom or principle is valid. Our thinker’s knowledge cannot always be explained as a result of his having explicitly inferred the validity of the axiom or principle from his explicit knowledge of the truth tables for the connectives involved. This cannot be a fully satisfying explanation for two reasons. First, our thinker can reflect and rationally appreciate the validity of these principles before having been explicitly taught any truth tables. Second, and crucially, we must also think about rational acceptance of the truth tables themselves. Each of us, when first presented with the truth tables for the unproblematic connectives, was able to reflect, and on the basis of our understanding of the expressions, come rationally to appreciate that the particular truth table is correct. This is itself a further illustration of the kind of phenomenon we are trying to explain. No doubt there are various detailed ways in which reflection may proceed in the original case of the axiom or principle, but one of them is as follows. Like the other variants in which the details differ, the reflection involves a simulation exercise. The thinker imagines—to start with one of the cases—that he has the information that A is true and B is false. His aim is to address the question of whether the alternation ‘A or B’ should be regarded as true or false in the imagined circumstances. As in any other simulation exercise, he then exercises a capacity offline. This capacity is the very same, understandingbased capacity he would be exercising in a real case in which he had the information that A is true and B is false and has to evaluate the alternation ‘A or B’. As in the corresponding real case, in the imaginative exercise he goes on to hold that ‘A or B’ will be true in the imagined circumstances. In coming to hold that ‘A or B’ is true in the simulated circumstances, our thinker employs only the information about the truth-values, within the simulation, of A and of B, together with his understanding of alternation. He does not draw on any other resources. Next our thinker proceeds to consider imaginatively another case, say that in which A is true and B is true. Soon, we hope, as he proceeds in this way it dawns on him that there will be no cases in which the antecedent, or premise, is true, and the consequent, or conclusion, is false for the axiom or inference-rule respectively. Thus he comes rationally to accept the axiom or rule as valid. The same procedure and resources will equally allow him to come to accept rationally each line of the truth table for some connective he understands. When axioms, inference rules, or lines of truth tables are reached in this way, it seems to me that the resulting judgments constitute knowledge.1 This account of the reflection gives a clear explanatory priority to the thinker’s understanding-based capacity to evaluate particular alternations, such as “Either he went left or he went right,” and particular conjunctions, and other complex statements, on the basis of information about their components. This is a capacity that a thinker can possess
Implicit Conceptions, Understanding, and Rationality
119
and exercise, and normally does do so, prior to having any explicit knowledge of general logical principles or of truth tables. It is this capacity that is run offline in the simulation. It is a capacity involved in the very understanding of the connectives. Its role in the imaginative exercise makes the case one in which the thinker draws on his own understanding of the expressions in coming to appreciate, via this reflection, that the axiom or principle is valid. I suggest further that the thinker’s understanding of the connective ‘or’ involves (and perhaps is even to be identified with) his possession of an implicit conception, a conception with the following content: that any sentence of the form ‘A or B’ is true if and only if either A is true or B is true. This implicit conception is influential in the thinker’s evaluation of alternations given information about the truth-values of their components. The influence is exerted not by the thinker inferring something from the content of the implicit conception. He need not have any explicit knowledge of its content. Rather, his having the implicit conception explains his particular patterns of semantic evaluation of the complex, given information about the truth-values of its constituents. Derivatively, it is this implicit conception that is influential in the simulational part of the reflection, which eventually leads him to accept certain primitive axioms and inferential rules involving alternation. This, then, is a description at the personal level of a way in which a thinker may come rationally to accept a logical principle, a way that is not simply a matter of inferring it from other previously accepted object-language principles. Before I give other examples of implicit conceptions, I want to clarify certain features of this noninferential but rational means of acceptance. (a) The very simple description I have given of the rational acceptance of a logical axiom is not meant to enable us to resolve the dispute between classical and constructivist, or any other, interpretation of the logical constants. Nor could it provide such a resolution. The phenomena cited in this simple description of the case are phenomena of a general kind that would equally need to be mentioned in an account of how it is that an ordinary, nonphilosophical thinker can come to appreciate that certain axioms are valid, even if a constructivist theory of meaning were correct. The constructivist is likely to elucidate validity of a transition as the transformability of any means of establishing its premises into a means of establishing its conclusion. To work out whether this definition applies to a particular form of transition, the ordinary thinker will have to use simulation to gain knowledge of the ways in which he takes statements of certain forms to be established. Imaginative simulation will be involved in any case in which the thinker is drawing, at least on early occasions, on the understanding he exercises in ordinary, real-world applications. This is something common to classical and to constructivist approaches. Any
120
Christopher Peacocke
resolution of the dispute between them must appeal to a quite different body of considerations. (b) The described means of rationally coming to accept a primitive law is a fallible means. A thinker may overlook a combination of truth values, or he may perform the simulation incorrectly. He may fail to run the very same procedure for evaluation offline as he would exercise online. He may misremember information derived from earlier simulations in which he was checking cases. He may use a procedure in imaginatively assessing particular cases that is not just understanding-based, but draws on auxiliary information specific to those particular cases. Much, then, may go wrong. Nonetheless, when the procedure is properly executed, the resulting belief in the logical law has an a priori status. No perceptual state, nor the deliverance of any other causally sensitive faculty for finding out about the world, is playing an essential justificational role in the thinker’s rational acceptance of the logical law when it is arrived at in this way. This combination of fallible capacities that, when exercised properly, are nevertheless capable of yielding a priori knowledge, is something with which we are very familiar in other routes to a priori knowledge. (c) An objector may protest that simulation can never give knowledge of what would be true in the circumstances imagined in the simulation, but can give knowledge only of what the simulating thinker would judge or believe in the imagined circumstances. This, though, seems to me to be false. Simulations, properly executed, can give information about the world, as well (of course) as information about the thinker’s mental state in various hypothetical circumstances. Suppose you are asked the question: “If you walk south down Whitehall, and turn left over Westminster Bridge, when you are on the bridge, what building is slightly to the left of straight ahead of you?” You answer this by imagining yourself following the described route. When, by this means, you reach the conclusion that when on the bridge, the former County Hall would be slightly to the left of straight ahead, this is a means of obtaining information about the world. If the thinker is drawing on knowledgeable states in performing the simulation, it is also a means of obtaining knowledge about the world. It is important to emphasize that the conditions initially specified to hold in the simulation, both in this spatial example and in our logical case, concern not merely what the subject believes in the simulated circumstances, but what is true in the simulated circumstances. (d) A thinker of a certain frame of mind sometimes classified as neo-Wittgensteinian may wonder whether there is really any objectivity in what is obtained by the simulation procedure as applied in the logical case. The point of this paper is not to take on central Wittgensteinian issues, and for present purposes I just note the plausibility of the following biconditional. The results of the simulation, properly carried out, will have the required
Implicit Conceptions, Understanding, and Rationality
121
objectivity if and only if there is objectivity in a thinker’s corresponding response to a new case in the real, nonsimulational, world. If there is objectivity of the latter, that is if it goes beyond merely an impression of correctness, then the capacity exercised online in the real-world cases can be drawn on in carrying through the simulation. That concludes the preliminary remarks on the nature of the simulation in this first example. It is not hard to reach, by reflection, principles distinctive of alternation, and in doing so to be appropriately influenced by one’s underlying implicit conception. It is not even hard, in that particular example, to make the content of the implicit conception explicit. In other examples, neither of these things is so. There are some cases in which a thinker has an implicit conception, but is unable to make its content explicit. The thinker may even be unable to formulate principles distinctive of the concept his possession of which consists in his having that implicit conception. One of the most spectacular illustrations of this type of phenomenon is given by the famous case of Leibniz’s and Newton’s grappling with the notion of the limit of a series, a notion crucial in the explanation of the differential calculus. It would be a huge injustice to Leibniz and Newton to deny that they had the concept of the limit of a series, or to deny that they had propositional attitudes describable by using the world ‘limit’ within the ‘that’ clauses. What they could do was to differentiate particular functions, and they had no difficulty in saying what the limit of a particular series of rationals was. I would say that each of these great thinkers had an implicit conception that guided his application of the phrase ‘limit of . . .’ in particular applications. What they could not do, despite repeated pressing by critics and well-wishers, was to make explicit the content of their implicit conceptions. When pressed for explications, Leibniz spoke of values that were infinitely close to one another. This is something we can now make sense of in the theory of infinitesimals, but it was quite illegitimate within the ontology of real numbers within which Leibniz was working. Newton spoke of ‘limiting values’, ‘ultimate ratios’, and the like, but these were not given a steady explanation. Sometimes the procedures given even seem to require dividing by zero. Newton comes extremely close to a correct explication at one point, but gives that explanation no special salience among the others. If their explications were really the best that could be given, it would be hard not to sympathize with Berkeley’s critique of the calculus. Even John Bernoulli, in trying to sort the matter out, wrote sentences like this: A quantity which is diminished or increased by an infinitely small quantity is neither increased nor decreased. (Stewart 1996, p. 77)
As is well known, it was not until Bolzano, Cauchy, and arguably even until Weierstrass in the mid–nineteenth century that a completely clear, unproblematic explication of the
122
Christopher Peacocke
limit of a series was achieved, the familiar epsilon-delta definition. L is the limit of the function f (x) as x approaches a if for any positive number e, there is some number d such that f (x) minus L is less than e whenever x minus a is less than d. In this explication there is, famously, no unexplained talk about ultimate ratios, infinitely small values, or anything that even appears to involve dividing by zero. To make an implicit conception explicit can, then, on occasion be a major intellectual achievement.2 The case of Leibniz, Newton, and limits also serves to illustrate another point. We do sometimes ascribe attitudes to contents containing a concept to a thinker, even when a thinker has only a partial understanding of the expression for the concept, provided that the thinker defers in his use of the expression to others in the community who understand it better. That phenomenon has been very well described by Burge (1979). But we ought not assimilate the example of early uses of the limit-concept to cases of deference, for the facts explained by implicit conceptions cannot be explained away by appealing to deference. To whom were Leibniz and Newton supposed to defer? There was no one else who understood the notion better. Nor, one may conjecture from each of their characters, was either of these two gentlemen of a mind to defer to anyone else on these matters. Leibniz’s and Newton’s use of the limit-concept is rather an example of what Frege called grasping a definite sense while also failing to grasp it ‘sharply’. The present paper is in effect an exploration of what is involved in employing concepts that are not ‘sharply grasped’. Actually, the very example of the limit-concept occurs in a list in the first section of Frege’s Foundations of Arithmetic: “The concepts of function, of continuity, of limit and of infinity have been shown to stand in need of sharper definition” (1953, §1). The early use of the concept of a limit in Leibniz and Newton is a concrete historical illustration of a state of affairs whose possibility is articulated by Burge. In the course of elaborating Frege’s conception of grasp that is not sharp, Burge writes: “The striking element in Frege’s view is his application of this distinction to cases where the most competent speakers, and indeed the community taken collectively, could not, even on extended ordinary reflection, articulate the ‘standard senses’ of the terms” (1990, p. 46; Burge’s italics). That was precisely the position of Leibniz and Newton in relation to terms for limits. So the Fregean view, Burge’s account, and the description I am in the course of developing would all firmly distinguish this phenomenon from that of attributions of concepts legitimized by the existence of deference in the use of expressions. Some of the intellectual skills required to succeed in making an implicit conception explicit will be skills useful in any enterprise of building an explanation from instances. Choosing the right classification of cases matters. The right classification of cases is a relatively trivial matter for the logical connectives (or at least, it is so once one has settled on a particular kind of semantic theory). It is somewhat less trivial to articulate the implicit conception involved in understanding the word ‘chair’. It is definitely nontrivial to make
Implicit Conceptions, Understanding, and Rationality
123
explicit what is involved in being the limit of a series. Equally, skill in appreciating the full range of cases matters too, as those failed attempts to define ‘chair’ that omitted skilift chairs showed. So, even though in trying to articulate one’s own implicit conceptions, one is trying to articulate what is influencing one in making judgments involving the concept in particular cases, the skills and methodology involved are those pertinent to any abductive investigation. Achieving such an articulation is not simply a matter of passively allowing the content of some implicit conception to float into consciousness from the subpersonal level. Since it can be hard to make explicit the content of one’s own implicit conceptions, we should equally not be surprised if thinkers sometimes mischaracterize the content of their implicit conceptions. A thinker’s explicit endorsement of an incorrect definition does not mean that he does not have an implicit conception whose content is the correct definition. The attribution of a content to an implicit conception is answerable to its role in explaining the thinker’s ordinary applications of the concept in question. Examples are primary in the attribution of content to the implicit conception. Thinkers can be good at classifying cases, and bad at articulating the principles guiding their classifications. Ordinary thinkers, who understand the predicate ‘chair’ perfectly well, often give an incorrect definition when pressed for one. And if Leibniz and Newton can mischaracterize their own grasp of a concept, how can the rest of us expect never to be in error on such matters? How wide is the range of concepts and expressions with which implicit conceptions are associated? The examples of implicit conceptions I have offered so far have been associated with logical and mathematical concepts, and have involved definitions. Implicit conceptions involving definition may, though, be found in almost any domain. A significant segment of moral and political thought, for example, consists in making explicit the implicit conceptions and constraints that guide our applications of such notions as fairness, equality, and opportunity. At the other end of the spectrum, we need to employ implicit conceptions in characterizing the mastery even of some observational concepts. In mastering the concept cube, taken as an observational concept, a thinker must have an implicit conception with roughly this content: that cubes are closed figures formed from square sides joined at right angles along their edges. Not all examples will be so trivial. In the case of any philosophically interesting concept, the question of the content of the implicit conception underlying it will be highly substantive. Answering the question will in such cases involve making some substantive advance in the subject matter in question. The benefits of successfully making explicit the content of some previously merely implicit conception are multiple and various. Since having a merely implicit conception is fundamentally tied to judgments about particular examples, the first benefit of an explicit statement is that of generality. Leibniz and Newton had no difficulty giving the limits of particular series of ratios. What they did not knowledgeably formulate was the general,
124
Christopher Peacocke
universally quantified biconditional stating the relation in which a number had to stand to a series to be its limit. The generality brings much in its wake. In particular, it provides a crucial tool needed to prove general theorems about limits. A second benefit of making the conception explicit, one for which the generality also matters, is the possibility of fully defending the legitimacy of the notion. Only with a general, explicit statement of what it is to be the limit of a series is a theorist in a position to give a fully satisfactory answer to Berkeley’s critique of the notion. A third benefit is one that Frege notes that proofs can also bring: correct definition can help to establish “the limits to the validity of a proposition” (Foundations of Arithmetic, §1). In general, proof and definition will do this handin-hand. Proofs usually require some definition of the notion in question. Equally, the fruitfulness of the definition can be established only by investigating what can be proved form it. A fourth benefit, like the second, also has to do with justification. Someone who knows the explicit characterization can give a rationale for his classification of particular examples. This applies both in mathematical and logical cases, and in the moral and political examples. Any general constraints on fairness, for instance, which we can discover and formulate with generality will allow us to argue much more forcibly that some particular procedure or arrangement is, or is not, unfair.3 If a thinker has an implicit conception, there will be a certain psychological relation in which he stands to a content that specifies the content of that conception. The nature of that psychological relation is something I will be discussing imminently. I do, though, want to distinguish sharply between this relation that is under investigation, and at least one familiar notion of tacit or virtual belief. The notion of tacit belief is most trivially illustrated by such examples as an ordinary person’s belief that cars are not edible, and perhaps less trivially by an ordinary person’s beliefs about an interlocutor with whom he is engaged in a conversation—his rationality, or perhaps some of his higher-order attitudes. Mark Crimmins seems to me to have made a good case that these examples of tacit belief can be elucidated as ones in which for a person to at least tacitly believe that p is for it to be as if the person has an explicit belief in p (1992, p. 248). The content of a tacit belief in Crimmins’s sense is not news to the tacit believer; whereas the Bolzano-CauchyWeierstrass definition was certainly news. (The point applies even to the modest case of the correct definition of ‘chair’.) It follows that the sense of ‘limit’ as used by Leibniz and Newton—its contribution to cognitive value—cannot be identified with the correct explicit definition of ‘limit’. Burge has made the same point forcefully for a different range of examples (1986, pp. 745–747). Note that the distinction between cognitive value and correct explicit definition applies both in cases that have no externalist character, such as the case of limits, and in cases like ‘chair’, which do. In both kinds of case, it is plausibly the close tie between ordinary employment of the sense and the ability correctly to classify examples that brings with it
Implicit Conceptions, Understanding, and Rationality
125
the distinction between the ordinarily used sense and the more theoretical explicit definition. The close tie with particular examples can be present both in cases where there is external individuation of the concept, and in cases where there is not. Maybe some substantial restriction on the range of phenomena considered in verifying the ‘as if’ clause in Crimmins’s characterization would capture tacit conceptions as a special case of a generic notion of virtual belief.4 The natural restriction would cut the range of phenomena down to certain canonical applications of the concept for which an implicit conception is being given. In the case of the limit example, we might be restricted to considering the thinker’s ability to calculate the limits of particular series. My point at present, however, is that some such substantial restriction is required. There may be a spectrum of tacit and virtual beliefs here, but implicit conceptions are not at the same point along it as many more familiar examples of tacit belief.5 2
Deflationary Readings Rejected
What I have said so far can be greeted with varying degrees and kinds of skepticism. One important deflationary reaction is the complaint that the implicit conceptions of which I have spoken are simply projected backwards from the actual inferential dispositions of thinkers. The complaint would run thus: insofar as it is legitimate to speak of implicit conceptions at all, they serve simply to summarize the actual inferential propensities of those who understand the expressions in question. But, the complaint continues, the implicit conceptions neither explain nor justify anything. What constitutes understanding are the particular inferential dispositions, rather than anything that underlies them. I offer two points in reply. The first is that a person’s understanding of an expression may outrun natural generalizations of all the principles he has ever encountered, or could be expected to come up with. A natural illustration of the point is provided by nonstandard models of first-order arithmetic, which contain blocks of ‘nonnatural’ numbers that follow after all the genuine natural numbers. It seems clear that an ordinary person’s understanding of the expression ‘whole number’ definitely counts nonstandard models as nonstandard. One principle whose truth excludes nonstandard models is the w-rule: in one form, this is the rule that if ‘F(0)’, ‘F(1)’, ‘F(2)’, . . . are all provable in the given system, then so is ‘All natural numbers are F’. Another such principle is a second-order axiom with a quite specific and highly general understanding of the range of the second-order quantifiers. Now ordinary thinkers, who use and understand the expression ‘whole number’, have no conception of any such principles. Nor, for many hundreds of years, did anyone else. All the same, it seems to me that the ordinary thinker’s understanding of the expression ‘natural number,’ and that of everyone more than a century and a half ago, would count the nonstandard models as nonstandard. That these models of first-order
126
Christopher Peacocke
formulations exist would hardly have been striking otherwise. Their designation as nonstandard was not simply a matter of stipulation or convention, or a resolution of some indeterminacy. At this point in the discussion, the deflationist about implicit conceptions may be tempted to appeal to counterfactuals. He may suggest that what matters is that our ordinary thinkers would acknowledge these principles as correct on their understanding of ‘whole number’, were they to be presented with these principles. This seems to me to be a decidedly optimistic view of the person in the street (or many other places) when we imagine that person presented with the w-rule, or with unrestricted second-order induction. But let us waive that. We will waive it by allowing, more specifically, that there may be some non-question-begging restriction R such that if someone has the ordinary concept of a whole number and meets this restriction R, then he would acknowledge such principles as the w-rule, or unrestricted second-order induction, as correct. The important issue here is: does that help the deflationary reading of implicit conceptions? It seems to me that it dos not. Intuitively, a person’s prior understanding of the predicate ‘is a whole number’ explains why the counterfactual is true of him. When all is working properly, a person who understands the predicate ‘is a whole number’ uses that understanding to work out that the w-rule is correct. The present deflationist is wrongly offering a kind of identification rather than an explanation.6 I would say that the counterfactuals, when they are true of a thinker and properly result from his prior understanding of the predicate ‘is a whole number’, are explained by his possession of a specific implicit conception of the range of that predicate. My own view is that the content of that particular implicit conception should make essential use of primitive recursion with a limiting clause. Its content is given by three primitive principles: (1)
‘is a whole number’ is true of 0;
(2)
‘is a whole number’ is true of the successor of anything it is true of; and
(3) nothing falls under ‘is a whole number’ unless it can be shown to do so on the basis of rules (1) and (2). Clause (3), on its intuitive understanding, excludes the nonstandard models. Note that no explicit use of the notion of finiteness, or second-order properties, or reference to reasoning by arithmetical induction occurs in this statement of the implicit conception.7 These points may just encourage our deflationist further, to say that a thinker’s understanding of ‘is a whole number’ consists in no more than his willingness to accept as correct an explicit statement of this primitive recursion with a limiting clause. But the distinctions of two paragraphs back remain. The implicit conception explains acceptance, when there is rational acceptance based on the thinker’s own understanding. This defla-
Implicit Conceptions, Understanding, and Rationality
127
tionist would also, of course, have to grapple with the problem of the willingness of some thinkers to accept incorrect explications of particular concepts. This first deflationist view I have been considering may seem like a no-nonsense position, opposed to mysterious views of understanding that transcend the knowable. But in fact nothing in what I have said should encourage the view that implicit conceptions somehow transcend the knowable. There would be a commitment to such transcendence if it were allowed as a possibility that there could be two thinkers whose rational judgments about particular applications of an expression, and about principles involving it, are in actual and counterfactual circumstances identical, and who yet have differing implicit conceptions. Nothing I have said entails that that is a possibility. I have, on the contrary, been emphasizing the role of implicit conceptions in the explanation of particular judgments involving the expression or concept. The upshot is, then, that insofar as we see the rejection of this sort of transcendence as desirable, its rejection is not unique to the deflationist. Rejection of transcendence cannot be used in support of the deflationist’s view. So far I have been concentrating on points about explanation; but I also promised a second point in reply to the deflationist’s objection that what I say about implicit conceptions is no more than a summary of truths about inferential dispositions. The second point emerges from the question: How is the deflationist to specify the inferential dispositions of which he says that implicit conceptions are not more than a summary? The second point starts from the fact not any old inferential disposition can be included. Ordinary logical inferences are rational transitions. They are not blind leaps into the dark, inclinations to make transitions in thought that just grip and take over the thinker’s rational self. (On this, I am in total agreement with Brewer 1995.) I tried at the start of this essay to say something about how a thinker’s implicit conception can make rational acceptance of even a primitive axiom or inference rule. The phenomenon I highlighted was not merely that of our learner of logic unable to see how a primitive logical law might fail to hold in the actual world. It is rather that he has a quite specific positive means of rationally reaching the view that the particular law in question will always be true. How might our deflationist try to account for the rationality of accepting primitive axioms or inference rules? He may just say that the rationality of acceptance is explained by the fact that these axioms and inference rules are evidently correct for the truth-functions, or higher-level functions, expressed by logical vocabulary. They are indeed evidently correct; but the point cannot serve the deflationist’s purpose, again for two reasons. One reason is that the deflationist had better say why these are the correct truth-functions and higher-level operations to associate with the logical expressions. Those specific semantic assignments are hardly given in advance, and what makes them the correct assignments must have something to do with what is involved in understanding these expressions. The theorist of implicit conceptions will insist that they are the correct assignments because they capture precisely
128
Christopher Peacocke
the contribution of the expression to truth-conditions given in the content of the implicit conception associated with the expression by one who understands it. The other reason the deflationist’s purposes are not served by this response is that the rationality of the acceptance of a logical principle must also somehow connect up with the truth-function, or higher-order function, that is the semantic value of the expression with the thinker’s own understanding. Saying the principles are correct for a certain semantic value does not explain the rationality of accepting the principles unless we make this semantic value something the thinker knows about. How is this connection with the thinker’s knowledge to be effected on the deflationist’s view? It cannot always be a matter of explicit knowledge of the semantic value. As we noted, the thinker who comes rationally to accept a logical principle does not always have such explicit knowledge. Moreover, such explicit knowledge seems obtainable by rational reflection on the part of one who understands the expression. If the deflationist tries, at this point, to retreat to the position that the thinker has implicit knowledge of the semantic value, he would thereby be embracing implicit conceptions after all. The deflationist might respond by taking a different route. He may say that the semantic value of a logical constant is simply fixed as that which makes truth-preserving the axioms and principles the thinker is willing, in some specially primitive way, to accept. This was the line I myself took in some of my earlier work (1987, 1992). It involves what is sometimes called a form of thinker-dependence. On the view proposed, what makes an axiom or principle correct is, as a constitutive matter, dependent on whether or not thinkers actually accept it (in some designated, specially primitive way). This is sometimes advertised as a virtue of the view. I think, however, that it makes it impossible to give a satisfactory account of the rationality, the nonblind acceptance of logical principles and axioms. The rationality of accepting some proposed axiom or principle containing already understood expressions involves aiming at correctness that is, as a constitutive matter, explained independently of acceptance of that particular principle. That sort of independence must be an illusion on a judgment-dependent view of these matters. Alternatively, a thinker-dependent view may mention not judgment, but rather how the principle strikes the thinker. How the principle strikes the thinker is quite properly to be distinguished from judgment, for a thinker’s judgment may either endorse or overrule how it strikes him. But we still conceive of validity as something that is equally neither constituted nor guaranteed by conditions involving how the principles strike the thinker. A proposed new logical principle may strike a thinker as correct. But he is not entitled to accept it until he has engaged in rational reflection on it, reflection of the sort we have been discussing. One of the points that distinguishes the logical case from that of color is that it is not plausible that, before a thinker makes a color predication of a perceived object, further rational reflection is required, of a thinker who experiences something as a shade
Implicit Conceptions, Understanding, and Rationality
129
of a certain color, and who has no reason to doubt that environmental and his own perceptual mechanisms are favorable. There is then no blanket objection in what I am saying that would apply to any thinker-dependent treatment of any concept whatsoever. My point is only that we have a conception of validity, and correspondingly of what is required for rational acceptance of logical principles, which makes thinker-dependent treatments of the validity of ordinary (nonmetalinguistic) principles inappropriate. A second, more persistent deflationary objector may still press his case. He may say: “Everything you explain by appeal to implicit conceptions can be explained by use of inferential dispositions run off-line. For instance, the lines of the truth-table for ‘or’ can be reached as follows. Our new student of logic treats any sentence A of English as interinferable with ‘A is true’, or, as we may say, he has the disquotational inference for truth. We can consider the thinker’s disposition to infer either ‘A or B’, or its negation, from each of the sets of premises {A, B}, {A, ~B}, {~A, B}, {~A, ~B}. These inferential dispositions, when exercised off-line, and employed in conjunction with mastery of the disquotational inference, allow him to attain each line of the classical truth table for alternation. From this he can also infer the validity of the schema A Æ (A or B). So we can explain all the phenomena without any appeal to implicit conceptions.”8 I reply that an inference such as that from the premises {~A, ~B} to ~(A or B), is— though no doubt automated for even elementary logicians—one that our student of logic has to work out to be correct on the basis of his existing understanding of alternation. It seems to me that this working-out must involve use of the concept of truth. It must involve reasoning tantamount to: “The premises imply that neither A nor B is true. ‘A or B’ is true, though, only if at least one of A and B is true; so when these premises hold, ‘A or B’ won’t be true, that is ‘~(A or B)’ will be true.” If this is right, then, even for students who do have the disposition to make the inference from the premises {~A, ~B} to ~(A or B), that disposition cannot be part of the explanation of his knowledge of the truth table for alternation. On the contrary, appreciation of the principles that fix the truth-value of an alternation is part of the rational explanation of the student’s appreciation of the validity of the transition. I have made the point with a more complicated inference, for the point is perhaps more vivid there. In fact, I suspect it applies equally to the rational acceptance of the general schema of alternation-introduction. In response to this, our second deflationary objector may shift his position slightly. He may say that it suffices for his purposes to consider a conceptual role mentioning metalinguistic transitions involving predications of truth and falsity themselves. It is metalinguistic inferential dispositions that are run offline, he may say, and that generate the truth table for ‘or’. Given the metalinguistic premises that A is true and B is false, for instance, the thinker will immediately be willing to infer that ‘A or B’ is true. I have some incidental doubts about this strategy for other lines of the truth table. I suspect that a
130
Christopher Peacocke
transition from the falsity of A and the falsity of B takes a bit more thinking about for an eighteen-year-old than for us experienced (elementary) logicians. I suspect he has to reason that if A and B are both false, then neither A nor B is true, that is, neither of the conditions at least one of whose truth is required for the truth of ‘A or B’ holds. But let us waive the incidental doubts. After all, I agree that the corresponding metalinguistic transitions are immediately compelling in the case of conjunction. So what do I say about this second variant of the deflationary objection? I say that, in moving to the metalinguistic level, it does not present a competitor to the theory of implicit conceptions. Finding such a metalinguistic transition as is cited in this objection to be a compelling transition is a manifestation of an implicit conception with the content that any sentence of the form ‘A or B’ is true iff either A is true or B is true. Our objector may protest, “Well that’s a spurious explanation: the alleged explanans is simply summarizing what needs to be explained.” But I dispute the objector’s claim that the attribution of an implicit conception simply summarizes the dispositions to be explained. The explanation makes quite specific commitments. One of these commitments is that what explains the transition is its having a certain form—rather than, say, the Gödelnumbers of its components standing in a certain relation, which is equally something that might be computed. This last issue is equally one that arises about the implicit conception underlying understanding of the predicate ‘chair’, and reflection on that case may help to make this part of the reply to the second objector more plausible. Implicit knowledge of the definition of ‘chair’ can explain a person’s applying the word correctly in central cases. To say that a person has a disposition to correct application in central cases is not by itself yet to specify which features of chairs in his environment are operative in leading him to apply the term. Saying that the thinker’s performance is explained by a specific implicit conception commits one to saying that his performance involved the identification of backs, seats, and the rest—the features mentioned in the content of the implicit conception involved in his understanding. I myself am very sceptical that there is one set of inference schemata acceptance of instances of which is absolutely constitutive of understanding classical alternation. Some thinkers are better at inferring to alternations, some are better at making inferences from them, some may have a better grasp of the way alternations interact with conditionals, others may find their interactions with negation easier. They may all nevertheless have the same core understanding of alternation and the same implicit conception of its contribution to truth conditions. In other parts of the philosophy of mind and language, we have become quite comfortable with the idea that some states are not definitionally tied to one kind of manifestation, but produce their effects only in combination with several other factors. A
Implicit Conceptions, Understanding, and Rationality
131
perceptual state’s having a particular spatial content is one such example. Such spatial content may explain all sorts of actions, in combination with other attitudes, abilities, and enabling conditions. It is, though, quite implausible that there is some privileged possible kind of explanandum that is canonical in legitimating that attribution of a spatial content. Seeing something as at a certain distance and direction from oneself may, in the presence of other attitudes, produce action directed at that position. But it may, as in the case of the prisoner in The Count of Monte Cristo, equally produce a certain sequence of winks of an eyelid as a message in code; or it may just result in the updating of some mental map on the part of someone incapable of movement at all. Another example of a psychological state not individuatively tied to just one kind of explanandum is that of a psychologically real grammatical rule. Its psychological reality may explain features of a person’s perception of heard utterances, or of his own productions; and the same rule may be real both for one thinker who can understand but not produce, and for another who can produce but not understand. I suggest that this feature, of having explanatory power that is not canonically or definitionally tied to one privileged kind of manifestation, is present also in the state of understanding logical expressions, and in having an implicit conception with a semantic content. There is yet a third deflationary critic to be considered, one who takes a rather different tack. He will say that we have no need of implicit conceptions. He will say that it suffices, in attaining the correct interpretation, to note that we maximize intelligibility of Newton and Leibniz, for instance, if we attribute to them the concept of the limit of a series. Now the description ‘maximizing intelligibility’ is a term of art, but on any natural reading, I doubt whether the reasons offered by this third critic are really incompatible with the existence of implicit conceptions. It cannot be a cosmic coincidence that interpreting Newton and Leibniz as having the concept of the limit of a series counts them as getting the answers to questions about series and gradients right. Interpretations must be counterfactually projectible, or they would be no use in either the explanation or the prediction of thought and action. If the interpretation of an expression that maximizes intelligibility is said to have no implications or commitments for the psychological explanation of why the expression is applied to the cases it is, the charge of cosmic coincidence would, it seems to me, be just. Indeed, I would make the charge even in the humble case of the concept chair. If someone is said to be interpretable as meaning chair by an expression, and gets its application correct, but is said not to have any tacit knowledge of its definition, then the charge of unexplained coincidence would stick against that view too. The coincidence in question is that of his applying the expression to all and only things that fall under the definition (independently certifiable illusions aside). Extending the coincidence to counterfactual circumstances would only increase the mystery. If, by contrast,
132
Christopher Peacocke
the definition, either of ‘chair’ or of ‘limit’, is regarded as the content of an implicit conception that contributes to the psychological explanation of why the expression is applied to the cases it is, there is no coincidence at all. The astute theorist who says that correct interpretation is to be elucidated in terms of maximizing intelligibility would do better to say the following. When we think through the consequences of maximizing intelligibility, we are forced by the need not to postulate massive cosmic coincidences—one indeed for each thinker and each such concept—to recognize the existence of implicit conceptions. This more astute position is then of course not in conflict with what I have been advocating. It is reaching some of the same conclusions by a (possibly) different route. It is not always the case that late theory simply articulates a concept that at an earlier time was not fully understood by its users. Sometimes later theoretical developments are refinements, precisifications that resolve earlier indeterminacies. This can happen as much in the physical and other empirical sciences as in the mathematical. Whether an example is one of articulation of a conception that was earlier merely implicit, or is rather one of refinement, has to be examined case by case and is often a complex and intriguing matter. It would be a brave soul who claims that we have a unique pretheoretical notion of set. Though the matter needs much argument, it would equally be a brave soul who denies that there was a determinate notion of whole number prior to the theoretical developments of the past hundred years. The theorist of implicit conceptions needs only the recognition that not all cases of theoretical development are resolutions of indeterminacies. I turn now to some further ramifications of the point. 3
Conceptual-Role Theories and the Phenomenon of New Principles
If we accept the existence of implicit conceptions, what are the consequences for conceptual-role theories of meaning? Conceptual-role theories were proposed by Sellars (1974) and were developed in one variety or another by Harman (1972), Block (1986), and Field (1977). It is consistent with the existence of implicit conceptions that in at least some cases, some part of the conceptual role of an expression or concept contributes to making the expression have the meaning it does, or contributes to the identity of the concept. In fact it is arguably not merely consistent, but on reflection a consequence of the position I have been outlining. For a good case could be made that it should be written into the principles for ascribing content to implicit conceptions, such as those associated with the concepts limit or chair, that they explain their possessor’s application of the corresponding concept in the circumstances in which it is correct to do so—and this is precisely for these expressions or concepts to have a certain conceptual role.
Implicit Conceptions, Understanding, and Rationality
133
This a priori connection between the content of the implicit conception and certain of its consequences should not be taken to mean that the implicit conception cannot be genuinely explanatory. The idea that certain states are individuated in ways that connect them a priori with what they are capable of explaining is one we have, quite properly, happily lived with in the philosophy of mind and psychology for many years now. The claim that a thinker’s practice with the concepts chair or limit is explained by his having a certain implicit conception is also one with quite specific import and other explanatory consequences too. In saying than an implicit conception with a certain content explains the practice, we are committing ourselves, for instance in the case of the concept chair, to the explanation of particular judgments implicating the thinker’s ability to distinguish seats, backs, the relation of support, and something with a certain function for human beings. If the case for the existence of implicit conceptions were confined to examples like that of chair and limit, they would force only a minor revision of conceptual-role theories of meaning and content. Examples like those would be accommodated by a small addition to conceptual-role theories. The addition would consist in saying that the thinker’s particular applications of the concept or expression in accordance with the specified conceptual role must have an underlying content-involving computational explanation. This addition the conceptual-role theorist could take in his stride; it does not seem to take back anything in his position. The general phenomenon that seems to me to preclude acceptance of conceptual-role theories—even those that are semantically constrained in one way or another—is rather that of the rational, justified acceptance of new principles involving a given concept, where these new principles do not follow from those principles (if any) immediate acceptance of which is required for possession of the concept. I label this the Phenomenon of New Principles. I am inclined to think that the Phenomenon of New Principles is as decisive an argument against pure conceptual-role theories as the phenomenon of understanding sentences one has never encountered before is decisive against theories of meaning that do not proceed compositionally. Let us consider yet another example. What might a conceptual-role theorist offer as the meaning-determining role for classical negation? He might include the conditions for assertion of the negations of observational sentences. He would need to do more, because negation must be fixed for all contents to which it can be applied, whether observational or not. At this point, the conceptual-role theorist is likely to be tempted to reach for and include the classical logical inferential principles for negation: that from ~~A one can infer A, and that if one can derive a contradiction from A, one can infer ~A. Yet again, it seems clear that these classical logical rules for negation are ones whose correctness can be, and needs to be, worked out from some philosophically prior understanding of negation. The prior understanding is simply possession of the implicit conception that a sentence prefixed with ‘it is not the case that’ is true just in case
134
Christopher Peacocke
the sentence is not true. In fact in this particular case, it seems to me that possession of this implicit conception does not involve drawing on anything new that was not involved in the understanding of sentences not containing negation. To understand the sentences not containing negation, the thinker must know their truth-conditions; and that is, ipso facto, to know their falsity-conditions. As Geach (1972) once emphasized, to know the truth-conditions of a sentence is in effect to know the location of the boundary between the cases in which it is true and those in which it is not. There is no such thing as knowing the location of this boundary without possessing knowledge of the falsity-conditions of the sentences. The implicit conception associated with the understanding of negation simply links the expression for negation with these already appreciated falsity-conditions. That this is the subject’s implicit conception may be manifested in all sorts of different ways. Conceptual-role theorists have not wholly neglected this sort of objection. The sorts of moves they have made to attempt to accommodate it, though, do not seem to me fully to resolve the problem. One move that suggests itself, and which I made in earlier work (1987), is for the theorist to say that a new principle, whose correctness can be rationally appreciated, is fixed by those that are mentioned in the conceptual role in some less direct way than by following by some consequence relation. In the earlier work, I spoke of the new principle as being determined as one made correct by, for instance, the strongest semantical assignment that validated some introduction rule mentioned in the conceptual role. In this way, for instance, one can explain why the natural deduction rule of orelimination is correct, even though it is not found immediately obvious by all those who understand ‘or’. Corresponding moves can be made for elimination rules too. The difficulty with this strategy is that even if it does succeed in fixing the right set of new principles as correct, it leaves at least three problems unresolved. The first problem is that the resources it employs give no plausible account of the ordinary thinker, like our new learner of logic, who works out the correctness of a new principle that does not follow from (say) the logical principles he already accepts for a given constant. When you worked out that or-elimination is a valid rule, you did not employ any premises, or tacit simulation that committed you to the proposition that the semantical value of a constant is the strongest that validates an introduction rule, or the weakest that validates an elimination rule. You had no such thought or commitment. If we are going to explain the rationality of acceptance of a new principle, we must appeal to something that is plausibly operative with a thinker engaged in rationally accepting it. The second problem is that in some cases, all of the inference rules distinctive of a concept have to be worked out by a thinker—we noted that this was plausibly the case for the natural-deduction rules of negation-introduction and negation-elimination. So in some cases, this strategy does not have the initial materials on which it needs to operate.
Implicit Conceptions, Understanding, and Rationality
135
The third problem with the strategy of appealing to the strongest semantical assignment that validates an introduction rule is more fundamental. It is that the strategy gives no rationale for this requirement itself. I do not think it can be founded in considerations of tightness of ascription of contents and semantical values. Suppose, for the sake of giving the view the best chance, we grant that if someone is using an introduction rule correctly, and that if the logical expression is meaningful, there must be some semantical assignment that validates it. It still does not follow that we must take as the semantic value the strongest such assignment. On the contrary, if we are appealing to considerations of tightness, with only that data, we should consider as semantical assignments only the whole class of those assignments that make valid the introduction rule, rather than the strongest. To select the strongest is actually to go beyond what is justified by the inferential practice. I conclude, then, that once we acknowledge the full range of phenomena explained by implicit conceptions, including the Phenomenon of New Principles, conceptual roles can neither fully determine meanings nor fully individuate concepts. For someone who occupies the Fregean standpoint and regards the examples as evidence of incompletely grasped—but nevertheless determinate—senses, none of this should be surprising. The Phenomenon of New Principles is only to be expected from that standpoint. The new principles that are rationally accepted reflect those aspects of the determinate sense, which is already employed in thought, but whose nature needs theoretical thought on the part of its ordinary users if it is to become “sharply grasped.” In criticizing conceptual role theories as constitutive theories of understanding and concept-possession, I have not committed myself to the view that meaning can go beyond the full range of correct conceptual roles for an expression. Equally, a realist about theoretically postulated magnitudes in a physical science should not assert that truths about them go beyond everything determined by possible observational consequences, when we are considering the full range of possible experimental setups. But it would also be almost universally agreed that acceptance of this last point does not mean that statements about the theoretically postulated magnitudes can be reduced to those about possible observations. Something analogous seems to me to be true of meaning and concept-possession, and their relation to conceptual roles. Indeed the very notion of a correct conceptual role is precisely one that I have been claiming the conceptual-role theorist cannot fully elucidate. In some cases, what is correct can be explained only by appeal to an underlying implicit conception. I also add a remark for enthusiasts who have followed the debates about conceptual role theories of meaning and concepts. For some years now I have argued that not every coherent conceptual role determines a meaning or concept. Only those roles that naturally correspond to a certain contribution to truth-conditions do so (1993).9 If not every conceptual
136
Christopher Peacocke
role determines a meaning, it would hardly be surprising if the specific contribution made to truth- or satisfaction-conditions features in the content of an implicit conception. For conceptual roles for which there is no corresponding contribution to truth- or satisfactionconditions, there is no content available to be the content of any corresponding implicit conception. Conceptual roles that correspond to no contribution to truth-conditions are, under the conception I am advocating, automatically excluded. 4
Explanation by Implicit Conceptions
Explanation by implicit conceptions raises a host of queries and doubts. There are doubts about the particular kind of psychological explanation in which they are said to be implicated. There are also more general philosophical doubts about whether implicit conceptions can ever properly be involved in a description of what is involved in possessing a concept or understanding an expression. Let us take first the issue of what kind of psychological explanation an explanation that appeals to implicit conceptions might be. An explanation by an implicit conception is a species of explanation by a contentinvolving state, the content being the content of the implicit conception. So the usual features of content-involving explanation apply. An explanation of a judgment involving a particular concept by citing the person’s implicit conception is not an explanation of a syntactic state by a syntactic state, not even if both implicit conception and judgment are realized in subpersonal syntactic states. An implicit conception contributes to the explanation of a judgment under its content-involving description as a judgment that p, for some particular p. The implicit conception does not explain the judgment under a merely syntactic description. Nor could we regard the explanation as one that is covered by a prima facie law relating some syntactic realization of the implicit conception to the occurrence of content-involving judgments. Explanations by content-involving conceptions can be the same across persons who realize contents in different subpersonal systems of representation, different mental “notations.” It is also not at all clear that a “syntax to content” prima facie law would be adequate to explain the knowledgeable status of the resulting judgments. The model, then, to illustrate it for the simple case of chair, would run thus. One of the thinker’s perceptual systems, say, identifies some object in the environment as having a supporting area and a back, and the subject has the background information that the object is used for sitting on. This information from the perceptual system, together with the background information, is combined, at a subpersonal level, with the content of the implicit conception underlying understanding of ‘chair’, and by some form of subpersonal computation it is computed, from the given contents, that the presented object is a chair. This in turn explains the thinker’s willingness to judge that that object, demonstratively given
Implicit Conceptions, Understanding, and Rationality
137
in perception, is a chair. In the case of other concepts, the role just played by the perceptual system will be played by some informational source or other. This source yields a content that, together with the content of the implicit conception underlying the concept, and possibly some background information or presupposition, permits computation of a content to the effect that some given object falls under the concept in question. Of course, in both perceptual and nonperceptual cases, we can expect all sorts of short-cuts to be used in reaching particular judgments. The full content of the implicit conception need not be online in every classification the thinker makes. All the implicit conceptions I have considered so far have contents that are correct. It is not impossible for there to be an implicit conception with an incorrect content. A thinker may misunderstand some word in the public language. False presuppositions about certain kinds of objects or events in his environment may also enter the content of his implicit conceptions. Nonetheless, there is a core of cases in which one can expect that the content of the implicit conceptions within that core will be correct. It is very plausible, on grounds having to do with the theory of interpretation and content, that there will be a core of cases in which a thinker will make judgments correctly, and will do so also in a range of counterfactual circumstances. If we accept any theory of content or interpretation on which that is so, then we can expect that any implicit conceptions explaining the applications of the concepts in those judgments will also be largely correct. If the implicit conceptions were not largely correct, the judgments would not be largely correct either. Having squarely accepted that explanation by implicit conceptions is content-involving explanation, there is still the question of whether the content of the implicit conception is, at the level of subpersonal mental representations, implicitly or explicitly represented. As with other kinds of tacit informational state, what has here been deemed important to an implicit conception is prima facie compatible both with explicit and with implicit representation, at the subpersonal level, of the content of an implicit conception (see Davies 1989). In the example of the limit of a series, the informational content might be explicitly formulated in a language of thought. There would be some stored formula that states the definition of a limit. But the content of the implicit conception could equally be grounded in the operation of a processor that does not involve, at the subpersonal level, explicit representation of the content of the implicit conception. We can certainly conceive of a processor that takes information about the numerical values approached by a function at a given point, and delivers as output information about the differential at that point. It must be at least partly an empirical question which kind of representation is operative in a given thinker. Whichever way the issue is resolved for a given subject and implicit conception, I would like to note one constraint that a fuller theory ought to satisfy. Thinkers can know that certain general principles hold for some concept for which they have only an implicit
138
Christopher Peacocke
conception. Even though a rigorous justification for these principles would need to draw on an explicit statement of that conception, it seems that these principles are known even though the conception is not explicitly known. An example is the multiplication principle that (dx/dy)·(dy/dz) = dx/dz. I think Leibniz and Newton knew this general multiplication principle. Again, they did not learn it by being told it by someone else. Though they did not know any adequate explicit definition of a limit, they had sufficient insight into what it must be to realize that this principle is correct. Any theory that characterizes their implicit knowledge as simply serving up the value for the differential of a particular function, and then claiming that such general principles as the one just mentioned are extracted inductively would be inadequate to the phenomena. Indeed, one does not have to be a Leibniz or a Newton to appreciate that the rate of change of one magnitude with respect to a third magnitude is identical with the rate of change of the first with respect to a second magnitude, multiplied by the rate of change of the second with respect to the third. Perhaps the correct description of the situation is that though they had only an implicit conception of the definition, they did know that limits are instantaneous rates of change of one magnitude (as one loosely writes) with respect to another; and they knew that relative rates of change respect that multiplication principle. This phenomenon, of knowledge of some general principles involving a concept in the absence of knowledge of any explicit definition, is found outside mathematical and logical cases. It applies in cases from the more interesting moral and political examples, right up to the humble case of the definition of ‘chair’. While we are on the topic of the nature of explanation by implicit conceptions, it may be helpful if I locate the position I have outlined in relation to the well-known theory of the psychology of inference expounded by Johnson-Laird in his book Mental Models (1983). Evidently I am committed to agreement with him on two of the distinctive claims of his approach. Like him, I have held that the validity of logical principles has to be worked out by thinkers on the basis of their prior understanding of the expressions they contain. I am also in agreement with him that this prior understanding takes the form of knowledge of contribution to truth-conditions. Thus Johnson-Laird writes: “What children learn first are the truth conditions of expressions: they learn the contributions of connectives, quantifiers and other such terms to these truth conditions. And, until the have acquired this knowledge about their language, they are in no position to make verbal inferences” (1983, p. 144). Some aspects of his theory of mental models could be integrated further with the position I have outlined. However, I do part company with any claim that there is no ‘mental logic’, that no form of mental reasoning is needed to explain explicit logical inference.10 There is, on the view I have put forward, inference at one and possibly two levels. First let us recall the example in which we envisaged the subject as working out, via a simulation procedure, at the personal level, the validity of some simple truth-
Implicit Conceptions, Understanding, and Rationality
139
functional principle. There the thinker had to use logical reasoning, for instance in drawing conclusions from the premise that these were all the truth-values that could be taken by the atomic components. Second, at the subpersonal level, in the explanation envisaged a few paragraphs back of a judgment ‘That’s a chair’, for instance, some form of subpersonal inference is essential. It was employed in moving from the information that the presented object has certain properties together with the content of the implicit conception to the conclusion that the presented object is a chair. Perhaps Johnson-Laird would say, as I think in consistency he should say, that the mental models should be used at that subpersonal level too. But it does become a real question then whether the procedures for constructing and operating on mental models should not be regarded as just the way the system is, subpersonally, encoding various inferential principles. It is true that the inferential principles need not be explicitly represented in a language of thought. (Perhaps that is all Johnson-Laird really wanted to claim about the subpersonal level, in which case our positions would not diverge.) But I noted only a few paragraphs back that absence of explicit representation at the subpersonal level does not mean that there are no psychologically real states that contain the content of those principles. The theory of implicit conceptions I have started to outline is committed to holding that there are some such psychologically real states whose content is that of the implicit conceptions. I turn now to two principled objections to the enterprise of employing implicit conceptions in explaining understanding. They could both be considered at some length, but here I will just try to indicate the lines of a response to them. The first set of concerns revolves around the ‘A(C) form’, the noncircularity constraint of A Study of Concepts. The other set starts from the views of the late Wittgenstein about meaning and understanding. What I have said about implicit conceptions is incompatible with adoption of the A(C) form of A Study of Concepts and involves abandonment of that constraint on the philosophical explication of concept-possession. The A(C) requirement on explicating possession of a given concept F was that the concept not feature in the explication, as the concept F, within the scope of attitudes attributed to the thinker. Implicit conceptions of the sort I have advocated violate this principle. I have been advocating implicit conceptions with such contents as ‘Any sentence of the form ‘not-A’ is true iff A is not true’, and ‘Any sentence of the form ‘A v B’ is true iff either A is true or B is true’. Here the occurrences of ‘not’ and ‘or’ on the right-hand side of these biconditionals violate the A(C) restriction when implicit conceptions with these contents are offered as explications of possession of the concepts of negation and alternation. There are various ways in which one might try to qualify the A(C) form to avoid an incompatibility, but I can only report that I have not been able to find any that are well motivated and also cover the ground. Violations of the A(C) form are unobjectionable in the explication of a concept F because one can use one’s own mastery of the concept F to assess what someone with an
140
Christopher Peacocke
implicit conception involving F could be expected to think or do in any given state of information. This is why a statement about what is involved in possession of a concept, and which does not respect the A(C) form, is not vacuous. It still makes an assessable claim. Each one of us, in evaluating the claim it makes, draws on his own mastery of the concept F being explicated. One draws on that mastery, and engages in simulations to assess what one would be obliged, or rational, to think or do in any given state of information. With information from these simulations, one is then in a position to assess the claim about possession of the concept in question. It is in just this way that one can evaluate the various claims I have made in this paper about the content of the implicit conceptions underlying various particular concepts. Drawing on one’s own mastery and using simulations in this way is sharply to be distinguished from making assessments by inference from any theoretical beliefs one may have about the conditions for possession of the concept F. Though of course if one uses the simulations, and draws on one’s own mastery of the concept, one will eventually end up with some such theoretical beliefs, the route by which they are attained essentially involves simulation. We could of course equally proceed this way in assessing what sentences someone would, in various circumstances, be likely to accept on the simple hypothesis that by ‘chair’ she means chair. We did, though, give specific motivations in the case of the logical constants for going beyond the disquotational form, and actually introduce semantic notions into the content of the implicit conception. Equally in the case of ‘chair’, too, there are facts about a thinker’s relations to seats, backs, and supporting humans in a seated position, which make it important to recognize an implicit conception underlying mastery of the predicate. Perhaps I should add explicitly that offering, for a given concept F, an implicit conception that violates the A(C) form is consistent with the existence of an A(C)conforming conceptual role that individuates F. The case of logical conjunction arguably shows the consistency of this combination. From the standpoint of the present theory, however, this is just a special case from which no general conclusions can be drawn. A defender of the A(C) form may be inclined to ask: Why cannot we proceed as follows? First, using our own understanding of negation, or alternation, or whatever is the target concept in question, we work out the inferential and transitional patterns distinctive of the target concept. These patterns will in general involve other concepts with which the target concept interacts in valid transitions, and they may involve complex principles. Then, this objector continues, we just take this totality of transitional patterns, and say that what is distinctive of the target concept is this: it makes rational that totality. Will this not then respect the A(C) form? I reply, first, that the totality of such transitions is quite open-ended. There is no limit to the valid interactions even of so simple an operator as negation or alternation not only
Implicit Conceptions, Understanding, and Rationality
141
with other logical concepts, but with any other concept, such as that of probability, evidence, arithmetical quantification, observational concepts—or indeed any other new concepts we may introduce. If we ask what unifies this open-ended totality, for instance in the case of negation, it seems to me that there is essentially only one answer: these are all the transitions that we would expect as consequences of possession of an implicit conception with the content that any sentence of the form ‘not-A’ is true iff A is not true. That is the only way of fully capturing the open-ended class of transitions whose rationality is distinctive of negation. If that is so, the way of capturing the totality is incompatible with, rather than supporting, the A(C) form. The second reply is a straightforward application of the earlier parts of this paper. If the target concept makes rational the totality of transitions mentioned by this defender of the A(C) form, there ought to be an answer to the question: How does it make these transitions rational? The theorist of implicit conceptions has an answer to this question, if he follows the approach of this paper. The defender of the A(C) form, it seems to me, does not. I promised also to discuss the relation of implicit conceptions to Wittgensteinian views on understanding and rule-following. The views I have been presenting are clearly incompatible with some parts of his thought. They are incompatible with his thesis that one’s understanding of an expression does not exceed what one can explain (1958, 209–210). The considerations I developed earlier do seem to me to show that some thinkers’ understanding of ‘chair’, ‘limit’, and even ‘natural number’ exceeds what they can in fact explain. The view I have been outlining would also endorse one reading of such a claim as ‘Once you have got hold of the rule, you have the route traced out for you’ (1974, VI, 31). Wittgenstein rejected that claim, though of course he was considering ‘rules’ of a sort available to guide a thinker at the reason-giving level in his intentionally making one application rather than another of the expression in question. Implicit conceptions as understood here are not rules of that sort. Finally, the whole idea of explaining rule-application was anathema to the later Wittgenstein. I am committed to the possibility of contentinvolving subpersonal computational explanations of thinker’s applications of expressions they understand. Wittgenstein’s objections to the possibility of explanation in “bedrock” cases where, Wittgenstein says, the person has nothing that is his reason for going on the way he does, are addressed either to the reason-giving level of explanation, or, on occasion, to physiological explanations. It would be wrong to assimilate contentinvolving subpersonal computational explanation to either of those very different cases. I have not myself found anything in Wittgenstein that can be extrapolated to give a sound argument against the possibility of subpersonal computational explanation. It is equally striking, however, how wide the area of agreement may be between a defender of implicit conceptions and the considerations marshaled in Wittgenstein’s
142
Christopher Peacocke
arguments about rule-following. That one’s reasons for saying that something is the result of adding 2, or is a chair, may in a certain sense eventually give out is entirely compatible with the existence of a content-involving computational explanation of why one applies these expressions in cases one does. The existence of implicit conceptions as understood here is consistent with Wittgenstein’s arguments that rulefollowing in the fundamental cases does not involve consciously consulting anything—as Crispin Wright puts it, there is “no essential inner epistemology of rule-following” (Wright 1989, p. 244). There is even a point of positive agreement, rather than mere consistency, between the present view and Wittgenstein’s. Wittgenstein insists at various points in his argument that the relation between understanding and correct application is not merely contingent.11 The way I have developed the account of implicit conceptions involves a commitment to precisely what Wittgenstein is here insisting on. An implicit conception has as its content a certain condition for falling under the expression it treats. We said that the principles for ascribing content to an implicit conception would ensure that, in certain basic and central applications, an expression associated with that implicit conception would be applied to things satisfying the condition in its content. So indeed the connection between possession of an implicit conception, in cases in which that is the nature of understanding, and correct application is not merely contingent. Hence we have a point of agreement with Wittgenstein. Indeed once the noncontingency is acknowledged, it even becomes possible for what a thinker finds compelling—the way he goes on—to enter the individuation of a concept, consistently with the theory of implicit conceptions. The content of the implicit conception can be fixed in part by the properties of the cases in which the thinker finds it compelling that the concept applies.12 So one should not exaggerate the divergence between implicit conceptions and Wittgenstein on rule-following. It is also fair to add, though, that if one can consistently accept these most recent points about rulefollowing while rejecting Wittgenstein’s claims about the extent and the explanation of understanding, we have to draw a certain conclusion. The conclusion must be that these most recent points, about the phenomenology of rule-following and about the noncontingent relation between understanding and application, offer no support for his position on the extent and the explanation of understanding. 5
Rationalism Supported
What I have said so far about implicit conceptions, together with the use to which I have put them, has the distinctive flavor of a classical rationalist position. Here, for instance, are six principles that can be supported by appeal to implicit conceptions, and which were held by that paradigm rationalist Leibniz.
Implicit Conceptions, Understanding, and Rationality
143
(1) The evidentness of particular axioms is grounded in the understanding of the terms they contain. In the New Essays, Leibniz’s protagonist quotes with approval the views of those philosophers who held that axioms “are evident ex terminis—from the terms—as soon as they are understood. That is, they were satisfied that the ‘force’ of their convincingness is grounded in the understanding of the terms” (1981, p. 406; Book IV, chapter 7, “Of the propositions which are named maxims or axioms”). The description I gave at the start of this paper of the way in which our logic student comes rationally to accept a logical axiom conforms to the description given by the philosophers with whom Leibniz’s protagonist agrees. According to that description, the student’s implicit conception is drawn on in the simulations that lead to rational acceptance of the axiom. In the account given, possession of the appropriate implicit conception was also identified with understanding. So acceptance of the axiom is grounded in understanding. The content of the student’s perceptual experience is justificationally irrelevant to his acceptance of the axiom. (2) Finding an axiom evident, when that is properly grounded in the understanding of the terms therein, is a way of coming to know what it states. The legitimacy of attributing knowledge when acceptance is reached via the understanding was essential to Leibniz’s case against Locke’s empiricism. Leibniz would hardly have had an anti-empiricist account of knowledge of these axioms if this understanding-based evidentness did not amount to knowledge. I noted early on in this essay that the reflections that can lead to rational acceptance of an axiom or inference rule plausibly yield knowledge. The innateness of axioms and inference-rules is not, however, something I am advocating. Chomsky, in his book Cartesian Linguistics, insisted that the rationalists were right in wanting a psychology that is “a kind of Platonism without preexistence” (1966, p. 63). In a similar spirit, I offer implicit conceptions as a rationalist account of understanding and certain kinds of knowledge, but without any commitment to innateness. Implicit conceptions can be acquired.13 In fact I think there are strands in Leibniz that suggest that what really mattered to him was independent of innateness taken literally. At one point in the New Essays, he writes that “quite often a ‘consideration of the nature of things’ is nothing but the knowledge of the nature of our mind and of these innate ideas, and there is no need to look for them outside oneself. Thus I count as innate any truths which need only such ‘consideration’ in order to be verified” (1981, p. 84). The distinctively purely understanding-based “consideration” can be applied whether or not the understanding is, in the literal sense, innate. Leibniz’s saying that he counts as innate any truth that can be attained by a certain kind of consideration suggests that this part, at least, of the
144
Christopher Peacocke
rationalist position may not need to involve literal innateness. The talk of verification in this passage also emphasizes Leibniz’s conception of this sort of consideration as a route to truth and knowledge.14 (3)
Logical axioms can be known a priori.
This was the burden of Leibniz’s dispute with Locke: we can have an entitlement or justification for accepting a logical axiom that is justificationally independent of perceptual experience or sensation, even if experience is an enabling condition for our attaining such an acceptance. The procedure by which, at the start of this essay, I envisaged someone coming rationally to accept a primitive logical axiom is also one that yields a priori knowledge. Nothing in that rationale for the subject’s belief involves perceptual experience or sensation. Of course Leibniz had what in our post-Fregean time we would regard as a very rudimentary conception of logic. The logical laws to which, according to him, all a priori truths could be reduced by means of substituting correct definitions were of such forms as all ‘As are As’, or ‘All ABs are As’, ‘All ABCs are As’, and the like. The “trifling” character of these axioms was a topic of some discussion in Leibniz’s writings. By contrast, someone acquainted with modern logic would be unlikely to characterize all of its axioms as trifling. Yet it seems to me that an understanding-based, nonempirical procedure for attaining knowledge of axioms, even when they are not merely trifling, still lies squarely within the spirit of the rationalists’ conception. Moreover, even on his simpler conception of logic, Leibniz still needs to rely on some of the apparatus I have been employing. Correct definitions, to which Leibniz repeatedly appeals in his characterization of demonstrations, are precisely definitions that correctly articulate the implicit conception involved in understanding the term being defined. To say, as I have, that knowledge of the axioms is grounded in understanding of the expressions in them, and is also a priori, is not at all to endorse the Carnapian view that they are true solely in virtue of the meaning of their constituent expressions. On the contrary, the sort of rationale I envisaged a thinker going through at the start of this essay is one that shows that, for any instance of a logical axiom schema, what it is true in virtue of is its disquoted truth-condition (as indeed would be the case for any other true sentence). I have been trying to develop the present view in a way that respects that point throughout and regards the phrase “true purely in virtue of meaning” as applying to no sentences whatever.15 For this reason, the views I am developing here are not in the target area of Quine’s formidable attack on Carnap’s view on truth-purely-in-virtue-of-meaning (particularly in “Carnap on Logical Truth,” in Quine 1976). The truth-condition for any sentence containing a logical operator, including primitive logical axioms, is determined in the same uniform way, by application of the clauses of a Tarskian truth-theory. In fact
Implicit Conceptions, Understanding, and Rationality
145
the ways of coming to know these axioms I have been identifying clearly rely on states whose informational content involves what is stated in the clauses of a truth-theory. (4)
Logical axioms are necessary.
This was a view Leibniz expressed repeatedly, and it was another plank in his criticism of Locke. Leibniz thought, rightly in my view, that Locke could explain neither the necessity of the axioms nor our knowledge of that necessity. Indeed, Leibniz repeatedly endorses the much stronger claim that for every necessary truth there is a reduction of it to logical axioms by means of substitution of definitions for defined terms.16 It would take us much too far afield to go into (4), or the stronger claim, in detail. It can be assessed only in the presence of a substantive theory of necessity. For now, let me simply say dogmatically that I think a correct account of the truth-conditions of necessity statements has the consequence that the semantical rules for logical operators hold not only in the actual world, but in any possible world.17 (Of course a theory must explain why this is so.) If it is so, then the sorts of rationale I have been considering for primitive logical axioms can be extended to show not only the validity of these axioms, but also their necessity. (5) Reflection is needed to discover the axioms of logic: it would “be wrong to think that we can easily read these eternal laws of reason in the soul, as the Praetor’s edict can be read on his notice-board, without effort or inquiry” (New Essays, p. 50). Leibniz in many places emphasizes that we need to reflect, to attend, to discover logical axioms. I have also emphasized that it can take reflection to appreciate that an axiom or primitive inference-rule is correct. If anything, I think Leibniz still overstates their ease of discovery, as in the passage quoted above in which he says they are evident as soon as their terms are understood. He writes as if reflection may be needed to discover the axioms, but that once stated, it will be evident that they are correct. This is too strong. I suspect that if Leibniz had been acquainted with modern logic, he would have withdrawn this point. As far as I can see, nothing in the rationalist conception rules out revision on this point. (6) There is an important distinction to be drawn between clear ideas and distinct ideas. A clear idea, for Leibniz, is one that enables one at least to recognize instances of the concept in question. A clear idea may nevertheless be indistinct, that is, “I am not able to enumerate separately the characteristics required to distinguish the thing from others, even though such characteristics and distinctions are really in the thing itself and the data which enable us to analyze the notion” (“Reflections on Knowledge, Truth and Ideas,” 1951, p. 284). It seems to me that the way to elucidate the distinction between someone whose
146
Christopher Peacocke
clear idea is indistinct and someone whose clear idea is distinct is to use the notion of an implicit conception. The thinker with the distinct idea is one who has succeeded in achieving an explicit formulation of the implicit conception that was all he had when he had only an indistinct, though clear, idea. Beyond this agreement on six particular theses, there is also an underlying sympathy between the approach I have been adopting and the general rationalist conception of knowledge. The views I have been putting forward are at home in a conception of knowledge as rationally or reasonably attained. No doubt, historically, the rationalists overshot in their enthusiasm for this idea. Nonetheless, if there is something in it, it would be a quite unstable position theoretically to hold that it applies to much of our knowledge, but fails to hold when we consider the case for primitive axioms and inference rules. If there are rationality or reasonability conditions for knowledge, they must apply in these basic cases too; and in effect I have been trying to argue that they do. 6
Consequences for the Theory of Justification, Rationality, and Understanding
What makes it rational to accept some logical truth of which one has a proof? These considerations suggest that the rationality of accepting it cannot be fully elucidated philosophically just by citing the proof. For what is at the start of this derivation? There are two sorts of cases to consider. In one type of case, the proof starts with primitive axioms and/or inference rules. The primitive axioms and/or inference rules may be universally quantified, as in Frege’s own formulations, or they may be schemata, as in current approaches. But whichever way the starting point of the proof is set out, under this first type of case the starting point of the proof itself does not involve metalinguistic notions. The notion of truth does not occur in the first line of the proof itself. Proofs of this first type we can call unsupplemented derivations. Proofs of the second type, by contrast, start with the semantic principles, stating the contribution of particular logical connectives to the determination of the truth-conditions of sentences or contents or thoughts containing them. They will start from such principles that ‘Any sentence of the form A … B is true iff A is false or B is true’. Proofs of this second sort then move from these semantic premises to the logical axioms or primitive inference rules, which the semantics validates. They then proceed as in the unsupplemented case. Proofs of this second kind we call supplemented derivations. Nothing in derivations of the first kind, the unsupplemented derivations, explains the rationality of accepting their staring point, their primitive axioms or inference rules. This state of affairs is especially perplexing for anyone who holds the highly intuitive and (it seems to me) correct view that it is something about the nature of the senses of the expressions in the primitive axioms and inference rules, and correspondingly about the thinker’s
Implicit Conceptions, Understanding, and Rationality
147
understanding, that makes it rational to accept them. So we may be tempted to turn to derivations of the second type. Yet it does not seem that they fully explain the rationality of accepting a logical truth, for two reasons. One is that a person can come to recognize the nonsemantic axioms as valid, while having no explicit knowledge of semantics. It is no doubt partly this point that attracted the classical rationalists to the view that axioms are known independently of other truths. Some prominent logicians, such as Russell, developed logical systems, and knew their axioms, before having been introduced to explicit semantical statements (in his case, by Wittgenstein). The other reason is that the rationality of accepting the semantic axioms is of course itself still unexplained. The problem is very sharp in Frege himself. Frege held that “it is part of the concept of an axiom that it can be recognized as true independently of other truths” (1979, p. 168). In conformity with this, in the formal system of the Basic Laws of Arithmetic, we have a system with primitive, nonsemantic axioms and inference rules. It was derivations in this system that were supposed to give the “ultimate justification” for arithmetical propositions. Yet there are many pressures in Frege to want a different position, and these pressures are reflected in the way Frege himself proceeds in the Basic Laws of Arithmetic. You might expect someone who holds the quoted rationalist doctrine that axioms can be recognized as true independently of other truths not to derive them from other truths. Yet at every point at which Frege introduces a new primitive axiom or inference rule in the formal language, he actually gives it, in the German, a justification. Indeed, a word meaning “therefore” immediately precedes the statement of the axiom or inference rule. And what precedes the “therefore” is a statement of the semantic rule—the rule giving the contribution to truth conditions—of the crucial expressions in the new axiom or inference rule. The very first axiom is introduced by a simple argument that it must always have the truthvalue true, given the truth-rule for the material conditional. Frege writes—I change to a more modern notation for the material conditional and for the variables— By § 12 [which states which truth-function the conditional denotes], (A … (B … A)) could be the False only if A and B were true while A was not the True. This is impossible; therefore (A … (B … A)) (I).
Here “(I)” is Frege’s notation for his first axiom. He does something similar for every other axiom and inference rule. This, incidentally, shows that it is quite false to say, as Ricketts (1996) does, that Frege never attempts any informal soundness proofs. These are precisely informal soundness proofs. Frege’s giving an informal soundness proof is not a mere quirk of exposition. It has deep roots in his, and indeed I think in the proper, conception of the relations between sense, justification, and truth. Frege held, like Leibniz, that “the truth of a logical law is immediately evident from itself, from the sense of the expression” (“Compound
148
Christopher Peacocke
Thoughts,” 1984, p. 405). There ought, if the evidentness of a logical law depends on the sense, to be some explanation of how it depends on the sense. Now Frege famously held that the sense of a sentence is given by its truth-conditions, and that the sense of its components by the contribution they make to these truth-conditions. This is precisely how Frege’s soundness proofs proceed. Moreover, in proceeding this way, he is giving a justification for thinking that the truth-condition for the axiom is fulfilled. So we are not, after all, lacking at the very foundation the sort of justification which it was the task of the rest of the structure to provide. Frege did not recognize the tension, given his account of axioms, nor did he resolve it. It is not only a problem for him; it is a problem for us too. Even if we do not accept his characterization of what an axiom is, and so have no problem with a semantic derivation of the truth of an axiom, there is an unresolved issue. Why is it rational to accept the semantic premises? There must be some answer to this. For if a logical derivation is not an example of a justified, rational route to a conclusion, then what is? And how can it be so unless its premises are justified? I suggest that both the supplemented and the unsupplemented derivations have a role to play in explaining the thinker’s justification, and that the way of solving the problem is to distinguish two very different relations a single thinker bears to derivations of each of the two kinds. Every step in the unsupplemented derivation corresponds to some transition explicitly made by a thinker who is inferring some logical truth from the primitive logical axioms. The thinker finds those axioms compelling, and does so by a proceeding along the lines we envisaged for our eighteen-year-old early on in this essay. Equally, when we consider the supplemented derivation, with its semantic premises, its earliest parts also capture something psychologically real, but they do so in a different way. The semantic premises of the supplemented derivation give the content of those of the thinker’s implicit conceptions that are operative in his rationally coming to appreciate that the nonsemantic axioms are valid. A statement of the implicit conception associated with understanding a truth-functional connective, for instance, would be a statement which determines its truth table. When our imagined thinker goes properly through the simulation-involving steps I described at the start of this essay, he will come to accept as valid the same principles as someone who is explicitly inferring from a statement of the truth tables. Moreover, the explanation of his doing so will be that his implicit conceptions entail the content stated in those truth tables. If we regard understanding the logical expression as involving association of the expression with the right implicit conception, we also see on this account how semantic understanding is the source of the thinker’s appreciation of the validity of the logical axioms (and primitive rules). I should add that in making these points, there is no commitment to the possibility of some level of theory at which everything, including all logical transitions, can be justi-
Implicit Conceptions, Understanding, and Rationality
149
fied. Even the mental activities of someone whose thought is captured by the supplemented derivation is still using logic at the very early states, for instance in moving from information gleaned by simulating the various cases to acceptance that an axiom or inference rule is valid. He will be making such transitions as: if these are possible combinations of truth-values, and there are no others, then this axiom (say) will always be true. We need some form of logic in any theoretical thought. So it seems to be incoherent to suppose that there is some level at which everything can be justified. What this shows is that an $", an existential-universal, proposition is false: it is false that there is a level at which everything can be justified. That is consistent with a weaker, coherent proposition of "$ form being true: that for every axiom and inference rule, there is some level at which it can be justified. This "$ proposition is much more plausible. It may be a requirement of reason. I draw two conclusions from these points. The first conclusion is that even in the area of logic, the rationality of accepting a proposition or schema on the basis of one’s understanding the expressions it contains cannot be explained solely in terms of proofs, not even supplemented proofs. For whence does our rational understander obtain his knowledge of the semantic premises? If he has that knowledge because others have informed him of these premises, then he is not obtaining his knowledge from his own understanding of the logical vocabulary. It is only if the knowledge is obtained ultimately by the simulation means we outlined that the source of the knowledge is his own understanding of the expressions. Of course he does not have to go through the simulation every time he needs to use a semantic premise. Knowledge obtained by the simulation can be stored for later use without the thinker having to rehearse its origins, just as knowledge obtained by any other means can be stored without rehearsing its origins. But the status of the stored information as knowledge derived from his own understanding rests on its having been reached by the simulation method. On this view, then, a full account of the rationality of accepting a logical law, when that derives from the nature of the thinker’s own understanding, has to mention implicit conceptions and the way in which reflective simulation provides a means of extracting the informational content of implicit conceptions. The other conclusion concerns the possibility of explicating Frege’s notion of sense in part by appeal to some ideal understander, and the axioms that an ideal understander would accept. Probably Frege was attracted to such an explication, and there are aspects of his thought and presentation that square with it. But it does not follow that we (or that he) should accept that explication in terms of what ideal understanders would accept. The resource to which I have appealed in explaining rational acceptance of the axioms is that of implicit conceptions of the semantic contributions of certain expressions. I have emphasized that these implicit conceptions play a part in ordinary thinkers’ evaluations of sentences. They are something possessed by ordinary, and not only idealized, thinkers. Insofar
150
Christopher Peacocke
as a proposed axiom can be recognized as true, it is recognized on a basis that is rooted in the ordinary, non-idealized understanding of the expression of the axiom. Acknowledgments Earlier versions of this material were presented at the 1996 Barcelona meeting of the European Society for Philosophy and Psychology, at the Universities of Hamburg and St. Andrews, and at CREA, Paris. It will soon be apparent to those who have followed my earlier efforts that the present paper involves a significant change of view. In grappling with these issues, I have been helped particularly by the comments and advice of Paul Boghossian, Bill Brewer, Tyler Burge, John Campbell, Martin Davies, Hartry Field, Wolfgang Künne, Stephen Schiffer, Stewart Shapiro, John Skorupski, and Crispin Wright. This paper also appears, with commentaries and a reply, in Villanueva (1998). Notes 1. A Study of Concepts claimed that certain axioms are found primitively compelling by those who understand their terms. The description I have just given in the text does not contradict that claim. The problem with A Study of Concepts is rather that it gives no elucidation of the rationality of accepting primitive axioms. Simply saying that they are noninferentially accepted is much too undiscriminating. Adding that they are noninferentially accepted on the basis of the thinker’s understanding at least makes clear that understanding plays an explanatory role. It does, though, still fail to say what ‘on the basis of’ amounts to here, or to say how rationality is implicated. 2. Did Newton and Leibniz actually operate with different, but equivalent, implicit definitions of the limit? Newton’s informal explications are closer to the Bolzano-Cauchy-Weierstrass definition, while, at first blush, Leibniz’s seem like those one would give in the theory of infinitesimals. Ishiguro (1990, chapter V), however, argues that ‘infinitely small’ was regarded as contextually defined by Leibniz, and so not thought by him to be referential vocabulary. 3. Here too I am at one with Burge’s elaboration of the Fregean position: “I think that Frege’s conception attempts to bridge the gap between actual understanding and actual sense expression by means of a normative concept— that of the deeper foundation or justification for actual understanding and usage” (1990, p. 47). 4. Crimmins considers a range of grades of ‘as if’ clauses at p. 257. 5. Near the start of his paper (p. 241), Crimmins also says that the notion of tacit belief may be needed to explain the relation between thinkers and nontrivial analyses of concepts. The Bolzano-Cauchy-Weierstrass definition is a nontrivial analysis of a concept. It is, then, a question whether it is quite the same standard, or grade of strictness, of ‘as if’ clause that we need to accommodate both the more trivial examples of tacit belief and the nontrivial analyses. 6. This first deflationist is also vulnerable to a near-ubiquitous problem with counterfactual analyses of categorical notions. We must be able to distinguish between someone who has an understanding of ‘is a whole number’ in advance, and someone who gains it in the course of his coming to meet the antecedent of the counterfactual. This distinction is incompatible with identifying understanding with something that simply has this satisfaction of the counterfactual as one primitive constituent. 7. Here I diverge from Field (2001), who brings in cosmological considerations to make sense of a determinate notion of finiteness, and to rule out nonstandard models. On the position I am advocating, primitive recursion with a limiting clause like (3) is explanatorily more fundamental than the general notion of finiteness. We do
Implicit Conceptions, Understanding, and Rationality
151
not need to rely on any empirical truths about the physical universe to classify the nonstandard models as nonstandard. It has of course to be part of this position that the modal ‘can’, occurs essentially in (3), is not itself to be elucidated in arithmetical terms not governed by an implicit conception—otherwise the problem of nonstandard interpretations would be with us again. I also note for the record that it is an obligation of the present position to say why the modal approach is to be preferred to second-order characterizations of the natural numbers. 8. I thank Stephen Schiffer for helping me improve on an earlier formulation of this objector’s position. 9. It is not only inconsistent conceptual roles, like that which Prior assigned to ‘tonk’, that fail to determine contributions to truth-conditions (Prior 1960). There can also be a failure to determine truth-conditions in the case of a consistent conceptual role, if the specification of the role also states that certain rules are invalid for the connective being explained. Examples are found in Peacocke (1993). 10. “Explicit inferences based on mental models, however, do not need to make use of rules of inference, or any such formal machinery, and in this sense it is not necessary to postulate a logic in the mind.” Johnson-Laird (1983, p. 131). 11. For instance, in Remarks on the Foundations of Mathematics (1978), VII §26 (p. 328): “But, if you have seen this law in it [a series of numbers—CP], that you then continue the series in this way—that is no longer an empirical fact. But how is it not an empirical fact?—for ‘seeing this in it’ was presumably not the same as: continuing it like this. One can only say that it is not an empirical proposition, by defining the step on this level as the one that corresponds to the expression of the rule.” 12. This is a point of contact between the theory of implicit conceptions and the position of A Study of Concepts. 13. Correspondingly, beliefs rationally explained by the possession of an implicit conception may be innate only in C. D. Broad’s “negative sense of internally generated” (1975, p. 138). 14. I am, though, uneasy about Leibniz’s implication that a priori knowledge is really knowledge about the thinker’s own mind. The content of a thinker’s understanding—the content of his implicit conceptions—can explain a thinker’s a priori knowledge without that knowledge being about his understanding. What his understanding makes available is not itself about his understanding. A later rationalist like Frege is clearly, and in Frege’s case explicitly, free of any commitment to the idea that a priori knowledge concerns the thinker’s own mind. 15. In this respect at least, my views have not changed since my paper “How Are A Priori Truths Possible?” (1993). 16. See Leibniz (1969, p. 646). Also in the New Essays, p. 86, of the kind of truths that are innate in Leibniz’s sense, he writes “and among necessary truths no other kind is to be found.” 17. For a theory of the truth-conditions of statements of necessity that I think can serve this purpose, see the sections on necessity in Peacocke (1999).
References Block, N. 1986. Advertisement for a Semantics for Psychology. In Midwest Studies in Philosophy. French (ed.). Minneapolis: University of Minnesota Press. Brewer, B. 1995. Compulsion by Reason. Aristotelian Society (suppl. vol.) 69: 237–253. Broad, C. 1975. Leibniz: An Introduction. Cambridge: Cambridge University Press. Burge, Tyler. 1979. Individualism and the Mental. Midwest Studies in Philosophy 4: 73–121. ———. 1986. Intellectual Norms and Foundations of Mind. Journal of Philosophy 83(12): 697–720. ———. 1990. Frege on Sense and Linguistic Meaning. In The Analytic Tradition, Bell and Cooper (eds.). Oxford: Blackwell. Chomsky, Noam. 1966. Cartesian Linguistics: A Chapter in the History of Rationalist Thought. N. Chomsky and M. Halle (eds.). New York: Harper and Row.
152
Christopher Peacocke
Crimmins, Mark. 1992. Tacitness and Virtual Beliefs. Mind and Language 7: 240–263. Davies, Martin. 1989. Connectionism, Modularity, and Tacit Knowledge. British Journal for the Philosophy of Science 40: 541–555. Field, Hartry. 1977. Logic, Meaning, and Conceptual Role. Journal of Philosophy 74: 379–409. ———. 1996. The A Prioricity of Logic. Proceedings of the Aristotelian Society 96: 359–379. ———. 2001. Which Undecidable Mathematical Sentences Have Determinate Truth Values? In his Truth and the Absence of Fact. Oxford: Oxford University Press. Frege, Gottlob. 1953. The Foundations of Arithmetic. J. L. Austin (trans.). Oxford: Blackwell. ———. 1964. The Basic Laws of Arithmetic. M. Furth (trans.). Berkeley: University of California Press. ———. 1979. Posthumous Writings. P. Long and R. White (trans.). Oxford: Blackwell. ———. 1984. Compound Thoughts. In Collected Papers on Mathematics, Logic and Philosophy, B. McGuinness (ed.). Oxford: Blackwell. Geach, P. 1972. Logic Matters. Oxford: Blackwell. Harman, G. 1972. Logical Form. Foundations of Language 9: 38–65. Ishiguro, H. 1990. Leibniz’s Philosophy of Logic and Language, second ed. Cambridge: Cambridge University Press. Johnson-Laird, P. N. 1983. Mental Models. Cambridge, MA: Harvard University Press. Leibniz. 1951. Leibniz Selections. P. Wiener (trans.). New York: Charles Scribner’s Sons. ———. 1969. Philosophical Papers and Letters. L. Loemker (trans.). Dordrecht: Reidel. ———. 1981. New Essays on Human Understanding. P. Remnant and J. Bennett (trans.). Cambridge: Cambridge University Press. Peacocke, Christopher. 1987. Understanding Logical Constants: A Realist’s Account. Proceedings of the British Academy 73: 153–200. ———. 1992. A Study of Concepts. Cambridge, Mass: MIT Press. ———. 1993. How Are A Priori Truths Possible? European Journal of Philosophy 1: 175–199. ———. 1994. Content, Computation, and Externalism. Mind and Language 9: 303–335. Reprinted in Contents (1995), Enrique Villanueva (ed.). Atascadero: Ridgeview. ———. 1999. Being Known. Oxford: Oxford University Press. Prior, A. 1960. The Runabout Inference-Ticket. Analysis 21: 38–39. Quine, W. V. O. 1976. Carnap and Logical Truth. In The Ways of Paradox and Other Essays, second ed. Cambridge, MA: Harvard University Press. Ricketts, T. 1996. Logic and Truth in Frege. Aristotelian Society (suppl. vol.) 70: 121–140. Sellars, Wilfrid. 1974. Meaning as Functional Classification. Synthese 27: 417–437. Stewart, Ian. 1996. From Here to Infinity. Oxford: Oxford University Press. Villanueva, Enrique (ed.). 1998. Philosophical Issues 9: Concepts. Atascadero, Calif.: Ridgeview. Wittgenstein, L. 1958. Philosophical Investigations, second edition. G. E. M. Auscombe (trans.). Oxford: Blackwell. ———. 1978. Remarks on the Foundations of Math-ematics, third edition. G. E. M. Auscombe (trans.). Oxford: Blackwell. Wright, C. 1989. Wittgenstein’s Rule-Following Considerations and the Central Project of Theoretical Linguistics. In Reflections on Chomsky, A. George, ed. Oxford: Blackwell.
8
Burge on Mentalistic Explanations, or Why I Am Still Epiphobic
Fred Dretske Tyler Burge and I agree about many things. We are both realists about psychological content. Neither of us sees any reason to believe that our use of intentional constructions in describing and explaining human behavior and abilities is “merely heuristic, instrumentalistic, or second class in any other sense” (1986, p. 8). We believe that appeal to such content is, and shows every promise of remaining, an essential part of understanding why people do the things they do. We are also semantic externalists or, to use Burge’s preferred term, nonindividualists about content. Thoughts can, and generally do,1 fail to supervene on the individual having the thought. What a person thinks can (in some suitably strong sense of ‘can’) change while the internal physical events (specified nonintentionally and individualistically) that constitute the thinking remain constant. Just as twin pieces of paper (money and perfect counterfeit, say) need not have the same (legal) monetary value, physical twins need not have the same thoughts. One can be a materialist about money without supposing that the material composition of money makes it money, and one can be materialist about thought without supposing that the material composition of thought makes it thought. This already constitutes a substantial body of agreement. But there is more. Unlike those philosophers who think the second point of agreement (externalism) conflicts with the first (the explanatory role of content), Burge and I see no incompatibility here. The external individuation of thought, the fact that physically identical individuals2 could be having different thoughts, does not mean that their behavior cannot be explained by what they think. This third point of agreement is much more contentious. Judging from the literature I have seen, Burge and I may be the only ones that agree about it. Lots of people think content is causally relevant to behavior. This is simple commonsense. Clyde goes to the fridge because he wants a beer and thinks there is one there. What he thinks—that there is a beer in the fridge—explains why he goes there. A fair number of philosophers also think psychological content—some of it, anyway—is relationally individuated. But precious few think both—that content can be externally individuated and causally relevant to the explanation of behavior. Burge and I are exceptions. Everyone else seems to think that externalism makes content—that much of it that is external, anyway—epiphenomenal. But here we part company. I think that epiphenomenalism is not a critical problem for semantic externalism because the problem (I regard it as a real one) is surmountable. Explaining Behavior (Dretske 1988) was devoted to surmounting it. Without some account of the sort provided there—or so I thought (and still do)—there is a tension, a mystery, a puzzle. What possible role could externally individuated thoughts, thoughts whose mental identity (as this thought rather than that thought) depends on how things are
154
Fred Dretske
outside the thinker, play in the explanation of changes that arise within the thinker (i.e., behavior)?3 Burge, on the other hand, does not even see a problem here—much less one that requires a book-length solution. Worries about epiphenomenalism, he thinks, reflect mistaken philosophical priorities (1993, p. 97). They have an air of make-believe about them (1993, p. 102). They have more in common with political or religious ideology than with a philosophy that maintains perspective on the difference between what is known and what is speculated (1993, p. 117). It is much surer that epiphenomenalism is false than that the (materialistic) assumptions leading to it are true (1993, p. 102). Instead of wasting time constructing elaborate naturalistic accounts of how (given our metaphysics) mental causation is possible, we should be developing a metaphysics of mental causation by looking at our explanatory practice. Instead of worrying about how mentalistic explanations can work, we should be using the fact that they do work as an instrument for clarifying beliefs about the place of mind in the natural scheme of things. Mentalistic explanations and causation do not need validation from metaphysics (1993, p. 117). The dog, the starting point, is explanatory practice; the tail it wags is metaphysics (1993, p. 118). I respect our explanatory practice as much as the next guy, and I agree with most of Burge’s methodological strictures (I do not, for instance, think epiphenomenalism is a serious philosophical option), but Burge’s embrace of explanatory and descriptive practice over metaphysical theory strikes me as tantamount to ignoring Zeno’s theoretical problems because they conflict with accepted locomotive practice. In both cases I know what the practice is. Despite Zeno, we go from here to there. Given the endless number of intermediate points, though, I don’t understand (at least I didn’t; I think I do now) how we manage to do it. Likewise, I know what our explanatory practice is: Clyde goes to the fridge because he thinks there is a beer there. That is why he does it. What I don’t understand, though, is how having this thought can explain his trip to the fridge if Clyde’s having it depends on conditions outside Clyde’s body, if the explanatory content is (as they say) broad. How can his thinking that get him off the couch and to the fridge? I’m willing to accept the fact that it does explain it, just as I’m willing to accept the fact that we do move from place to place. I agree with Burge: These facts are more obvious than the metaphysical assumptions that appear to threaten them. But that leaves me puzzled about how things can work that way. I can be puzzled about this while conceding that things do work that way. One can see a problem here without supposing that nonindividualism threatens to rob thought of all explanatory relevance. No; it all depends on what is being explained. Psychology, as Burge reminds us, is not a monolith: Different explanatory tasks and types of explanation coexist within it (1986, p. 10). We are, for instance, sometimes interested in explaining successes and failures, reliabilities and achievements (Burge 1986, pp. 25, 29,
Burge on Mentalistic Explanations
155
32; 1995, p. 228). Since success, failure, reliability, and achievement typically depend on being suitably situated in an environment, it should be no great surprise to find that relational properties of the performer (and therefore, possibly, externally individuated thoughts) might be necessary (or, at least, eligible) to explain it. There is also the related fact that what is classified as behavior in both everyday affairs and psychology often fails to supervene on the individual whose behavior is being explained (see Burge 1986, p. 11; Dretske 1988, chapters 1 and 2). The behavior we use mental content to explain, therefore, is often as implicated in external affairs as are the facts about what we believe and desire that (according to semantic externalism) we invoke to explain it. When this is so, there is, I agree, no particular mystery about why externally individuated mental states play a role in mentalistic explanations. Since we can easily imagine identical persons doing entirely different things as a result of being situated in different circumstances (one points at Tom; the other points at Jerry) I see no reason why differences in such behavior should not be explained by differences in the relations people bear to their environment (one sees Tom, for example; the other Jerry) and, therefore, possibly differences in their beliefs about these relations. All this seems reasonably straightforward. Nonetheless, there is behavior that is not (at least not on the surface) externally individuated. We explain why the witness pointed at the accused, yes, and pointing at the accused is not something you can do all alone in the universe. You need the accused. But we also explain why the witness raised her arm, changed positions, rubbed her nose, closed her eyes, lowered her head, stood up, wept, and fainted. You don’t need anyone else to do these things. Even when we describe behavior in such a way as to necessitate other objects, such behavior is always carried out by doing something that does not depend (at least not in the same way) on the existence of external objects. One points at the accused by raising one’s arm and extending one’s finger toward a person one takes to be the accused. You don’t need the accused to do this. You need a person, yes, but not the accused. You don’t even need a person to perform an even more basic action the performance of which (given a properly situated accused) is pointing at the accused—namely, raising one’s arm and pointing in the direction one takes an accused to be. This being so, it seems plausible to suppose that persons who do different things because of the external individuation of what they do will always be doing the same thing at some more basic level of description, and the question that bothers me is whether their externally individuated beliefs are relevant to what they do at this more basic level. If I, wanting to point at the accused, point at the man on my left because I think the man on my left is the accused, my behavior is explained by a belief (that the man on my left is the accused) that—so expressed—I cannot have unless there is a man on my left. In this way my behavior is explained by an externally individuated belief. But in what way is the fact that this belief requires a man on the left (and is, thus, externally individuated) relevant to the way it explains what I do? The fact
156
Fred Dretske
that there is a man on my left explains why, in pointing to the left, I point at a man on my left, but it does not explain why I point to the left. I would have pointed to the left even if there had not been a man there—for example, if I had mistakenly thought the accused was there. So why is there being a man on my left—and, therefore, my having a belief whose verbal expression requires it—relevant to understanding why I do what I do? Let’s be clear about what such examples show. They do not show that externally individuated beliefs do not figure in the explanation of behavior. They, in fact, show just the reverse. I pointed at the man on my left because I thought he was the accused, and the belief that he (the man on my left) was the accused is an externally individuated belief. It does not supervene on the person whose behavior it explains. No, what such examples suggest (and they only suggest it)4 is not that externally individuated beliefs do not figure in the explanation of behavior, but, rather, that they can always be eliminated with no loss of explanatory force. The aspects of the belief in virtue of which they are externally individuated, those aspects that require the existence of external objects, are not relevant to understanding why the agent moves in such a way (pointing to his left) so as to give rise (given appropriate circumstances: the accused on the left) to behavior (pointing at the accused) that is being explained. I have not yet said anything that Burge doesn’t already know. I have merely been describing the kinds of suspicions—the threats, if you will—that bother me. The worry is not that externally individuated beliefs do not play a role in the explanation of behavior. Of course they do. See above. The worry is that their role is dispensable—that the same explanatory results can always be achieved without mentioning those aspects of belief that fail to supervene on the believer. Broad content, qua broad content, doesn’t do anything— nothing, at least, that isn’t already being done by narrow content and facts about the circumstances in which what is done is done. I agree with Burge (1986, p. 15) that materialism per se does not support individualism. Our classificatory practice—both in science and in common sense—is rife with nonindividualistic notions. Concepts like information and representation are obvious and relevant instances (see Dretske 1981, 1988; also Burge 1989, pp. 306, 309–310 for other telling examples). The disagreement, then, is not about the existence—not even the prevalence—in both science and common affairs of relational concepts. The issue, at its most general level, is about the explanatory role of relational properties. It is not about whether classifying people as aunts and uncles is a useful way of classifying people, but about whether anything aunts and uncles do is explained by their being aunts and uncles. Since I can become an uncle as the result of events occurring far away (theoretically, outside the light cone that defines the limits of causal interaction), there is a question about how anything I do can be explained by my being an uncle (isn’t this, potentially at least, a violation of Relativity Theory?). Nonindividualism says that beliefs are like uncles—entities
Burge on Mentalistic Explanations
157
that are relationally individuated. Just as uncles can be “affected” (i.e., their status as uncles depends on) events with which they need have no causal contact, beliefs can be “affected” (their status as beliefs5 depends on) events with which they need have no causal commerce. This is a point that, Burge warns (1986, p. 16), is easily missed if one conflates causation with individuation. As far as I can see, though, this makes the causal and explanatory status of beliefs, qua beliefs, as mysterious as that of uncles qua uncles. Since this is (or was)6 a mystery to me, but not to Burge, I infer that I have some metaphysical preconceptions (presumably about explanation) that he doesn’t have. It may be that my naturalistic inclinations, my tendency to understand mentalistic notions (belief, experience, perception, knowledge) in terms of their physical realization (information, function, representation), give me special problems that those who (like Burge) are free of reductive proclivities do not have. This may be so. Therefore, in an effort to identify the source of our differences (and cognizant of the fact that Burge is going to reply to this essay) I propose to put my metaphysical cards on the table. If nothing else, this may help Burge identify why I worry about things (like epiphenomenalism) that, according to him, really shouldn’t bother me. I am driven to semantic externalism by general considerations about representation. Thought is a species of representation, and since I cannot think of any plausible theory of representation that is individualistic,7 I see no way to avoid externalism about thought (and experience).8 What gives an instrument, a gauge, a diagram, or a road sign the power to represent some object, a, to be F? What makes such artifacts say or mean that a is F?9 It is, I submit, some fact about their relations to external affairs—typically (and, I think, unexceptionally) some fact about their function. What makes the gauge say or mean that the tank is empty (whether or not it is empty), what gives it the power to represent the tank as being empty, is that it has the function (deriving from our design and use of it) of indicating how much gas is in the tank. It is something about the instrument’s history— its design and use—that gives it this power. The gravitational force on the bolts holding the gas tank to the frame of the car also indicate (i.e., co-vary with) the amount of gas in the tank, but, unlike the position of the needle on the gauge, these bolts do not represent10 the amount of gas in the tank. These bolts do not have that function; they cannot, therefore, as can the gauge, misrepresent the amount of gas by failing to indicate it when things go wrong. If one takes (as I do) such artifacts as a model for understanding representation in general, then biological representations (those whose indicator functions—unlike that of artifacts—are not determined by the intentions and purposes of designers and users) will have representational powers (and, thus, content) determined by whatever natural processes confer functions on bodily organs and states. The only processes that do this are natural selection and certain types of individual learning (the sort involved in concept
158
Fred Dretske
formation). Evolution and learning are processes that confer indicator (informationproviding) functions on various bodily organs and states. They give rise, therefore, to internal representations of whatever conditions these organs and states have the function of indicating. Since both natural selection and learning are historical processes, the resulting account of representation is nonindividualistic. What some material state of the brain represents, then, depends not just on its material constitution, but on its history, on what it was “designed” (by learning or natural selection) to indicate. Just as one can imagine physically identical instruments being assigned different indicator functions (one is used as a pressure gauge, the other as an altimeter), a condition of the body can be imagined to represent different external magnitudes by means of different developmental or evolutionary histories. On this account of representation, what systems represent does not supervene on their material constitution. Biological twins might represent their environments in much different ways. Since this is the only way I can imagine a material system acquiring the power to represent, the power of saying something false, I am, as a materialist, and as long as I cannot imagine a more plausible alternative, committed to an externalistic view of thought. That is the only way I can imagine something meaning that a is F (in a way compatible with a’s not being F) without the help of systems (conscious beings) who can already represent (i.e., think of) something as F. Not everyone—in fact, probably no one—agrees with me about all this. That isn’t necessary. What I expect agreement about—especially from materialists—is that whatever makes one bit of matter about something else—about, say, football rather than philosophy—what gives it its intentional character, is not intrinsic to that bit of matter. It is something about the way this bit of matter is related to the rest of the world, something, perhaps, about its history, about the context in which it occurs, or about its causal and informational connections to other things. With this much, I am confident, Burge agrees. Indeed, he is the most articulate and insightful proponent of such a view. So even if he does not endorse my specific story about which relations are crucial, or what bit of history is the important part, he agrees that it is relations to the stuff out there (and the stuff back then) out of which the mind—intentional content—emerges. So, given that Burge thinks causation is a local affair (1986, p. 17), that causal influences on (and, presumably, by) mental states are transmitted only via neighboring stimulations on (and by) physical states of the body, why isn’t this enough to give Burge the same problems I have? If causation is local but individuation is not, this means that the conditions that (help) make beliefs the beliefs they are need have no causal commerce with the beliefs they help individuate. How, then, can the fact that it is a belief that P (not Q or R) explain whatever effects the belief has on the believer’s bodily movements and,
Burge on Mentalistic Explanations
159
thus, behavior?11 Isn’t this like saying that uncles are empowered simply by being uncles— by the mere activities of (with or without causal connections to) remote siblings? To press this point, let me change examples in order to get something that more nearly approximates the situation that actually prevails in mind–body interaction. This will, I think, give us a better picture of what is at stake. The United States Treasury does its best to make real $20 bills intrinsically different from all other physical objects. Special watermarks, intaglio printing, high-quality paper, and security strips (visible only with transmitted light to frustrate photocopying) make successful counterfeiting difficult and increasingly rare. As a result of such efforts, and for practical monetary purposes, the property of being $20 supervenes on the paper’s intrinsic properties. You can, by looking carefully enough, tell a genuine $20 bill from a fake. All this is true even though, of course, being a real $20 bill is an extrinsic property of the paper, a causal-historical property, that has to do with where (and by whom) the paper was printed. Unless the paper was printed by the U.S. Treasury Department’s Bureau of Engraving and Printing it is not a $20 bill. That is why you cannot make $20 bills in your basement even though, were you clever enough, you could print paper identical to them. Thanks to the Treasury Department, then, the property of being real money supervenes on the money’s intrinsic features (the way it looks and feels). As a result, we can (and do) use the fact that something is a genuine $20 bill to predict and, sometimes, even “explain” the effects this paper has in transactions of all sorts. Why did the cashier give me $8? Because lunch cost $12 and I gave her $20. Is that really why I got $8 change? Giving the cashier $20 is what caused her to give me $8 in change. About that there is no argument. The question is whether its being $20 is what explains this result. What if I had, instead, given the cashier a piece of paper that looked and felt exactly like a real $20 bill? Would anything have changed if we suppose, counterfactually, that the bill were perfect counterfeit? Would she have given me $8 change then? Yes, of course. Well, then, why say that the cashier gave me $8 change because I gave her $20? That fact that I gave her $20 is, apparently, not the explanation of why she gave me $8 change. The right explanation is that I gave her a piece of paper that looked and felt (to her) like a $20 bill. The causally effective properties, those that explain why the effect occurs, are the intrinsic properties of the paper on which its being a $20 bill supervenes, the properties cashiers use to tell whether it is a $20 bill. I know this sounds a little silly, but I don’t see what is wrong with it. Giving the cashier a $20 bill (together with the price of lunch) may be the cause of my receiving $8 in change, but it is not the explanation. Its being $20 is not the effective property of the cause, the property that explains why the cause has that effect. Thanks to supervenience (thanks, that is, to the fact that anything—or most things—that look and feel like a $20 bill are $20
160
Fred Dretske
bills) there is a regularity, a pattern, a lawlike regularity, that we can express in terms of the paper’s monetary value. Give a cashier $20 for a $12 item and, ceteris paribus, you will receive $8 in change. More generally, give someone $X for something costing $Y and, ceteris paribus, you will get $(X–Y ) in change. That is a perfectly familiar pattern. We all rely on it for predicting and (if we don’t think about it too much) “explaining” the results of monetary transactions—why, for example, we get the change we do. If truth be known, though, giving the cashier a $20 bill is not the explanation (the cause, yes, but not the explanation) of why I received $8 change. I am not, of course, recommending that we change explanatory practice. I (for one) will go right on explaining monetary transactions in terms of the money exchanged even though I am convinced that its being money is (in most imaginable cases) totally irrelevant to the results obtained. I know the vending machine would have given me a Coke even if I hadn’t put quarters in it. Four pieces of metal of the right size and shape would have done quite as well. Coke machines, as we all know, are designed to detect the shape, size, and density, not the history, of the objects placed in them. Nonetheless, I will “explain” the machine’s behavior (giving me a Coke) by mentioning the historical properties (they were quarters), not the intrinsic properties (shape, etc.) of the objects that cause the machine to behave this way. When we operate in areas where an externally individuated property E supervenes (for practical purposes) on some intrinsic property P, we tend to cite E in our “explanations” of why the thing possessing P (and E) has the effects it has. Supervenience on a wide enough scale gives rise to patterns that can be formulated in the nonindividualistic vocabulary of “E” instead of “P.” Despite this practice, though, these generalizations are only lawlike; they need not be laws (and in the case of money, are not laws)12 nor need the “explanations” they support be real explanations. In more careful moments—when, for instance, we are doing metaphysics—we acknowledge that it was the fact that the object was P, not the fact that it was E, that was the causally relevant fact. Our explanatory practice does not respect metaphysical scruples. There is no reason it should. In giving and receiving these explanations we are not doing metaphysics. We take explanatory shortcuts. We leave the metaphysics for later. Or never. We can, if we like, describe transactions between quarters and Coke machines as monetary–machine interactions. Such interactions may be made to appear mysterious for the same reason mind–body interactions appear mysterious: An externally individuated type (money) is invoked to explain a physical result (dispensing a Coke). Those who regard this as a mystery will insist (rightly, as far as I can see) that, despite what we say, despite our explanatory practice, the monetary facts, the fact that you put quarters in the machine, are epiphenomenal.13 What explains the result are “narrow” facts about the size, shape, and density of the coins deposited. They will also insist (and with this I disagree) that, for exactly the same reasons, what you believe (when externally individuated) is irrelevant to
Burge on Mentalistic Explanations
161
the effects a belief has on bodily behavior. Though I disagree with this latter claim, I am sympathetic with the feeling that something has to be said to distinguish these cases from one another. Without a model of mind–body interaction that distinguishes it from monetary–machine interaction, one is left with the impression that these processes are, in effect, the same—that the causal relevance of beliefs on the behavior of people is the same as the relevance of quarters on the behavior of machines. I do not understand Burge’s indifference to this felt need.14 For reasons that such examples make clear, I am suspicious of what one can conclude from the facts of explanatory practice. Frankly, I do not trust explanatory practice—not, at least, when I’m doing metaphysics. To change the metaphor, I don’t see why beliefs are not our mental “quarters”—things inside us whose “monetary value” (i.e., content) is irrelevant to what we do, but which, nonetheless, cause us to behave the way we do and whose content correlates sufficiently well with causally effective properties to let us cite it in explanations of behavior. It is this suspicion, this possibility, that leaves me dissatisfied with Burge’s answer to Fodor’s (1987) arguments. Fodor asks why believing P isn’t like being an H-particle (a particle that exists at a time when Fodor’s dime is heads up). I do not think it enough to say, as Burge does (1989, p. 308), that (unlike the “physics” of Hparticles) there are, in the psychology of believers, lawlike generalizations describing actual causal patterns that are formulable in the vocabulary of intentional content. That is true, but it isn’t enough to show the causal relevance of the properties picked out by that vocabulary. If it were, then we could conclude that monetary properties—a coin’s being an actual quarter—was causally relevant to the behavior of vending machines, and this is something that neither I nor (I assume) Burge (but see below) is willing to conclude from our explanatory practice. For the same reason, the fact that narrow content plays no explicit role in actual psychological explanations (1995, pp. 226–227) strikes me as no more relevant to what really explains a person’s behavior than does the fact that we hardly ever (and then only if things go wrong) mention an object’s intrinsic properties in explaining the effect these objects have on the behavior of vending machines. Most people, I assume, don’t mention these properties because they don’t know what they are. Burge is right that the special sciences, including psychology, study patterns of causation that exist in an environment that is considered normal for entities of the sort studied by that science (1989, p. 312). Physiology, for instance, individuates its kinds with a view to understanding the regularities that exist in an actual biological environment. What hearts do and what hearts cause is determined by the patterns that exist when hearts are located where they are normally located and when they are doing what they normally do. “It follows that the causal powers relevant to type individuating the explanatory kinds of a special science need not be ‘locally supervenient’ on causal powers recognized by
162
Fred Dretske
sciences that deal with underlying constituents” (1989, p. 317). As a result, a heart and a waste pump need not have the same causal powers just because they are physically the same (i.e., have the same causal powers from the point of view of physics). If they have different normal environments (as their names suggest they do), they would (in their respective environments) have different effects (one pumps blood, the other eliminates waste) and, therefore, different causal powers. The same could be said about monetary systems. Explanations couched in terms of the monetary value of the coins or paper occur within a normal context, a context in which monetary values supervene on a familiar set of intrinsic properties. Given appropriately different normal environments, we can imagine physically indistinguishable pieces of paper having quite different effects. Within a sufficiently different environment, giving a cashier an object physically indistinguishable from a $20 bill after an indistinguishable lunch might typically result in almost anything— including arrest. As long as we are interested in the patterns that exist in a normal environment, though, we can frame our “lawlike” regularities in terms of the monetary value of the objects being exchanged. Given a normal environment, giving a cashier a $20 bill for a $12 lunch will result in $8 change. All this is true, but what does it show about the causal efficacy of monetary value? Not much. All it shows is that monetary value can be made to appear causally efficacious if we restrict ourselves to environments in which monetary value supervenes on causally effective properties. Ordering a diaphragm of a microphone to vibrate rapidly makes it vibrate rapidly, but the device does not, for this reason, obey the command. It doesn’t obey because, although it does what it is told to do, it does not do what it is told to do because it is told to do it. It is not the meaning of what one says that is causally relevant to the microphone’s behavior. I mention this point, once again, in order to remind the reader (not Burge; he already knows it) that the issue is not about whether reasons are causes. Of course they are. Our problem is to understand how being a reason—this property—figures in the causal efficacy of objects having it. This possible confusion of causes with their causally effective properties is a confusion easy to avoid in the case of microphones and vending machines, but, for some people, much harder to avoid when they think about applications (e.g., computers) where the supervenience of the extrinsic on the intrinsic is more systematic and widespread. In the case of computers, careful programming ensures that there is a supervenience of extrinsically individuated meanings on intrinsically individuated symbols, a correlation that allows us to use meanings as explanatory surrogates for symbols in the same way we use an object’s monetary value as an explanatory substitute for its physically relevant properties in explanations of machine and cashier behavior. Given widespread supervenience, we talk, effortlessly and without metaphysical anxiety, about giving the machine commands and about its obeying these commands. The computer does this, we say, because we told it to. We say this often enough that we end up believing it. The
Burge on Mentalistic Explanations
163
worry, then—at least my worry—is that a corresponding fact (vouchsafed by Mother Nature) about the supervenience of content on neural properties has led us to speak—with the same ease but with as little truth—of doing things because of what we believe. Having said all this, I find myself wondering whether Burge will disagree with any of it. Maybe he thinks that psychological explanations of human behavior are like monetary explanations of machine behavior—the only difference is that, unlike mind–body interactions, we have, in the case of monetary–machine interactions, a convenient model (in terms of the intrinsic properties of the coins) for the underlying physical processes. In monetary–machine transactions we have a pretty clear picture of what the underlying causal processes are and what the causally effective properties are. The coins actually have properties to which we know the machine is (and was designed to be) sensitive, and on which the monetary value of the coins supervenes. So we understand how the historical events that make objects into quarters (that gave them their monetary value—this nonindividualistic property), give rise, in causally local ways, to properties (particular size, shape, etc.) that cause machines to dispense Cokes. In the case of mind–body interactions, though, we know of no corresponding neural properties (on which content supervenes) to which behavior-regulating mechanisms are sensitive. There must be such underlying (local) causal processes, of course, but we have no idea what they are. The only difference between mind–body and monetary–machine interactions, then, is that we don’t happen to know what are the causally effective properties of the mental coins that drive bodily behavior. So, from an epistemological point of view, we are stuck with talking about the “value” of these coins—what it is we believe and want. If this is what Burge believes, I must say that we agree about much less than I thought we did. For I think this view is merely a tricked-up version of epiphenomenalism. It makes intentional content as relevant to behavior as is the monetary value of coins to vending machines. This, I grant, is a kind of relevance, but it is not good enough to quiet my epiphobia. Notes 1. I say “generally do” because Burge thinks (1986, p. 6) that the sort of arguments he uses to show that thoughts about aluminum and arthritis can vary while an individual’s physical and functional history remains constant could be given for any notion that applies to public types of objects, properties, or events that are typically known by empirical means. This covers a lot. 2. Specified nonintentionally and individualistically. Hereafter, when speaking of the failure of content to supervene on the material (biological) constitution of the thinker (and his or her history), I will take these qualifications to be understood. 3. In speaking of behavior as consisting of changes that arise within the individual I do not mean to be suggesting that behavior consists of events occurring wholly (or even primarily) in the individual. The changes originate from within, but they need not (and generally do not) end there. Closing a door may begin within the door-closer, but it doesn’t end there. See Dretske (1988), chapter 1.
164
Fred Dretske
4. I realize that this example is not entirely appropriate to the point under dispute. Burge’s thesis concerns variations in belief that derive from the obliquely (intensionally) occurring expressions in content clauses—not, as in this example, variations in the reference of names and descriptions. But I’m not now arguing (I’ll get to that later)—only trying to suggest the source of a worry. The example does that well enough. 5. If not their status as beliefs, then at least which beliefs they are can be “affected” by events with which they have no causal contact. I am unsure what Burge thinks about the dependence of beliefs, qua beliefs (not just qua this or that belief), on external factors. 6. Before Dretske (1988). 7. Burge and I agree about this: “I think it questionable whether there is a coherent conception of objective representation that can support an individualistic language of intentional attitude attribution” (Burge 1986, p. 24). He certainly thinks the representational language of psychological theories is thoroughly nonindividualistic (see 1986, pp. 34, 40–41). 8. I’m a representationalist—hence, an externalist—about experience (sensations, qualia) also (see Dretske 1995), but there is no need to get into that here. We have problems enough with conceptual forms of representation (thought, belief, judgment). 9. In such a way that it can misrepresent a to be F. That is, when I speak of representation, I mean to be speaking of a power (to say or mean that a is F) the exercise of which does not depend on a’s being F. 10. They indicate it, yes, but they do not represent it. I use the word “indicate” to refer to that relation Grice (1989) called “natural meaning” and “represent” to stand for the relation he called “nonnatural meaning.” Nothing can naturally mean (i.e., indicate) that P unless P, but something can nonnaturally mean P (represent P to be true) without P being true. 11. It is the “thus” that I challenge in Dretske (1988). It is the conflation of behavior (moving your arm) with bodily movements (arm movement) that (I argue) prevents understanding how content can explain behavior without explaining (biology explains this) the bodily movements that constitute behavior. 12. Though they are, of course, supported by laws in another sense of “law.” 13. Once again, it is important to emphasize that it is not being denied that putting quarters in the machine causes it to dispense Cokes—only that their being quarters explains this result. 14. In the case of mind–body transactions, Burge (1993, p. 114) denies that a model is appropriate or even needed.
References Burge, Tyler. 1986. Individualism and Psychology. Philosophical Review 95: 3–45. ———. 1989. Individuation and Causation in Psychology. Pacific Philosophical Quarterly 70: 303–322. ———. 1993. Mind-Body Causation and Explanatory Practice. Mental Causation, John Heil and Alfred Mele (eds.), pp. 97–120. Oxford: Clarendon Press. ———. 1995. Reply: Intentional Properties and Causation. In Philosophy of Psychology: Debates on Psychological Explanation, Cynthia and Graham MacDonald (eds.), pp. 226–235. Oxford: Blackwell. Dretske, Fred. 1981. Knowledge and the Flow of Information. Cambridge, MA: MIT Press/A Bradford Book. ———. 1988. Explaining Behavior. Cambridge, MA: MIT Press/A Bradford Book. ———. 1995. Naturalizing the Mind. Cambridge, MA: MIT Press/A Bradford Book. Fodor, Jerry. 1987. Psychosemantics: The Problem of Meaning in the Philosophy of Mind. Cambridge, MA: MIT Press. Grice, P. 1989. Studies in the Way of Words. Cambridge, MA: Harvard University Press.
9
Mental Paint
Ned Block The greatest chasm in the philosophy of mind—maybe even all of philosophy—divides two perspectives on consciousness. The two perspectives differ on whether there is anything in the phenomenal character of conscious experience that goes beyond the intentional, the cognitive, and the functional. A convenient terminological handle on the dispute is whether there are “qualia,” or qualitative properties of conscious experience. Those who think that the phenomenal character of conscious experience goes beyond the intentional, the cognitive, and the functional believe in qualia.1 The debates about qualia have recently focused on the notion of representation, with issues about functionalism always in the background. All can agree that there are representational contents of thoughts, for example the representational content that virtue is its own reward. And friends of qualia can agree that experiences at least sometimes have representational content too, for example, that something red and round occludes something blue and square. The recent focus of disagreement is on whether the phenomenal character of experience is exhausted by such representational contents. I say no. Don’t get me wrong. I think that sensations—almost always—perhaps even always—have representational content in addition to their phenomenal character. What’s more, I think that it is often the phenomenal character itself that has the representational content. What I deny is that representational content is all there is to phenomenal character. I insist that phenomenal character outruns representational content. I call this view “phenomenism.” Phenomenists believe that phenomenal character outruns not only representational content but also the functional and the cognitive; hence they believe in qualia. This paper is a defense of phenomenism against representationism. Hence issues of reduction of the phenomenal to the functional or the cognitive won’t play much of a role. First I will briefly discuss an internalist form of representationism, and then I will go on to the main topic of the paper, externalist forms of the view. Internalism One form of representationism holds that the phenomenal character of experience is its “narrow intentional content,” intentional content that is “in the head” in Putnam’s phrase. That is, heads that are the same in ways that don’t involve relations to the environment share all narrow intentional contents. A full dress discussion of this view would discuss various ideas of what narrow intentional content is supposed to be. But this isn’t a full dress discussion. I will simply say that all versions of this view that I can think of that have even the slightest plausibility (and that aren’t committed to qualia) are functionalist.
166
Ned Block
They are functionalist in that they involve the idea that narrow intentional content supervenes on internal functional organization as well as physico-chemical configuration. That is, there can be no differences in narrow intentional contents without corresponding differences at the level of causal interactions of mental states within the head. The view comes in different flavors: Functional organization can be understood in terms of the interactions of common-sense mental states or in terms of the causal network of computational states. In both cases, there is a level of “grain” below which brain differences make no difference. One functional organization is multiply realizable physico-chemically in ways that make no difference in narrow intentional content. In other words, there is a level of organization above the level of physiology (“mental” or “computational”) that determines narrow intentional content. (Tye 1994 takes this view and I understand Rey 1992a,b and White 1995 as endorsing it.) Of course phenomenists can (and should) be internalists about phenomenal character too. But phenomenists can allow that phenomenal character depends on the details of the physiology or physico-chemical realization of the computational structure of the brain. Of course, there are also dualist forms of phenomenism, but both the physicalist and dualist forms of phenomenism agree that there is no need to suppose that qualia supervene on functional organization. There is a very simple thought experiment that raises a serious (maybe fatal) difficulty for any such (functionalist) internalist form of representationism. Suppose that we raise a child (call her Erisa) in a room in which all colored surfaces change color every few minutes. Further, Erisa is allowed no information about grass being green or the sky being blue, and so on. The result is that Erisa ends up with no standing beliefs that distinguish one color from another. Suppose further that she is not taught color words, nor does she make them up for herself. (There are many languages that have only two words for colors, for example the language of the Dani famously studied by Rosch 1972.) Now we may suppose that the result is that there is little in the way of abiding functional differences among her color experiences. Most important, Erisa has no associations or behavioral inclinations or dispositions toward red that are any different from her associations or inclinations or dispositions toward blue. Of course, she responds to color similarities and differences—she groups blue things together as having the same color, and she groups red things together as having the same color. But her ability to group under the same color relation does not distinguish her reaction to red from her reaction to blue. The experience as of red is vividly different from the experience as of blue. But what difference in function in this case could plausibly constitute that difference? The challenge to the internalist representationist, then, is to say what the difference is in internal intentional content between the experience of red and the experience of blue. The only resources available to the internalist representationist are functional. There is a
Mental Paint
167
difference in phenomenal character, so the internalist representationist is committed to finding a difference in function. But the example is designed to remove all abiding differences in function. The functionalist can appeal to temporary differences. Erisa will say “The wall is now the same color that adorned the table a second ago,” and “For one second, the floor matched the sofa.” But these beliefs are fleeting, so how can they constitute the abiding differences between the phenomenal character of her experience of red and green? The differences between these phenomenal characters stay the same (for us) from moment to moment, day to day, and there is no reason to suppose that the same cannot be true for Erisa. The point of the thought experiment is to make it plausible that color experiences can remain just as vivid and the differences between them just as permanent as they are for us even if the functional differences between them attenuate to nothing that could plausibly constitute those differences. Of course, there is one abiding difference in functional role between the experience of red and the experience of green—the properties of the stimuli. Since we are talking about internalist representationism, the stimuli will have to be, for example, light hitting the retina rather than colored surfaces. But these differences in the stimuli are what cause the differences in the phenomenal character of experience, not what constitute those phenomenal differences. I don’t expect diehard functionalists to recant in response to this point, but I really don’t see how anyone with an open mind could take being caused by certain stimuli as constituting phenomenal characters of color experiences. Of course, there may be innate behavioral differences between the experience of red and the experience of blue. Perhaps we are genetically programmed so that red makes us nervous and blue makes us calm. In my view it is a bit silly to suppose that the phenomenal character of the experience as of blue is constituted by such factors as causing calmness. But since so many philosophers feel otherwise, I will also note that despite claims of this sort (Dennett 1991), such assertions are empirical speculations. (See Dennett’s only cited source, Humphrey 1992, which emphasizes the poor quality of the empirical evidence.) And of course anyone who holds that representationism is a conceptual truth will be frustrated by the fact that we don’t know without empirical investigation whether it is a truth at all. Perhaps the internalist will say that there would be no differences among Erisa’s experiences. Red would of necessity look just the same to her as yellow. But this is surely an extraordinary thing for the internalist to insist on. It could be right of course, but again it is surely an unsupported empirical speculation. I claim that an Erisa of the sort I described is conceptually possible. The replies I have just considered do not dispute this, but instead appeal to unsupported empirical claims.
168
Ned Block
But physicalism is an empirical thesis, so why should the representationist be embarrassed about making an empirical claim? Answer: Physicalism is a very general empirical thesis having to do with the long history of successes of what can be regarded as the physicalist research program. Internalist representationism, by contrast, depends on highly specific experimental claims. For example, perhaps it will be discovered that newborn babies hate some colors and love others before having had any differential experience with these colors. I doubt that very many opponents of qualia would wish their point of view to rest on speculations as to the results of such experiments. The defender of qualia does not depend on a prediction about the results of such experiments. Our view is that even if such experiments do show some asymmetries, there are possible creatures—maybe genetically engineered versions of humans—in whom the asymmetries are ironed out. (See Shoemaker 1982.) And those genetically engineered humans could nonetheless have color experience much like ours. Any view that rests an argument against qualia on the existence of such asymmetries would seem to be committed to the peculiar claim that qualia are possible in creatures that differ from us in ways so subtle that we are not now sure whether we are these possible creatures. Further, given the necessity of identity, if phenomenal characters are identical to internal functional states, then there are no creatures, even possible creatures, who have different phenomenal characters that do not differ functionally. Externalism That is all I will have to say about internalist representationism. Now I will move to the main topic of this paper, externalist representationism. In this section, I will try to motivate externalist representationism. The consideration I will advance in motivating it will be of use to me later. Then I will advance various considerations that cause one or another sort of difficulty for the view. Often, when I see water I see it as water; that is, my visual experience represents it as water. Most of you have seen water as water all your lives, but I’m different, I’m a foreigner. I was born on Twin Earth and emigrated to Earth at age 18. When I was 15 and looked at the sea on Twin Earth, my visual experience represented the twin-water as twinwater (though of course we didn’t call it that). Perhaps you are skeptical about whether visual experience represents such properties, but please bear with me. When I first got here, I saw water as twin-water, just as you, if you went to Twin Earth, would see the twin-water in the oceans as water. Now, many years later, my practices of applying concepts are relevantly the same as yours: My practices show that I am committed to the concepts of my adopted home, Earth. Now when I look at the sea of my adopted home, my visual experience represents the water as water just as yours does. (This is controversial
Mental Paint
169
but again, let’s suppose.) So the representational content of my experience of looking at the sea has changed. But my visual experience is nonetheless indistinguishable from what it was. If you took me up in a space ship and put me down by some seaside, I wouldn’t know whether I was on Earth or Twin Earth. (See Stalnaker’s 1996 commentary on Lycan 1996c on Block 1990.) The representational content of my experience has changed, but the phenomenal character has stayed the same. And that shows, someone (but not me) might argue, that there is some sort of gap between representational content and phenomenal character. (I think that there is such a gap, but this case doesn’t reveal it.) One way for the representationist to answer would be to note that though all phenomenal character is (according to representationism) representational content, the converse is not true. Not all representational contents are phenomenal characters. For example, the thought content that mumesons have odd spin is not a phenomenal character. But how is this point supposed to apply to the issue at hand? The putative gap between representational content and phenomenal character has to do with the representational content of visual experience, not thought. Here is one way of extending the representationist response. The representationist should hold that visual experience has two kinds of representational content: One kind of representational content can be identified with phenomenal character, and another kind of representational content is distinct from phenomenal character. The (relevant) phenomenal character of experience that has remained the same throughout my life is a matter of the observable or appearance properties of the liquids in oceans, rivers, and streams, their color, sheen, motion, taste, smell, and so on. It is these properties that my experience has represented liquids as having that have been shared by my experiences of both twin-water and water. Those are the representational contents that make my visual experience of the Twin Earth ocean at age 15 indistinguishable from my visual experience of the ocean now. (Many representationists go further, saying that these observational representational contents are nonconceptual, but nothing I will be saying here will depend on whether or not this is so.) Of course, as Burge (1979 and elsewhere) has noted, color concepts and other “appearance concepts” are vulnerable to the same sort of Twin Earth arguments for externalism as is the concept of water. (Perhaps only truly phenomenal concepts are invulnerable to such arguments—as is suggested by Burge’s 1979 remark that in his thought experiment, phenomenal character does not change.) Further, as I will argue later in the essay, a sort of Twin Earth argument can be run on perceptual contents whether or not they are conceptual. So in my view, the reply just discussed will ultimately fail to sustain the representationist point of view. But for the moment, we can take the representationist reply just given as a success for representationism. We started with an experiential continuity despite representational change that challenged the representationist. But then it turned out that the representationist can give a representational explanation of the difference. The main
170
Ned Block
burden of this essay is to explore some reasons for thinking that the representationist cannot always repeat this success. Two preliminary issues: First, one might object to my claim that visual experience can ever represent anything as water. Perhaps, you might say, visual experience can represent something as round or at least red, but not as water. I use the notion of visual experience representing water as a way of setting up the issue, but it does not play any essential role in my argument. Further, it should be noted that disallowing visually representing water as water has the effect of limiting the resources available to the representationist. Next, a terminological reminder. I take a qualitative character or quale as a phenomenal property of an experience that eludes the intentional, the functional, and the purely cognitive. “Phenomenal character” is a more neutral term that carries no commitment to qualia. Both the representationist and the phenomenist can agree that there are phenomenal characters, even though the former but not the latter thinks phenomenal characters are wholly representational. Supervenience If phenomenal character supervenes on the brain, there is a straightforward argument against representationism. For arguably, there are brains that are the same as yours in internal respects (molecular constitution) but whose states represent nothing at all. Consider the Swampman, the molecular duplicate of you who comes together by chance from particles in the swamp. He is an example of total lack of representational content. How can he refer to Newt Gingrich when he has never had any causal connection with him; he hasn’t ever seen Newt or anyone else on TV, never seen a newspaper, and so on. But although he has no past, he does have a future. If at the instant of creation, his brain is in the same configuration that yours is in when you see a ripe tomato, it might be said that the Swampman’s state has the same representational content as yours in virtue of that state’s role in allowing him to “track” things in the future that are the same in various respects (e.g., color) as the tomato. It might be said that he will track appearances of Bill Clinton in the future just as well as you will. But this invocation of the Swampman’s future is not very convincing. Sure, he will track Bill Clinton if he materializes in the “right” environment (an environment in which Bill Clinton exists and has the superficial appearance of the actual Bill Clinton), but in the wrong environment he will track someone else or no one at all. And the same point applies to his ability to track water and even color. If put on Twin Earth, he will track twin-water, not water. Of course, you will have the same tracking problems if suddenly put in “wrong” environments, but your references are grounded by the very past that you have and the Swampman lacks.
Mental Paint
171
If we can assume supervenience of phenomenal character on the brain, we can refute the representationist. The phenomenal character of the Swampman’s experience is the same as yours but its experiences have no representational content at all. So phenomenal character cannot be representational content. And the case of two different Swampmen who have different phenomenal characters but the same (null) representational content shows phenomenal character does not even supervene on representational content. If this point is right, there will be a great temptation for the representationist to deny supervenience of phenomenal character on the brain. And in fact, that’s what representationists often do (Dretske 1995; Lycan 1996a,b; McDowell 1994). This seems to me to be a desperate maneuver with no independent plausibility. The independent arguments against representationism to follow constitute a partial justification of supervenience, since if representationism were true, supervenience would be false. What Is the Issue? One of the ways that I have been framing the issue is this: Is there more to experience than its representational content? But if experiences are brain states, there will be more to experiences than their representational contents just as there is more to sentences in English than their representational contents. For example, the size of the font of this sentence is something more than (or anyway, other than) its representational content. So the question is better taken as: Is there anything mental in experience over and above its representational content? I say yes, the representationist says no. Harman (1990, 1996) expresses his version of representationism about experience by claiming that in experience we are aware only of properties of what is represented, not the vehicle of representation. When we look at a red tomato, no matter how hard we try to introspect the aspect of the experience that represents redness, all we succeed in doing is focusing our attention on the redness of the tomato itself. Harman relies on the diaphanousness of perception (Moore 1922), which may be defined as the claim that the effect of concentrating on experience is simply to attend to and be aware of what the experience is of. As a point about attention in one familiar circumstance—for example, looking at a red tomato—this is certainly right. The more one concentrates on the experience, the more one attends to the redness of the tomato itself. But attention and awareness are distinct, and as a point about awareness, the diaphanousness claim is both straightforwardly wrong and misleading. One can be aware of what one is not attending to. For example, one might be involved in intense conversation while a jackhammer outside causes one to raise one’s voice without ever noticing or attending to the noise until someone comments on it—at which time one realizes that one was aware of it all along. Or consider the familiar experience of noticing that the refrigerator compressor has gone off and that one was
172
Ned Block
aware of it for some time, even though one didn’t attend to it until it stopped. (These and other examples of attention without awareness are discussed in Block 1995a, and in Tyler Burge’s reply in this volume. I am grateful to Burge for discussion of this issue.) So Harman is wrong about awareness, even if he is right about attention. Further, though Harman is right about attention in one common circumstance, I believe he is wrong about attention in other circumstances. For example, close your eyes in daylight and you may find that it is easy to attend to aspects of your experience. If all experiences that have visual phenomenology were of the sort one gets with one’s eyes closed while awake in daylight, I doubt that the thesis that one cannot attend to or be aware of one’s experience would be so popular. Another way to appreciate the point: Stick your finger in front of your face, then focus on something distant. It does not seem so hard (at least to me) to attend to and be aware of aspects of the experience of the finger as well as of the finger. The points just made are preliminary. They are just appeals to introspection—mine, and I hope yours—that go against representationism. If you agree, you are ready to reject representationism. But the main purpose of this section is to get clearer on what exactly you would be rejecting. Harman can allow that there are many properties of experience that we are aware of. For example, we know when an experience happens and how long it lasted. But that will no doubt be glossed by him as a matter of when we looked at the tomato and how long we looked. Harman concludes that introspection gives us no access to anything “nonrepresentational” about the experience, nothing whose identity is not given representationally, no access to mental paint. He argues that the contrary view confuses properties of what is represented, the intentional object of perception, with properties of the vehicle of representation, what is doing the representing. I don’t agree with the imputation of fallacy. But my point right now is more preliminary: that if Harman means to define representationism in this way, his definition is too narrow. Harman writes as if the issue is whether we can introspect the representational features of the experience, the mental paint that represents the redness of the tomato. But there are two deeper issues. 1. The first is whether there is mental paint, even if Harman were right that we cannot become aware of it when we are seeing a tomato. One way—not the only way—of seeing why there is a real issue here is to consider the idea that the possibility of an inverted spectrum shows that there is more to experience than its representational content. According to this argument (Shoemaker 1982; Block 1990; see also Block and Fodor 1972), your experience and my experience could have exactly the same representational content, say, as of red, but your experience could have the same phenomenal character as my experi-
Mental Paint
173
ence as of green. Shoemaker (1994a,b) agrees with Harman’s views about introspection. He agrees that we cannot be aware of mental paint, that when we try to introspect the experience of the redness of the tomato, all we succeed in doing is attending to and being aware of the represented color of the tomato itself. But according to Shoemaker, we have a kind of indirect introspective access to the phenomenal character of the experience via our intuitions about the inverted spectrum. By imagining that things we both call red look to you the same way that things we both call green look to me, we succeed in gaining indirect introspective access to mental paint. Thus there is mental paint. So Shoemaker’s view gives us an example of how on certain assumptions it would be reasonable to think that there is mental paint even if we can’t be directly aware of it, and that is one way of illustrating the distinction. Shoemaker’s view is highly paradoxical—there is mental paint but it is a theoretical entity. It is of the essence of mental paint to be something of which we are aware. This view would be less paradoxical if Shoemaker’s “phenomenal properties” of objects were really just phenomenal characters of experiences projected onto objects. The phenomenal character of an experience determines what phenomenal property it represents, and the phenomenal property is individuated in accord with and gets its identity from the phenomenal character that determines it. Thus far, Shoemaker’s phenomenal properties look like mere projections of phenomenal characters onto the world. But there is one crucial feature of Shoemaker’s view that resists this attempt to blunt the paradox, namely that Shoemaker takes phenomenal properties to be causally efficacious, indeed to cause the instantiation of the phenomenal characters that help to individuate them. Shoemaker’s view is discussed in more detail in note 2. 2. A second issue is whether there are phenomenal features of experience that are not even vehicles of representation. For example, according to me, the phenomenal character of the experience of orgasm is partly nonrepresentational. Such a nonrepresentational mental feature would be (in this respect) like the oil in oil-based paint. So we could put the two issues as whether there is mental paint and whether there is mental oil. To sum up then, we can distinguish three things: 1. The intentional content of an experience. I am currently looking at a tomato and my experience represents the tomato as red. 2. Mental properties of the experience that represent the redness of the tomato. This is mental paint. According to me, the phenomenal character of the experience is such a mental property: It represents the tomato as red. According to me, one can attend to this phenomenal character and be aware of it even when one is not attending to it. Representationists would deny both.
174
Ned Block
3. Mental properties of the experience that don’t represent anything. This is mental oil. I don’t know whether there are any such properties in the case of a normal experience of a red tomato, but I do claim that such properties are involved in orgasm-experience, pain, and other bodily sensations. These distinctions allow us to see an ambiguity in ‘nonrepresentational’. In the most straightforward sense, if as I claim there are phenomenal properties of orgasm-experience that don’t represent anything (the ones that make orgasm experience something one wants to have), then those properties are nonrepresentational. But there is also a weaker sense: The phenomenal character of color experience, for example, could be said to be nonrepresentational in that the identity of that phenomenal character is not given by its representational content. In this weak sense, we can agree that color experience has representational content while at the same time regarding it as nonrepresentational because that representational content is not the whole essence of the experience. As I mentioned, Harman says that we are aware only of what is represented by our experience, the intentional object of the experience, not what is doing the representing, not the vehicle of representation. But what will Harman say about illusions, cases where the intentional object does not exist? Surely, there can be something in common to a veridical experience of a red tomato and a hallucination of a red tomato, and what is in common can be introspectible. This introspectible commonality cannot be constituted by or explained by the resemblance between something and nothing. It would be better for the representationist to say that what is in common is an intentional content, not an intentional object. Disjunctivists like McDowell deny that there is anything introspectible in common. But how can they understand the perceptual situations in which one can be reliably fooled, in which one has no idea whether the perception is veridical or not? “It only seems that there is something perceptual in common.” But why doesn’t whatever mental aspect that grounds the seeming constitute the phenomenal similarity? On the face of it, the disjunctivist has a liability beyond those of the representationist. The representationist can appeal to the intentional content that is shared by the two experiences, the content that there is a red tomato in front of me. Suppose Harman were to hold that we are aware of the shared intentional content; what would that come to? What is it to be aware of the intentional content that I am seeing a red tomato or that two experiences have that intentional content? I don’t see what awareness of an intentional content could come to if not awareness that some state has that intentional content. And, one might speculate, awareness that two experiences have the same intentional content requires awareness that each has that intentional content. So if Harman were to give this representationist account of the introspectible similarity, he would have to concede that we have introspective awareness of some mental properties of experience, not just of the intentional objects of experience.
Mental Paint
175
The representationist view I’ve just mentioned is taken by Lycan (1995)—that one can be aware of a family of mental properties of an experience, namely that the experience represents something, that it represents a tomato, that it represents the tomato as red, and so on. So Lycan can deal with illusions by saying that when one hallucinates a red tomato, there is something introspective in common with a normal veridical perception of a tomato, namely in both cases one is aware that the experience represents a red tomato.2 Though Lycan’s position does accommodate common sense better than the disjunctivist, it is nonetheless implausible. A child who has no concept of representation or of intentional content can be aware of what is in common to two experiences that represent that a tomato is red. Some representationists combine externalism and internalism. For example, Rey (1992a,b) individuates color experience partly in terms of what colors it represents and partly in terms of what he sees as syntactic properties of the vehicle of representation. (There is a similar view in Lycan 1996b.) I won’t try to consider such mixed views here. Those who deny both mental paint and oil are representationists; those who countenance one or the other are phenomenists. The representationists include Byrne and Hilbert (1997a), Dretske (1995, 1996), Harman (1990, 1996), Lycan (1995, 1996), McDowell (1994), Rey (1992a,b) and Tye (1995, 1996). (See also White 1995 for a representationist view of color experience.) The phenomenists include Burge (1997), Block (1990, 1994a), Loar (1990), McGinn (1991), Peacocke (1983), and Shoemaker (1982, 1994a,b). (Shoemaker’s view combines aspects of both representationism and phenomenism, though I count him as a phenomenist here.)3 Shoemaker and I hold that the Inverted Spectrum argument and the Inverted Earth argument make strong cases for phenomenism. (Loar, McGinn, and Peacocke have declared doubts about the Inverted Spectrum arguments; I don’t know what they think of the Inverted Earth argument.) I won’t go into the Inverted Spectrum argument here, and I will only be able to mention Inverted Earth briefly. I will mention some other considerations that are less effective, but which I hope put some pressure on the representationist. Bodily Sensations Is the experience of orgasm completely captured by a representational content that there is an orgasm? Orgasm is phenomenally impressive and there is nothing very impressive about the representational content that there is an orgasm. I just expressed it and you just understood it, and nothing phenomenally impressive happened (at least not on my end). I can have an experience whose content is that my partner is having an orgasm without my experience being phenomenally impressive. In response to my raising this issue (Block 1995a,b), Tye (1995a) says that the representational content of orgasm “in part, is that
176
Ned Block
something very pleasing is happening down there. One also experiences the pleasingness alternately increasing and diminishing in its intensity.” But once again, I can have an experience whose representational content is that my partner is having a very pleasing experience down there that changes in intensity, and although that may be pleasurable for me, it is not pleasurable in the phenomenally impressive way that that graces my own orgasms. I vastly prefer my own orgasms to those of others, and this preference is based on a major league phenomenal difference. The location of “down there” differs slightly between my perception of your orgasms and my own orgasms, but how can the representationist explain why a small difference in represented location should matter so much? Of course, which subject the orgasm is ascribed to is itself a representational matter. But is that the difference between my having the experience and my perceiving yours? Is the difference just that my experience ascribes the pleasure to you rather than to me (or to part of me)? Representational content can go awry in the heat of the moment. What if in a heated state in which cognitive function is greatly reduced, I mistakenly ascribe your orgasm to me or mine to you? Would this difference in ascription really constitute the difference between the presence or absence of the phenomenally impressive quality? Perhaps your answer is that there is a way in which my orgasm-experience ascribes the orgasm to me that is immune to the intrusion of thought, so there is no possibility of a confused attribution to you in that way. But now I begin to wonder whether this talk of “way” is closet phenomenism. No doubt there are functional differences between my having an orgasm-experience and merely ascribing it to you. Whether this fact will help to defend representationism depends on whether and how representationism goes beyond functionalism, a matter to be discussed in the section after next. Lycan (1996c) appeals to the following representational properties (of male orgasm): It is “ascribed to a region of one’s own body,” and the represented properties include “at least warmth, squeezing, throbbing, pumping and voiding. (On some psychosemantics, I suppose, impregnating is represented as well)” (p. 136). Lycan says that it is “impracticable to try to capture detailed perceptual content in ordinary English words, at least in anything like real time,” but he thinks he has said enough to “remove reasonable suspicion that there are nonintentional qualitative features left over in addition to the functional properties that are already considered characteristic of the sensation.” But Lycan’s list of properties represented seems to me to increase the suspicion rather than removing it. Everything that matters (phenomenally speaking) is left over. According to me, there are features of the experience of orgasm that don’t represent anything; so mental oil exists. I don’t expect this example to force representationists to concede that mental oil exists. Appeals to intuitions about relatively unstructured cases are rarely successful. That is why the complex thought experiments such as those involving the inverted spectrum and inverted earth are useful. I believe that complex cases can be
Mental Paint
177
framed that reflect our concepts so directly that everyone can be brought to agree with them. The argument can leave the intuitions behind and move to the philosophical question of what they show. Of course, we should not demand that a representationist be able to capture his contents in words. But if we are to try to believe that the experience of orgasm is nothing over and above its representational content, we need to be told something fairly concrete about what that representational content is. Suppose the representational content is specified in terms of recognitional dispositions or capacities. One problem with this suggestion is that the experience of orgasm seems on the face of it to have little to do with recognizing orgasms. Perhaps when I say to myself “There’s that orgasm experience again” I have a somewhat different experience from the cases where no recognition goes on. But there is no plausibility in the insistence that the experience must involve some sort of categorization. And if you are inclined to be very intellectual about human experience, think of animals. Perhaps animals have the experience without any recognition. The representationists should put up or shut up. The burden of proof is on them to say what the representational content of experiences such as orgasm is. Phosphene-Experiences Harman (1990) says “Look at a tree and try to turn your attention to intrinsic features of your visual experience. I predict you will find that the only features there to turn your attention to will be features of the represented tree.” But the diaphanousness of perception is much less pronounced in a number of visual phenomena, notably phosphene-experiences. (I use the cumbersome ‘phosphene-experience’ instead of the simpler ‘phosphene’ by way of emphasizing that the phenomenist need not have any commitment to phenomenal individuals.) If all of our visual experiences were like these, representationism would have been less attractive. Phosphene-experiences are visual sensations “of” color and light stimulated by pressure on the eye or by electrical or magnetic fields. (I once saw an ad for goggles that you could put on your eyes that generated phosphenes via a magnetic field.) Phosphene-experiences have been extensively studied, originally in the nineteenth century by Purkinje and Helmholz. Close your eyes and place the heels of your hands over your eyes. Push your eyeballs lightly for about a minute. You will have color sensations. Can you attend to those sensations? I believe I can. Even if you can’t attend to them, are you aware of them? According to the representationist, all awareness of those sensations could consist in is awareness of the colored moving expanses that are represented by them. My view is that one can be aware of something more. Again, I don’t think this sort of consideration can change anyone’s mind, but I hope it will have an impact on the noncommitted.
178
Ned Block
Lycan (1987) says: “given any visual experience, it seems to me, there is some technological means of producing a veridical qualitative equivalent—e.g. a psychedelic movie shown to the subject in a small theater” (p. 90). But there is no guarantee that phospheneexperiences produced by pressure or electromagnetic stimulation could be produced by light. (Note I don’t say there is a guarantee that phosphene-experiences could not be produced by light, but only that there is no guarantee that they could; I have no idea whether they could or not.) I do wonder if Lycan’s unwarranted assumption plays a role in leading philosophers to suppose that the phenomenal characters of phosphene-experiences, afterimage-experiences, and the like are exhausted by their representational content. Bach-y-Rita According to me, in normal perception one can be aware of the mental paint—the sensory quality that does the representing. This idea can be illustrated (this is more of an illustration than it is an argument) by Bach-y-Rita’s famous experiment in which he gave blind people a kind of vision by hooking up a TV camera that produced tactual sensations on their backs. Bach-y-Rita says that the subjects would normally attend to what they were “seeing.” He says “unless specifically asked, experienced subjects are not attending to the sensation of stimulation on the skin of their back, although this can be recalled and experienced in retrospect” (quoted in Humphrey 1992, p. 80). The retrospective attention of which Bach-y-Rita speaks is a matter of attending in retrospect to a feature of one’s experience that one was aware of but not attending to when the perception originally happened, as with the jackhammer and refrigerator examples mentioned earlier. Of course, the analogy is not perfect. In attending to visual sensations, we are not normally attending to sensations of the eye (Harman 1996). I think that the Bach-y-Rita experiment is useful in thinking about the two versions of representationism mentioned above. Let me remind you about the difference. Harman seems to say that all we can introspect in experience are the intentional objects of experience. Lycan, however, allows that we can actually introspect certain properties of the experiences themselves. At first glance, reflection on the Bach-y-Rita experiment provides support for Lycan over Harman. For the ability of Bach-y-Rita’s subjects to introspect their tactual sensations helps to remind us that we really can notice features of our own visual sensations. But Lycan’s concession, you will recall, was to allow introspection of the property of having certain intentional properties. But is that what Bach-y-Rita’s subjects were doing? Were they introspecting that the sensations on their backs represented, say, a couch? Perhaps occasionally, but I doubt that that’s what Bach-y-Rita was talking about. I think he meant that they were attending to the experiential quality of the feelings on their backs. And I think that this case helps to remind us that at least sometimes when we intro-
Mental Paint
179
spect visual experience, we are attending to the phenomenal properties of experience, not the fact that they have certain intentional properties. So although Lycan’s version of representationism is superior to Harman’s in allowing the existence of introspection of something other than intentional properties of experience, it is not true to what that introspection is often like.4 Moving now to the orgasm and pain cases, as I mentioned earlier, there is a challenge here for the representationist. Just what is the representational content of these states? In vision, it often is plausible to appeal to recognitional dispositions in cases where we lack the relevant words. What’s the difference between the representational contents of the experience of color A and color B, neither of which has a name? As I mentioned earlier, one representationist answer is this: The recognitional dispositions themselves provide or are the basis of these contents. My experience represents A as that color, and I can misrepresent some other color as that color. But note that this model can’t straightforwardly be applied to pain. Suppose I have two pains that are the same in intensity, location, and anything else that language can get a handle on—but they still feel different. Say they are both twinges that I have had before, but they aren’t burning or sharp or throbbing. “There’s that one again; and there’s that other one” is the best I can do. If we rely on my ability to pick out that pain, (arguably) we are demonstrating a phenomenal character, not specifying a representational content. (Note the difference between Loar’s 1990 proposal of a recognitional view of phenomenal concepts and the current suggestion that a recognitional disposition can specify phenomenal character itself. Phenomenal character is what a phenomenal concept specifies or refers to.) The appeal to recognitional dispositions to fill in representational contents that can’t be specified in words has some plausibility, so long as the recognitional dispositions are directed outward. But once we direct them inward, one begins to wonder whether the resulting view is an articulation of representationism or a capitulation to phenomenism. I will return to this point.5 Is Representationism Just a Form of Functionalism? Consider what I call quasi-representationism. Quasi-representationists agree with phenomenists that there are differences between sensory modalities that cannot be cashed out representationally. One modality is flashing lights, another is tooting horns. But quasirepresentationists agree with representationists that within a single modality, all phenomenal differences are representational differences. (I think that this is the view that Peacocke 1983, ch. 1 argues against.) Some philosophers are attracted to representationism but can’t bring themselves to treat the experiential differences between, say, vision and touch as entirely representational. So they appeal to the fact that visual and touch representations of, say, an edge, function
180
Ned Block
differently. They plug a gap in representationism with functionalism. (Block 1996 and Robinson 1998 interpret Lycan 1996c in this way.) But they should tell us why they don’t reject representationism altogether in favor of functionalism. Some philosophers start with functionalism, but don’t see how to handle afterimages (and perhaps other putatively intentional phenomena) functionally (Lycan 1996a,b). So they add representationism (Lycan 1996c). Perhaps both doctrines are wrong by themselves, but there is a third that draws on the resources of both that works. Many philosophers in the representationist ballpark are rather vague about whether they are pure representationists or quasi-representationists, but Tye (1995b) makes it clear that he is a pure representationist. (Harman tells me he is a quasi-representationist and Lycan 1996c declares quasi-representationism.) How can we decide whether the antiphenomenist needs both representationism and functionalism? Suppose I both touch and see a dog. Both experiences represent the dog as a dog, but they are different phenomenally. Representationists are quick to note that the two experiences also differ in all sorts of other representational ways. (See Tye 1995a, for example.) The visual experience represents the dog as having a certain color, whereas the tactual experience represents it as having a certain texture and temperature. In Block (1995a,b) I tried to avoid this type of rejoinder by picking experiences with very limited representational content. If you wave your hand in the vicinity of your ear, your peripheral vision gives you an awareness of movement without size, shape, or color. You have a visual experience that plausibly represents something moving over there and nothing else. And I imagined that there were auditory experiences with the same content. But my expert consultants tell me that I was wrong. There is no auditory analogue of peripheral vision. For example, any auditory experience will represent a sound as having a certain loudness. But that does not ruin the point. It just makes it slightly harder to see. Imagine the experience of hearing something and seeing it in your peripheral vision. It is true that you experience the sound as having a certain loudness, but can’t we abstract away from that, concentrating on the perceived location? And isn’t there an obvious difference between the auditory experience as of that location and the visual experience as of that location? If so, then there is either mental paint or mental oil. (The ways in which representationally identical experiences might be phenomenally different could involve differences in either paint or oil.)6 Seeing Red for the First Time Marvin is raised in a black and white room, never seeing anything of any other color. Further, as with Erisa, he never learns that fire engines and ripe tomatoes are red or that grass is green, and so on. Then he is taken outside and shown something red without being
Mental Paint
181
told that it is red. (I’ve changed ‘Mary’ to ‘Marvin’ so as to emphasize the small differences between this and Jackson’s 1982 argument.) He learns what it is like to see red, even though, unlike Mary, he is not told what that color is called. He might say: “So that’s what it is like to see that color.” Lewis (1990) (following Nemirow) says that Marvin aquires an ability, some sort of recognitional know-how. But as Loar (1990) notes, this idea can’t account for embedded judgments. Here’s an example that fits the Marvin case: “If that’s what it is like to see red, then I will be surprised.” (Loar is applying a standard argument against nondescriptivism in ethics. Perhaps followers of Lewis and Nemirow will try to utilize nondescriptivist attempts to deal with the problem by Blackburn 1993 and Gibbard 1990.) What does the representationist say about what Marvin has learned? If Marvin is told that what he sees is red, the representationist might say that he has acquired a visual representational concept, the concept of red. But can the representationist say this if Marvin doesn’t know that it is red? Perhaps the representationist will say this: He acquires the concept of red without the name ‘red’. What Marvin acquires is a recognitional concept. After all, he can say “There’s that color again.” He has a recognitional concept that he applies on the basis of vision, even though it doesn’t link up to his linguistic color concepts. But there is a trap for the representationist in this reply. For what, according to the representationist, is the difference between Marvin’s concept of red and Marvin’s concept of blue? He recognizes both. When he sees a red patch he says “There’s that color again” and when then sees a blue patch he says “There’s that other color again,” each time collating his outer ostension with an inner ostension. But what, according to the representationist, is the difference between Marvin’s concept of red and Marvin’s concept of blue? The phenomenist will link the difference to an internal difference, the difference in the phenomenal qualities of the experience of blue and the experience of red. But the representationist can’t appeal to that without changing sides. What else is there of a suitably internal sort for the representationist to appeal to? (Remember, we are supposing, as with Erisa raised in the room of changing colors, that Marvin knows nothing that distinguishes the unnamed colors, no abiding beliefs to the effect that this color has certain properties that that color doesn’t have.) The appeal to a recognitional disposition suggests that the representationist is appealing to the colors themselves. What makes his concept of red different from his concept of blue is that he applies the former in recognizing red and the latter in recognizing blue. In a response to an earlier version of this objection to representationism (Block 1996), Lycan (1996c, p. 137) gives an answer to my question about what the difference is between Marvin’s concept of red and Marvin’s concept of blue. “Answer: that the former
182
Ned Block
represents the (or a) physical property of objective redness while the latter represents objective blueness.” I believe that this is an important concession on the part of the externalist representationists because it eases the way for a variety of thought experiment arguments, some of which will be considered later in this essay. (Here’s one that won’t be considered later: Imagine that Marvin has somewhat different color vision in one eye from the other. Can’t he have phenomenally different experiences of the same color? Here’s another: Suppose a Cartesian demon fools Marvin. His two concepts that he takes to pick out different colors actually pick out the same color. But surely the concepts are different independently of what they pick out!) But even independently of such thought experiments, if the representationist is willing to recognize a color concept that has been cut loose from everything but recognition, why shouldn’t he also recognize such concepts turned inward? Why can’t he have a recognitional concept of his own phenomenal state— “There’s that experience again.” And that would just be a phenomenal concept. (I don’t mean to imply that I think it would be natural for Marvin to invoke such a highly sophisticated concept in recognizing that he is seeing the same color again. Phenomenal concepts, that is concepts of phenomenal characters, are highly sophisticated.) Rey (1992a,b) postulates that color experiences involve the tokenings of special restricted predicates in the language of thought. So he would say that Marvin tokens ‘R’ when he sees red and ‘B’ when he sees blue. Is that a suitable representationist answer? Recall that we are now discussing externalist representationism, and Rey’s view would deal with this problem by bringing in an internalist element. Recall my objection to internalism in terms of Erisa, the girl raised in the room in which everything changes colors. Erisa perhaps has more or less normal color experience but may have no abiding asymmetrical associations in her color experience. So what’s the representational difference between her ‘R’ and ‘B’? Suppose Rey says: “who needs a representational difference; the syntactic difference is enough.” Then it becomes difficult to see why he is not just an oldfashioned eliminativist. Let me explain. I am a reductionist, a physicalist phenomenist. I believe that the difference between Marvin’s experience of red and of blue is a physical difference. I suspect it is a difference in brain events that is not naturally capturable in terms of “syntax,” since talk of syntax in connection with brain processes will most likely apply, if anywhere, to language areas of the brain rather than to visual areas, but I am flexible on this issue. Certainly no position on the syntax issue is an important feature of the physicalist-phenomenist position. I allow that it is an empirical question (of course) what the physical natures of phenomenal qualities are. Perhaps these natures are syntactical or even functional (psychofunctional, that is, in the terminology of Block 1978). In sum, no doubt the internalist antiphenomenist can find some physicalistic surrogate for qualia, but a suitable surrogate will be the very item that the physicalist phenomenist will think of qualia as reduced to. And then the difference will be like the difference between the reduc-
Mental Paint
183
tionist who says that there is water and it is H2O and the eliminativist who says there is H2O but no water. And eliminativism about phenomenal character is even less plausible than eliminativism about water. Jackson uses his “Mary” case to argue against physicalism. Mary knows all the physicofunctional facts, but nonetheless learns a new fact, so there are facts that aren’t among the physico-functional facts. I accept the familiar refutation of this argument along the lines of: Mary learns a new concept of something she already knew. She acquires a phenomenal concept of a physical fact that she was already acquainted with via a physical concept (Loar 1990). The point of this section is that an adaptation of Jackson’s case is more effective in setting up an argument against representationism than it is against physicalism. Externalist Memory At the beginning of the discussion of externalism, I discussed the thought experiment in which I, a native of Putnam’s Twin Earth, emigrated to Earth. When I first looked at water, I thought it was twin-water and, we are agreeing to suppose, my visual experience represented it as twin-water. Much later, after learning everything that I’ve just told you, my allegiances shifted; in effect, I decided to become a member of the Earth language community, to speak English, not Twenglish. Now, when I look at the sea, I take what I am seeing to be water and my visual experience represents it as water (let’s suppose), not twin-water. (How could a conceptual change affect a visual representation? Though nothing here depends on the matter, I will briefly mention the rationale. There is a difference between seeing a group of buildings as a hospital and seeing them as a nuclear reactor. These are conceptual differences but they make a visual difference. Seeing as is both conceptual and visual.) Though my conceptual and visual representation of water has changed during my stay on earth, in some very obvious sense, water looks the same to me as it did the first time I saw it even though my representational content has changed. If you blindfolded me and put me down at the seaside, I would not know from looking at the liquid in the ocean whether it was water or twin-water. My phenomenal character has stayed the same even though the representational content of my visual experience changed. But this doesn’t yet show that there is anything nonrepresentational about phenomenal character. For the shared representational contents can be appealed to to explain why there’s no difference in what it’s like to see water. Here is the representationist picture: Experiences have representational properties of two types, and the phenomenal character of an experience can be identified with one of those two types. The nonphenomenal type includes the representation of water as water, the phenomenal type includes representations of such
184
Ned Block
“appearance properties” as color. (See Block 1995a,b and Tye 1995a.) The visual representation of these appearance properties includes nonconceptual representations according to many representationists, and these are the ones whose representational content is identified with phenomenal character. (This is the view of Dretske, Lycan, and Tye.) What I will argue is that there is a Twin Earth case that turns on a property that does not allow a reply corresponding to the one just made with respect to water. The property is color, which is an “appearance property” if anything is. The upshot, I will argue, is that there is mental paint (but there is no argument here for mental oil). I will illustrate this point with an argument from Block (1990, 1994a) about color. I won’t go into this argument in full detail.7 Inverted Earth is a place that differs from Earth in two important ways. First, everything is the complementary color of the corresponding Earth object. The sky is yellow, the grasslike stuff is red, and so on. Second, people on inverted earth speak an inverted language. They use ‘red’ to mean green, ‘blue’ to mean yellow, and so forth. If you order a sofa from Inverted Earth and you want a yellow sofa, you fax an order for a “blue” sofa (speaking their language). The two inversions have the effect that if wires are crossed in your visual system (and your body pigments are changed), you will notice no difference when you go to Inverted Earth. After you step off the space-ship, you see some twin-grass. You point at it, saying it is a nice shade of ‘green’, but you are wrong. You are wrong for the same reason that you are wrong if you call the liquid in a Twin Earth lake ‘water’ just after you arrive there. The grass is red (of course I am speaking English not Twenglish here). But after you have decided to adopt the concepts and language of the Inverted Earth language community and you have been there for fifty years, your word ‘red’ and the representational content of your experience as of red things (things that are really red) will shift so that you represent them correctly. Then, your words will mean the same as those of the members of your adopted language community and your visual experience will represent colors veridically. Your color words and color concepts shift in meaning and content, and your color experiences shift in representational content, but the explanations are not exactly the same. Concepts shift for reasons familiar from the work of Burge and Putnam. But the representational contents of color experience may be nonconceptual and therefore not linked to the use of concepts in the language community. Still, nonconceptual contents arguably get their content causally. My dog recognizes me and has experiences that represent me even if my dog has no concept of me. If my dog goes to Twin Earth, she will react to Twin Block in just the way she reacts to me. She will mistakenly represent Twin Block as Block. But a Block-recognitional capacity is not a Twin-Block recognitional capacity, and that has to be because the Block-recognitional capacity involves causal contact with Block rather than Twin-Block. Further, any recognitional capacity can be “swamped” by a new causal source. Suppose that my dog meets Erisa for ten minutes and develops an Erisa
Mental Paint
185
recognitional capacity. Then the dog goes to Twin Earth and meets Twin Erisa, whom she misrecognizes as Erisa. Twin Erisa then adopts the dog. After ten years, the dog’s Erisarecognitional capacity has been replaced by a Twin-Erisa recognitional capacity. Evans introduced the dominant causal source account of reference, but the account covers some aspects of visual representation as well as linguistic representation. It may be said that whereas the visual or olfactory representation of Erisa has a current causal source, the visual representational of colors has its source in evolutionary history. That issue will be taken up later. In the old version of the thought experiment, you are kidnapped and inserted in a niche in Inverted Earth without your noticing it (your twin having been removed to make the niche). In the new version, you are aware of the move and consciously decide to adopt the concepts and language of the Inverted Earth language community. The change has the advantage of making it clearer that you become a member of the new community. On the old version, one might wonder what you would say if you found out about the change. Perhaps you would insist on your membership in the old language community and defer to it rather than to the new one. Your color concepts might be regarded as indeterminate in reference. Dennett (1991) made objections to the inverted spectrum case, which, applied to Inverted Earth, would say that there is no coherent interpretation of the conceptual representational contents of the traveler in the thought experiment. The point concerns conceptual contents, but applies nonetheless to nonconceptual contents. For suppose the traveler’s memory images of the sky on his fifth birthday, taken to represent the same color as now, are intimately involved in every identification of the color of any yellow thing on Inverted Earth. We may feel that radical error suffuses all his color representations, nonconceptual as well as conceptual. But thinking of the traveler as knowing all along all we know shows how incoherence can be avoided. (In Block 1990, this problem—though with respect to the inverted spectrum rather than inverted earth—is dealt with by changing the thought experiment to involve amnesia for the earlier life. The same idea would work here as well, as Burge hints in his response.) The upshot is: 1. The phenomenal character of your color experience stays the same. That’s what you say, and why shouldn’t we believe you? 2. But the representational content of your experience, being externalist, shifts with external conditions in the environment. (Recall that I am now discussing representationists who are externalists; I discussed internalist representationism at the beginning of the essay.) Your phenomenal character stays the same, but what it represents changes. This provides the basis of an argument for mental paint, not mental oil. Mental paint is what stays the same; its representational content is what changes.
186
Ned Block
What exactly is the argument for mental paint? Imagine that on the birthday just before you leave for Inverted Earth, you are looking at the clear blue sky. Your visual experience represents it as blue. Years later, you have a birthday party on Inverted Earth and you look at the Inverted Earth sky. Your visual experience represents it as yellow (since that’s what color it is and your visual experience by that time is veridical, let us suppose—I’ll deal with an objection to this supposition later). But the phenomenal character stays the same, as indicated by the fact that you can’t tell the difference. So there is a gap between the representational content of experience and its phenomenal character. Further, the gap shows that phenomenal character outruns representational content. Why? How could the representationist explain what it is about the visual experience that stays the same? What representational content can the representationist appeal to in order to explain what stays the same? This is the challenge to the representationist, and I think it is a challenge that the representationist cannot meet. The comparison with the water case is instructive. There, you will recall, we also had phenomenal continuity combined with representational change. But the representationist claimed that the phenomenal continuity itself could be given a representational interpretation. The phenomenal character of my visual experiences of twin-water and water were the same, but their representational contents differed. No problem, because the common phenomenal character could be said by the representationist to be a matter of the representation of color, sheen, flow pattern, and the like. But what will the representationist appeal to in the Inverted Earth case that corresponds to color, sheen, flow pattern, and the like? Objections There are many obvious objections to this argument, some of which I have considered elsewhere. I will confine myself here to two basic lines of objection. Bill Lycan has recently objected (1996a,b) that the testimony of the subject can show only that the phenomenal character of color experience is indistinguishable moment to moment, and that allows the representationist to claim that the phenomenal character shifts gradually, in sync with the shift in the representational content of color experience. (I raised this objection in my 1990, p. 68.) The gradual shift of phenomenal character in sync with the gradual shift of representational content avoids any gap between them. But this objection ignores the longer-term memories. The idea is that you remember the color of the sky on your birthday last year, the year before that, ten years before that, and so on, and your long-term memory gives you good reason to think that the phenomenal character of the experience has not changed gradually. You don’t notice any difference between your experience now and your experience five years ago or ten years ago or sixty years ago. Has the color of the American flag changed gradually over the years? The stars used
Mental Paint
187
to be yellow and now they are white? No, I remember the stars from my childhood! They were always white. Of course, memory can go wrong, but why should we suppose that it must go wrong here? Surely, the scenario just described (without memory failure) is both conceptually and empirically possible. (As to the empirical possibility, note that the thought experiment can be changed so as to involve a person raised in a room who is then moved to a different room where all the colors are changed. No need for a yellow sky, a yellow ceiling will do.) Now a different objection may be mounted on Lycan’s behalf—that the externalist representationist should be externalist about memory. According to the first version of my story, the representational contents of the subject’s color experience have shifted without his knowing about it. So if my story is right, Lycan (if he is to be an externalist about memory) should say that the subject’s color experience has shifted gradually without the subject’s knowing it. And that shows that the subject’s memory is defective.8 Why should we believe that memory is defective in this way? One justification is simply that the nature of phenomenal character is representational (and externalist), so the phenomenal character of experience shifts with its representational content. Since memory is powerless to reveal this shift, memory is by its nature defective. But this justification is weak, smacking of begging the question. The Inverted Earth argument challenges externalist representationism about phenomenal character, so trotting in an “error theory,” an externalist representationism about memory of phenomenal character to defend it is not very persuasive. The idea of the Inverted Earth argument is to exploit the first-person judgment that in the example as framed the subject notices no difference. The subject’s experience and memories of that experience reveal no sign of the change in environment. Yet his representational contents shift. Since the contents in question are color contents, the move that was available earlier about a set of representational contents that capture what stays the same is not available here. And that suggests for reasons that I just gave that there is more to experience than its representational content. The defender of the view that memory is defective must blunt or evade the intuitive appeal of the first-person point of view to be successful. It is no good to simply invoke the doctrine that experience is representational. But the reply to the Inverted Earth argument as I presented it above does something close to it. It says that the memories of the representational contents are wrong, so the memories of the phenomenal characters are wrong too. But that is just to assume that as far as memory goes, phenomenal character is representational content. For the argument to have any force, there would need to be some independent reason for taking externalism about phenomenal memory seriously. The representationist may reply that there is no question begging going on, but only thoroughgoing externalism about the phenomenal character of experience in all domains— in both perception and memory. But externalism about phenomenal memory has nothing
188
Ned Block
to recommend it aside from its use in defending externalist representationism about phenomenal character. It is an error theory—it postulates that ordinary memory, for example about what it was like to see the sky a few minutes ago, is inherently defective. It has the consequence that there can be changes in the phenomenal character of my experience (due to changes in the world) that I am in principle incapable of detecting, no matter how large they are or how fast they happen. The founders of externalism (such as Tyler Burge) should not be pleased by such an invocation of the externalist point of view. It will be useful to consider briefly a related objection to the Inverted Earth argument. Suppose it is said that the subject’s (that is, your) representational contents don’t ever switch. No matter how long you spend on Inverted Earth, the sky still looks blue to you. After all, as I have insisted, you notice no difference. So doesn’t the sky continue to look the same, namely blue? This line of objection has more than a little force,9 but it can easily be seen not to lead away from my overall conclusion. For it is hard to see how anyone could accept this objection without also thinking that the subject (viz., you) on Inverted Earth has an “inverted spectrum” relative to the other denizens of Inverted Earth. The sky looks yellow to them (recall that the sky there is yellow) but blue to you. And you are as functionally similar to them as you like. (We could even imagine that your monozygotic twin brother is one of them.). The sky on earth is blue and looks blue to normal people. The sky on Inverted Earth is yellow and looks yellow to normal residents. But the sky on Earth looked the same to you as the sky on Inverted Earth now looks. So you must be inverted either with respect to Earthians or Inverted Earthians.10 Shifted Spectra A shifted spectrum would obtain if, for example, things that we both call ‘orange’ look to you the way things we both call ‘reddish orange’ look to me. There is an argument for shifted spectra that appeals to the fact that color vision varies from one normal person to another. There are three kinds of cone in the retina that respond to long, medium, and short wave light. The designations “long,” “medium,” and “short” refer to the peak sensitivities. For example, the long cones respond most strongly to long wavelengths but they also respond to medium wavelengths. Two normal people chosen at random will differ half the time in peak cone sensitivity by 1–2 nm (nanometers) or more. (More precisely, the standard deviation is 1–2 nm. See Lutze et al. 1990.) This is a considerable difference, given that the long wave and middle wave cones differ in peak sensitivities by only about 25 nm. Further, there are a number of specific genetic divisions in peak sensitivities in the population that are analogous to differences in blood types (in that they are genetic polymorphisms, discontinuous genetic differences coding for different types of normal individuals). The most dramatic of these is a 51.5 percent / 48.5 percent split in the population
Mental Paint
189
of two types of long wave cones that differ by 5–7 nm, roughly 24 percent of the difference between the peak sensitivities of long and middle wave cones (Neitz and Neitz 1998).11 This characteristic is sex-linked. The distribution just mentioned is for men. Women have smaller numbers in the two extreme categories and a much larger number in between. As a result, the match on the Rayleigh test (described below) “most frequently made by female subjects occurs where no male matches” (Neitz and Jacobs 1986).12 These differences in peak sensitivities don’t show up in normal activities, but they do reveal themselves in subtle experimental situations. One such experimental paradigm uses the anomalo-scope (devised in the nineteenth century by Lord Rayleigh), in which subjects are asked to make two halves of a screen match in color, where one half is lit by a mixture of red and green light and the other half is lit by yellow or orange light. The subjects can control the intensities of the red and green lights. Neitz et al. 1993 note that “People who differ in middle wavelength sensitivity (M) or long wavelength sensitivity (L) cone pigments disagree in the proportion of the mixture primaries required” (p. 117). That is, whereas one subject may see the two sides as the same in color, another subject may see them as different—for example, one redder than the other. When red and green lights are adjusted to match orange, women tend to see the men’s matches as too green or too red (Neitz and Neitz 1998). Further, variation in peak sensitivities of cones is just one kind of color vision variation. In addition, the shape of the sensitivity curves varies. These differences are due to differences in macular pigmentation, which vary with “both age and degree of skin pigmentation”(Neitz and Jacobs 1986). Hence races that differ in skin pigmentation will differ in macular pigmentation. There is also considerable variation in amount of light absorption by preretinal structures. And this factor also varies with age. I emphasize gender, race, and age to stifle the reaction that one group should be regarded as normal and the others as defective. There are standard tests for defective color vision such as the Ishihara and Farnsworth tests, and it is an empirical fact that most men and almost all women have nondefective color vision as measured by these tests. My point is only that the facts about variation that I have presented give us no reason at all to regard any gender, race, or age as abnormal in color vision. Hardin (1993) mentions a classic study (by Hurvich, Jameson, and Cohen 1968) of the spectral location of unique green in a group of 50 normal subjects. Here is a table of locations: 5 subjects located unique green at 11 15 11 5 2
490 nm 500 nm 503 nm 507 nm 513 nm 517 nm
190
Ned Block
As Hardin notes, this is an enormous range, as can be seen in a number of ways. Take a look at a spectrum (such as the one in Hardin’s book) and block off the other areas. Or look at the Munsell chips, noting that this range goes from 5 blue-green to 2.5 green. Or simply note that the 27 nm span of this group’s location of unique green is 9 percent of the visible spectrum. The upshot is that if we take a chip that any one subject in this experiment takes as being unique green, most of the others will see it as slightly off. Thus we are justified in supposing that the way any chip looks (colorwise) is unlikely to be exactly the same as the way that chip looks to most other people, especially if they differ in sex, race, or age. So now we have the beginnings of an argument against representationism. Jack and Jill have experience that represents red things as red even though they very likely experience red slightly differently. But the argument doesn’t quite work, for as representationists could reply, the representational contents of Jack’s and Jill’s color categories may differ too, so there is still no proven gap between representational content and phenomenal character. “Color categories?” you say. “I thought the representationist was talking about nonconceptual contents?” True, but the representationist has to allow that our visual experiences represent a scarlet thing as red as well as scarlet. For we experience scarlet things as both red and as scarlet. We experience two red things of different shades as having the same color, though not the same shade, so a representationist has to concede a component of the representational contents of experience that captures that fact about experience. The representationist has to allow representational content of both color and shade. Further, pigeons can be conditioned to peck when presented with things of a certain color, as well as of a certain narrow shade. Even if the pigeon lacks color concepts, it has something short of them that involves some kind of categorization of colors as well as shades, red as well as scarlet. Let’s use the term ‘category’ for this aspect of the nonconceptual contents that are conceptlike but can be had by animals that perhaps can’t reason with the contents. Now we can see why the argument I gave doesn’t quite work against the representationist. Jack’s and Jill’s experiences of a single red fire hydrant may differ in phenomenal character but also in representational content, because, say, Jack’s visual category of red may include a shade that is included instead in Jill’s visual category of orange. Furthermore, because of the difference in Jack’s and Jill’s color vision, the fire hydrant may look more red than orange to one, more orange than red to the other. So we don’t yet have the wedge between phenomenal character and representational content. Indeed, it is quite plausible that the varying nature of our perceptual apparatuses determines different extensions for common color words. (By ‘extension’, I don’t mean just what they apply or would apply their color words to, but what they would apply their words to in ideal conditions.) Some things that Jack categorizes as blue will be catego-
Mental Paint
191
rized by Jill as green, and it is implausible to regard either as mistaken. A sensible conclusion is that they use the words blue and green in somewhat different senses, both correctly. The objective nature of color, it might be said, derives from the overlap between persons with normal color perception. There are objects that should be agreed to be blue by everyone with normal color vision, and that’s what makes them objectively blue. The objects that are not objectively blue, but are said to be blue by one normal person but not another are indeterminate in color. I endorse this point of view, though I think that it is of less value to the representationist than might appear at first glance. The way to get the wedge between phenomenal character and representational content is to apply the argument just given to shades rather than colors. Let us co-opt the word ‘aquamarine’ to denote a shade of blue that is as narrow as a shade can be, one that has no discriminable subshades. If Jack’s and Jill’s visual systems differ slightly in the ways that I described earlier, then we can reasonably suppose that aquamarine doesn’t look to Jack the way it looks to Jill. Maybe aquamarine looks to Jack the way turquoise (a different minimal shade, let’s say) looks to Jill. But why should we think that there is any difference between the representational contents of Jack’s experience as of aquamarine and Jill’s? They both acquired their categories of aquamarine by being shown (let’s suppose) a standard aquamarine chip. It is that objective color that their (different) experiences of aquamarine both represent. The upshot is that there is an empirically based argument for a conclusion—what one might call “shifted spectra”—that, while not as dramatic as an inverted spectrum, has much the same consequences for representationism and for the issue of whether there are uniform phenomenal characters corresponding to colors. There probably are small phenomenal differences among normal people that don’t track the colors that are represented. Genders, races, and ages probably differ by shifted spectra. Thus, if representationism is right, if aquamarine things look aquamarine to men, they probably don’t look aquamarine to women. And if aquamarine things look aquamarine to one race or age group, they probably don’t look aquamarine to others. In sum: If representationism is right, color experience probably cannot be veridical for both men and women, both blacks and whites, both young and old. Hence representationism is not right. I mentioned above that there is an objection to my first try at refuting representationism: Maybe Jack’s visual category that represents red includes a shade that is included in Jill’s visual category that represents orange. The present point is that the same argument does not apply to minimal shades themselves.13 This possibility should not disturb the functionalist, however, for even if there are phenomenal differences among representationally identical experiences as just supposed, the phenomenal differences might be revealed in subtle empirical tests of the sort I mentioned. That is, perhaps shifted spectra always result in different matches on a Rayleigh anomaloscope or other devices. But shifted spectra would still count against representationism.
192
Ned Block
There is a complication that I can’t treat fully here. If you regard a certain mixture of blue and green as matching the aquamarine chip, but I don’t, then our categories of aquamarine are applied by us to different things, and in that sense have different extensions. I don’t regard this as showing our categories have different representational contents, since representational contents have to do with what objective colors are represented and the example given exploits an indeterminacy in objective color. There is no determinate answer as to whether the color of the mixture of blue and green (that matches aquamarine according to me but not you) actually is aquamarine. It is an objective fact that the standard aquamarine chip is aquamarine, but there is no fact of the matter as to whether the two mixtures of blue and green are aquamarine. Representationist Objections I will put an objection in the mouth of the representationist: Whatever the differences in their visual systems, if Jack and Jill are normal observers, then in normal (or anyway ideal) conditions, the standard aquamarine chip has to look aquamarine to Jack and it also has to look aquamarine to Jill. After all, ‘looks aquamarine’ just means that their perceptual contents represent the chip as aquamarine, and you have already agreed [above] that both Jack’s and Jill’s visual experience represent the chip as aquamarine. You have argued that the representational content of their visual experience is the same (viz., aquamarine), but the phenomenal character is different. However, we representationists don’t recognize any kind of phenomenal character other than that which is given by the representational content—which is the same for Jack and Jill. The chips look aquamarine to both Jack and Jill, so they look the same to both Jack and Jill, so Jack and Jill have the same phenomenal characters on viewing the chips in any sense of ‘phenomenal character’ that makes sense. If Jack and Jill have different brain states on viewing the aquamarine chip, that just shows that the different brain states differently realize the same phenomenal character. Reply: We phenomenists distinguish between two senses of ‘looks the same’. In what Shoemaker (1981) calls the intentional sense of ‘looks the same’, the chips look the same (in respect of color) to Jack and Jill just in case both of their perceptual experiences represent it as having the same color. So I agree with the objection that there is a sense of ‘looks the same’ in which the aquamarine chip does look the same to Jack and Jill. But where I disagree with the objection is that I recognize another sense of ‘looks the same’, (the qualitative or phenomenal sense) a sense geared to phenomenal character, in which we have reason to think that the aquamarine chip does not look the same to Jack as to Jill. (The same distinction is made in somewhat different terms in Shoemaker 1981, Peacocke
Mental Paint
193
1983, and Jackson 1977.) But the case at hand supports the phenomenist rather than the representationist. For we have reasons to believe that there is a sense in which the aquamarine chip does not look the same to Jack as to Jill. One reason was given earlier: The chip that Jack regards as unique green (green with no hint of blue or yellow) is not regarded as unique green by most other people. So it looks different to Jack from the way it looks to most others, including, we may suppose, Jill. And the same is likely to be true for other chips, including the aquamarine chip. But there is another reason that even functionalists should agree to: Jack and Jill match differently on the anomaloscope. Recall that the match “most frequently made by female subjects occurs where no male matches.” If Jack produces a mixture of blue and green that he says matches the aquamarine chip, Jill will be likely to see that mixture as either “bluer” than the chip or “greener” than the chip. The big division in the ballpark we are talking about is between those who accept and those who reject qualia, that is, features of experience that go beyond the experience’s representational, functional, and cognitive features. In effect, the argument just given uses functionalism against representationism. The functional differences between different perceivers suggest phenomenal differences, but we have yet to see how those phenomenal differences can be cashed out representationally, even if they can be cashed out functionally. So the argument does not show that there are qualia, though it does go part way, by challenging one of the resources of the anti-qualia forces. Another representationist objection: These empirical facts show that colors are not objective. A given narrow shade looks different to different groups, so the different groups represent it as having slightly different colors. Thus it does in fact have slightly different colors relative to these different groups. Famously, phenylthiocarbamide tastes bitter to many people but not to many others. Phenylthiocarbamide is not objectively bitter, but it is objectively bitter relative to one group and objectively nonbitter relative to another. Color is the same, though not so dramatically. There are no absolute colors—color is relative, though only slightly so. Reply: The problem with this objection derives from a difference between our concept of taste and our concepts of at least some colors, or rather shades. We are happy to agree that phenylthiocarbamide has no objective taste—it tastes bitter to some but not others. But we do not agree that Munsell color chip 5 Red has no objective hue. Its objective hue is 5 Red no matter whether it looks different to people of different genders, races, and ages. The whole point of the Munsell color scheme, the Optical Society of America Uniform Color Space, and other color schemes is to catalog objective colors (see Hardin 1988). Every American grade school child knows the colors named by the Crayola company, despite differences in the way Burnt Sienna or Brick Red probably looks to different
194
Ned Block
children. If you paint your living room wall Benjamin Moore Linen White, it is an objective fact that you have not painted it Cameo White, Dove White, Opal White, or Antique White. If you have ordered White 113 but the paint store gives you White 114, you can get your money back. (The premixed colors have names in the Benjamin Moore scheme; the custom colors, of which there are very many, as anyone who has ever picked out one of their paint colors knows, have numbers.) If the paint dealer says, “Hey, color is relative to gender and we are different genders. Your white 113 is my 114, so I didn’t make a mistake,” he is wrong. So the problem for the representationist is this: The standard aquamarine chip is objectively aquamarine. If it looks different to men and to women, then at least one gender’s visual experience is representing it as some other shade, and that is an unacceptable consequence. Representationism is empirically false.14 There is a type of difficulty with the argument presented that I have not been able to discuss here. Perhaps Jack sees aquamarine as greener than Jill does, so there is a representational difference after all. For example, Jill may see aquamarine as greenish blue whereas Jack sees it as greenish greenish blue. I argue in Block (1999) that given that there are tens of thousands of shades of greenish blue that persons with normal vision can discriminate, it is unlikely that we (or our visual systems) have available to us (as part of our normal visual competence) representational resources that would distinguish close shades of greenish blue. Alternatively, suppose that Jack and Jill both see aquamarine as greenish blue, but their visual categories corresponding to the terms “green” and “blue” have slightly different extensions. Can their different phenomenal impressions of aquamarine be explained in terms of different representations of green and blue? I argue not. See the discussion of “subjective color” in Block (1999). Now we are in a position to counter another of Lycan’s (1996a,b) arguments. He notes that I concede that our inverted earth subject has experiences whose representational contents (on looking at the inverted earth sky) shift from looking blue to looking yellow. And he concludes that this undermines the subject’s claim that there is no difference between the way the sky looked to him on earth and the way the sky looks to him now fifty years later on inverted earth. I have admitted that the sky looked blue to him at the beginning of our story on earth, and that the sky looks yellow to him at the end of our story on inverted earth. So how can I (or he) claim that the sky looks the same to him as it always did? Since he can’t remember any change, we must conclude that the reason is that the change was gradual. But this argument ignores the distinction made earlier between the two senses of ‘looks the same’. ‘Looks blue’ does not express a phenomenal character but rather a representational content! We can all agree that his color representational contents have changed. But it is phenomenal character, not representational content, that is relevant to noticing a difference, and phenomenal character has remained the same. Repre-
Mental Paint
195
sentational content has changed purely externally without any corresponding change in phenomenal character. Tyler Burge suggests (see his reply to me in this volume) that commonly, the phenomenal character of color perception is a factor in individuating the sense component of color concepts. He leaves it open whether phenomenal character is literally a part of the sense of color concepts or part of the intentional content of thoughts involving color concepts. For simplicity, let’s take phenomenal character to constitute the sense of color concepts. Then inverted spectrum cases fail to show that phenomenal character doesn’t supervene on representational content. For if representational content includes phenomenal character, any difference in phenomenal character will, ipso facto, constitute a difference in representational content. But representationists of the sort I have been talking about here should not be pleased by this Pickwickian victory. They have tried to cash phenomenal character of color experience in terms of the color represented. That is, representationists have construed phenomenal character purely referentially. The phenomenal experience as of red is a matter of visual experience representing something as red. My arguments are supposed to show that view is wrong even if we follow Burge in taking phenomenal character to be an individuating factor in the representational content of color experience. I began the discussion of externalism by discussing a thought experiment involving Putnam’s Twin Earth. The idea was that I had emigrated from Twin Earth to Earth and that after many years on Earth the representational contents of my visual experiences of the liquid in the oceans shifted even though the phenomenal character of the experiences stayed the same. I noted that there is no immediate problem for the representationist here, since the constant phenomenal character can be understood in representational terms. However, there is no corresponding move available to the representationist in the case of an emigration to or from Inverted Earth. This is one of a number of reasons given in this essay to resist the identification of the phenomenal character of the experience as of red with representing red.15 Notes 1. Controversially, I think that the scientific nature of qualia could be discovered to be functional or representational or cognitive. Thus I actually prefer a definition of qualia as features of experience that cannot be conceptually reduced to the nonphenomenal, in particular the cognitive, the representational, or the functional. But this view will play no role in this essay. 2. These mental properties aren’t mental paints because they don’t represent. However, my definition of mental oil does, unfortunately, count them as mental oils because they are mental properties of experience that do not represent. So let’s understand the definition of mental oil as containing a qualification: The property of having certain intentional contents does not count as a mental oil. 3. Shoemaker (1994a,b) holds that when one looks at a red tomato one’s experience has a phenomenal character that represents the tomato as having a certain phenomenal property and also as being red, the latter via the
196
Ned Block
former. On his view, the objects have certain phenomenal characters. Each phenomenal property of an object can be defined in terms of its production in certain circumstances of a certain nonrepresentational phenomenal character of experience. The view is motivated in part by a consideration of the Inverted Spectrum. For concreteness, suppose that George’s spectrum is inverted with respect to Mary’s. George’s experience of the apple represents it both as red and as having phenomenal property Q (the former via the latter). Mary’s experience represents the apple as red and as having phenomenal property P. (George’s experience represents grass as green and Q.) What determines that George’s experience represents phenemonal property Q is that it has the nonrepresentational phenomenal character Q*; and Q gets its identity (with respect to George) from the (normal) production of Q*. Red can be identified with the production of Q* in George, P* in Mary, etc. The phenomenal character Q* is in a certain sense more basic than the phenomenal property Q, for Q* is what makes the experience represent Q. Still, it could reasonably be said that there is nothing in Q* over and above its representation of Q, and so Shoemaker’s view qualifies as representationist about phenomenal character Q* on my definition. However, there is more to Q* than the representational content red. Shoemaker’s view could therefore be said to be representationist with respect to phenomenal character Q*; so phenomenal character supervenes on the representation of phenomenal properties. But phenomenal character does not supervene on the representation of colors. George and Mary both have experiences that represent red, but one has phenomenal character Q*, the other P*. So if inverted spectra are possible, representationism with respect to colors is wrong. It is that kind of representationism that is at issue in this essay (except in the internalism section). That is, the argument against representationism in this essay is directed against the view that the representational content of color experience is color content. In my view as well as in Shoemaker’s the phenomenal character of an experience as of red is not exhausted by the representation of what is seen as red. I say that representationists hold that the phenomenal character of an experience is or at least is determined by its representational content. I use the phrase ‘is determined by’ rather than ‘supervenes on’ because I think the supervenience relation as it is normally interpreted is too weak to capture the representationist ideology. Suppose, for example, that a dualist agrees that phenomenal character supervenes on representational content, but only because the dualist believes in irreducible phenomenal—representational laws of nature, laws of nature that describe the effect of representational content on the soul. This would fit most definitions of supervenience but would certainly be incompatible with the doctrine that representationists intend. 4. Harman tells me that his view is actually the same as what I ascribe to Lycan. 5. As I mentioned earlier, there is also a problem for the recognitional view in the plausibility of the idea that we (or animals) can have an experience without any sort of categorization or recognition. 6. See the papers in Crane (1992) for more on this issue. 7. I will make use of Harman’s (1990) Inverted Earth example. In my (1978) I used a cruder example along the same lines. (Pages 302–303 of Block 1978—reprinted on p. 466 of Lycan 1990 and p. 227 of Rosenthal 1991). Instead of a place where things have the opposite of their normal colors, I envisioned a remote arctic village in which almost everything was black and white, and the subject of the thought experiement was said to have no standing color beliefs of the sort of ‘grass is green’. Two things happen to him: He confuses color words, and a color inverter is placed in his visual system. Everything looks to have the complementary of its real color, but he doesn’t notice it because he lacks standing color beliefs. Harman used the Inverted Earth example to make a point orthogonal to the point made here. His conclusion is that representational content does not supervene on the brain. He does not consider someone emigrating to Inverted Earth or a pair of twins one of whom is on earth the other on Inverted Earth. Instead, he describes the Inverted Spectrum thought experiment in a form in which a person puts on color-inverting lenses. Then he describes Inverted Earth and notes that the inverting lenses could be donned by someone on Inverted Earth with the upshot that the brain state that normally represents blue would then represent orange. There is no discussion of sameness or difference of phenomenal character. Here is Harman’s complete discussion of the matter: Consider Inverted Earth a world just like ours, with duplicates of us, with the sole difference that there the actual colors of objects are the opposite of what they are here. The sky is orange, ripe apples are green, etc. The inhabitants of Inverted Earth speak something that sounds like English, except that they say the sky is ‘blue’, they call ripe apples ‘red,’ and so on. Question: what color does their sky look to them? Answer: it looks orange. The concept they express with the word ‘blue’ plays a relevantly special role in the normal perception of things that are actually orange.
Mental Paint
197
Suppose there is a distinctive physical basis for each different color experience. Suppose also that the physical basis for the experience of red is the same for all normal people not adapted to color inverting lenses, and similarly for the other colors. According to (nonsolipsistic) conceptual role semantics this fact is irrelevant. The person who has perfectly adapted to color inverting lenses \will be different from everyone else as regards the physical basis of his or her experience of red, but that will not affect the quality of his or her experience. Consider someone on Inverted Earth who perfectly adapts to color inverting lenses. Looking at the sky of Inverted Earth, this person has an experience of color whose physical basis is the same as that of a normal person looking at Earth’s sky. But the sky looks orange to the person on Inverted Earth and blue to normal people on Earth. What makes an experience the experience of something’s looking the color it looks is not its intrinsic character and/or physical basis but rather its functional characteristics within an assumed normal context. 8. I took this point to be raised by some of the discussions of Lycan’s paper at the meeting of Socioedad Filosofica Ibero-Americano in Cancun in June, 1995. If I had to credit it to anyone, it would be Alan Gibbard. 9. In my (1998) I give a different response in terms of the Swampman visiting Inverted Earth. The Swampman response is also used there to deal with Dretske’s and Tye’s plausible view (shared by Burge) that color representation is a product of evolution, hence if the sky on Inverted Earth continues to produce the same phenomenal experience in our traveler, it also is represented as blue. 10. It may be thought that the Inverted Spectrum argument is superior because it is a case of same representational content, different phenomenal character, and that this yields a more direct argument that phenomenal character goes beyond representational content. Inverted earth might be said to provide only the converse: same phenomenal character, different representational content. The upshot, it might be said, is that I have had to resort to a burden of proof argument. I have had to challenge the representationist with the question: What kind of representational content of experience stays the same? This thought, which I have been guilty of, makes a simple error. Both the Inverted Spectrum and Inverted Earth cases involve counterexamples in both directions. Consider the Inverted Earth twin looking at a lemon and compare that with the Earth twin looking at the sky. This is a case of same representational content (both states represent blue, the color of Inverted Earth lemons), different phenomenal character. So we can squeeze both same representational content/different phenomenal character and same phenomenal character/different representational content out of the thought experiment. Similarly, we can imagine an Inverted Spectrum subject looking at a lemon while his inverted twin looks at the sky. This is a case of the same phenomenal character/different representational content, the converse of the usual Inverted Spectrum conclusion. 11. Neitz, Neitz, and Jacobs (1993) report a figure of 62 percent / 38 percent. 12. Neitz and Neitz (1998) explain the result as follows. Genes for long and medium wave pigment are on the X chromosome. Men have a single X chromosome, which is roughly equally likely to be each of the two forms, and hence they show a matching distribution with two spikes corresponding to the peak sensitivities of the two kinds of cones. Women have two X chromosomes, and in roughly half the cases, they have different alleles of the long wave gene in the two chromosomes. When this happens, one gene deactivates the other. But that happens independently in each cell, the result being that the average in these women is intermediate between the extreme values, and so they have long wave absorption peaks roughly in between the two groups of men. 13. Some may wish to try to avoid this conclusion by insisting that colors are not real properties of things, that our experience ascribes phenomenal properties to physical objects that the objects do not and could not have (Boghossian and Velleman 1989, p. 91). Recall that representationism as I am understanding it says that the phenomenal character of a visual experience as of red consists in its representing something as red. Are the phenomenal properties (1) colors or (2) phenomenal properties in something like Shoemaker’s (1994a,b) sense? If the latter, the view countenances unreduced phenomenal characters and is therefore incompatible with representationism as I understand it. (See the discussion of Shoemaker’s views in the penultimate section of this essay.) The former interpretation is that our experiences represent objects as having colors such as red or orange, but objects do not and could not have those colors. Colors are in the mind, not in the world outside the mind. The point I will be making contains the materials for refuting this view. Briefly, the picture of colors as in the mind rather than in the world has to explain our agreement on which Munsell chip is 4 Red. But how can the Boghossian-Velleman picture, on this interpretation of it, explain this agreement, given that we have somewhat
198
Ned Block
different experiences, colorwise, when we see that chip? If your experience represents the 4 Red chip the way mine represents the 5 Red chip, how can we explain our agreement on which chips are 4 Red and 5 Red? Perhaps Boghossian and Velleman will say that you and I have different phenomenal characters that represent the same color. But this line of thought makes sense only if phenomenal characters are in the mind and colors are in the world, contrary to the current interpretation of Boghossian and Velleman. 14. I present the Shifted Spectrum argument in more detail in Block (1999). 15. I am grateful for discussions with Tyler Burge, Brian Loar, Paul Horwich, Pierre Jacob, and Georges Rey, and to Bill Lycan and his NEH Summer Seminar, 1995. I am grateful to Burge, Lycan, Rey, and Sydney Shoemaker for helpful comments on an earlier draft. This essay is a descendant of “Mental Paint and Mental Latex” in E. Villaneuva, ed., Philosophical Issues 7, Ridgeview: Atascadero, 1996.
References Berlin, Brent and Paul Kay. 1969. Basic Color Terms. Berkeley: University of California Press. Blackburn, Simon. 1993. Essays in Quasi-Realism. New York: Oxford University Press. Block, Ned. 1978. Troubles with Functionalism. Minnesota Studies in the Philosophy of Science (C. W. Savage, ed.), vol. IX: 261–325. Reprinted (in shortened versions) in Block (ed.), 1980, Readings in Philosophy of Psychology 1; in W. Lycan (ed.), 1990, Mind and Cognition, Oxford: Blackwell: 444–469 (435–440 in the second edition); in D. M. Rosenthal (ed.), 1991, The Nature of Mind, Oxford: Oxford University Press: 211–229; in B. Beakley and P. Ludlow (eds.), 1992, Philosophy of Mind, Cambridge, MA: MIT Press; and in Goldman, Alvin (ed.), 1993, Readings in Philosophy and Cognitive Science, Cambridge, MA: MIT Press. ———. 1980. Readings in Philosophy of Psychology. Cambridge, MA: Harvard University Press. ———. 1981. Psychologism and Behaviorism. Philosophical Review 90(1): 5–43. ———. 1990. Inverted Earth. Philosophical Perspectives 4: 51–79. Reprinted in W. Lycan (ed)., 1999, Mind and Cognition (second edition), Oxford: Blackwell. Also reprinted in Block, Flanagan, and Güzeldere (1997). ———. 1994a. Qualia. In A Companion to Philosophy of Mind, S. Guttenplan (ed.), 514–520. Oxford: Blackwell. ———. 1994b. Consciousness. In A Companion to Philosophy of Mind, S. Guttenplan (ed.), 209–218. Oxford: Blackwell. ———. 1995a. On a Confusion about a Function of Consciousness. Behavioral and Brain Sciences 18: 227–247. Reprinted in Block, Flanagan, and Güzeldere (1997). ———. 1995b. How Many Concepts of Consciousness? Behavioral and Brain Sciences 18. ———. 1996. Mental Paint and Mental Latex. In Philosophical Issues 7: Perception, E. Villanueva (ed.). Ridgeview: Atascadero. ———. 1998. Is Experiencing Just Representing? Philosophy and Phenomenological Research 58(3): 663–670. ———. 1999. Sexism, Racism, Ageism, and the Nature of Consciousness. In The Philosophy of Sydney Shoemaker, Philosophical Topics, Richard Moran, Jennifer Whiting, and Alan Sidelle (eds.), 26 (1&2). Block, Ned and Jerry Fodor. 1972. What Psychological States Are Not. The Philosophical Review 81(2): 159–181. Reprinted in Block (ed.), Readings in Philosophy of Psychology, volume 1, 1980. Block, Ned, Owen Flanagan, and Güven Güzeldere. (eds.). 1997. The Nature of Consciousness: Philosophical Debates. Cambridge, MA: MIT Press. Boghossian, Paul and David Velleman. 1989. Color as a Secondary Quality. Mind 98: 81–103. Burge, Tyler. 1979. Individualism and the Mental. Midwest Studies in Philosophy 4: 73–121. ———. 1997. Two Kinds of Consciousness. In Block, Flanagan, Güzeldere (1997). Byrne, Alex and David Hilbert. 1997a. Colors and Reflectances. In Byrne and Hilbert (1997b). ———. 1997b. Readings on Color: The Philosophy of Color, volume 1. Cambridge, MA: MIT Press.
Mental Paint
199
Crane, Tim (ed.). 1992. The Contents of Experience: Essays on Perception. New York: Cambridge University Press. Dennett, Daniel C. 1991. Consciousness Explained. Boston: Little Brown. Dretske, Fred. 1995. Naturalizing the Mind. Cambridge: MIT Press/A Bradford Book. ———. 1996. Phenomenal Externalism, or If Meanings Ain’t in the Head, Where Are Qualia? In Philosophical Issues 7: Perception, E. Villanueva (ed.). Atascadero: Ridgeview. Gibbard, Alan. 1990. Wise Choices, Apt Feelings: A Theory of Normative Judgment. Cambridge, MA: Harvard University Press. Hardin, C. L. 1988. Color for Philosopher. Indianapolis: Hackett. Harman, Gilbert. 1982. Conceptual Role Semantics. Notre Dame Journal of Formal Logic 23: 242–256. ———. 1990. The Intrinsic Quality of Experience. In Philosophical Perspectives 4, Action Theory and Philosophy of Mind, James Tomberlin (ed.), pp. 31–52. Atascadero: Ridgeview. ———. 1996. Explaining Objective Color in Terms of Subjective Reactions. In Philosophical Issues 7: Perception, E. Villanueva (ed.), pp. 1–17. Atascadero: Ridgeview. Humphrey, N. 1992. A History of the Mind. New York: Simon and Schuster. Hurvich, L. M., D. Jameson, and J. D. Cohen. 1968. The Experimental Determination of Unique Green in the Spectrum. Perceptual Psychophysics 4: 65–68. Jackson, Frank. 1982. Epiphenomenal Qualia. Philosophical Studies 32: 127–36. ———. 1977. Perception. Cambridge: Cambridge University Press. Lewis, David. 1990. What Experience Teaches. In Mind and Cognition: A Reader, W. Lycan (ed.). Oxford: Blackwell. Loar, Brian. 1990. Phenomenal States. In Philosophical Perspectives 4, Action Theory and Philosophy of Mind, James Tomberlin (ed.), pp. 81–108. Atascadero: Ridgeview. A much revised version of this paper is found in Block, Flanagan, and Güzeldere (1997), pp. 597–616. Lutze, M., N. J. Cox, V. C. Smith, and J. Pokorny. 1990. Genetic Studies of Variation in Rayleigh and Photometric Matches in Normal Trichromats. Vision Research (30)1: 149–162. Lycan, William G. 1987. Consciousness. Cambridge, MA: MIT Press. ———. 1990. Mind and Cognition, A Reader. Oxford: Blackwell. ———. 1995. We’ve Only Just Begun. Behavioral and Brain Sciences 18: 262–263. ———. 1996a. Layered Perceptual Representation. In Philosophical Issues 7: Perception, E. Villanueva (ed.), pp. 81–100. Atascadero: Ridgeview. ———. 1996b. Consciousness and Experience. Cambridge, MA: MIT Press. ———. 1996c. Replies to Tomberlin, Tye, Stalnaker, and Block. In Philosophical Issues 7: Perception, E. Villanueva (ed.). Atascadero: Ridgeview. McDowell, John. 1994. The Content of Perceptual Experience. Philosophical Quarterly 44(175): 190–205. McGinn, Colin. 1991. The Problem of Consciousness. Oxford: Blackwell. Moore, G. E. 1922. Philosophical Studies. London: Routledge. Neitz, J. and G. Jacobs. 1986. Polymorphism of Long-wavelength Cone in Normal Human Color Vision. Nature 323: 623–625. Neitz, J., M. Neitz, and G. Jacobs. 1993. More than Three Different Cone Pigments among People with Normal Color Vision. Vision Research (33)1: 117–122. Neitz, M. and J. Neitz. 1998. Molecular Genetics and the Biological Basis of Color Vision. In W. G. Backhaus, R. Kliegl, and J. S. Werner, (eds.), Color Vision: Perspectives from Different Disciplines. Berlin: de Gruyter. NIH. 1993. Mixed-up Genes Cause Off-Color Vision. Journal of NIH Research 5(February): 34–35. Peacocke, Christopher. 1983. Sense and Content. Oxford: Oxford University Press.
200
Ned Block
Rey, Georges. 1992a. Sensational Sentences. In Consciousness: Psychological and Philosophical Essays, M. Davies and G. Humphreys (eds.). Oxford: Blackwell. ———. 1992b. Sensational Sentences Switched. Philosophical Studies 68: 289–331. Robinson, William. 1998. Intrinsic Qualities of Experience: Surviving Harman’s Critique. Erkentniss 47: 285–309. Rosch, Eleanor. 1972. [E. R. Heider]. Probabilities, Sampling and Ethnographic Method: The Case of Dani Colour Names. Man 7: 448–466. Rosenthal, David. 1991. The Nature of Mind. Oxford: Oxford University Press. Shoemaker, Sydney. 1981. Absent Qualia are Impossible: Reply to Block. Philosophical Review 90: 581–600. ———. 1982. The Inverted Spectrum. Journal of Philosophy 7(79): 357–381. ———. 1994a. Self-Knowledge and Inner Sense; Lecture III: The Phenomenal Character of Experience. Philosophy and Phenomenological Research 2(54): 291–314. ———. 1994b. Phenomenal Character. Noûs 28: 21–38. Stalnaker, Robert. 1996. On a Defense of the Hegemony of Representation. In Philosophical Issues 7: Perception, E. Villanueva (ed.). Atascadero: Ridgeview. Strawson, Galen. 1994. Mental Reality. Cambridge, MA: MIT Press. Tye, Michael. 1994. Qualia, Content, and the Inverted Spectrum. Noûs 28: 159–183. ———. 1995a. Blindsight, Orgasm, and Representational Overlap. Behavioral and Brain Sciences 18: 268–269. ———. 1995b. Ten Problems of Consciousness. Cambridge, MA: MIT Press. White, Stephen. 1994. Color and Notional Content. In Philosophical Topics 22: 471–504.
10
Mental Content and Hot Self-Knowledge
Bernard W. Kobes In this essay I explore the implications of a view of conscious thought as an activity that a thinker S engages in, rather than as something that S undergoes or as something that merely occurs in S. Such an account of cognitive agency promotes our understanding of a thinker’s authority in making certain attributions of mental states and events to himself. The theme will be that a thinker’s authority stems from two facts: first, that his selfattributions are not merely passive registrations of inner goings-on but have instead a certain performative character, and second, that he will typically have a kind of spontaneous knowledge of his performances. In some of the recent philosophical literature, authoritative self-attributions have been seen as problematic in view of externalist accounts of mental content. I shall argue that the view of conscious thought as active, as having a certain performative character, can deepen our understanding of how it is that mental content is externally fixed. For example, it helps us respond to certain puzzles about how to ascribe mental content in Twin Earth scenarios, puzzles that call into question whether we have genuine knowledge of our current, conscious thoughts. Assumptions and Background I will assume that we can usefully distinguish mental content from mental relation. Thus, in the linguistic schema for ascribing propositional attitudes, ÈS js that p˘, the mental content is given by the sentential complement p, while the mental relation is given by the predicate j—believes, desires, hopes, wishes, doubts, and so on. The mental relation is broadly functional in nature. A mental relation is a characteristic kind of role that the content, or rather the thinking or representing of that content, plays in the thinker’s cognitive economy. If, for example, a thinker on reflection moves from doubt to belief with respect to some content p, then one and the same content representation plays first one broadly functional role, then another.1 Mental relations may be broadly categorized into two kinds, depending on their direction of fit. Speech acts have been described as having either world-to-word or word-toworld direction of fit. Applying the distinction to Elizabeth Anscombe’s vivid example, we may say that a list of grocery items prepared by a man going shopping has world-toword direction of fit, while the same list of items prepared by a detective who follows him around in the grocery store has word-to-world direction of fit (Anscombe 1957, p. 56). (The direction of fit is typically the opposite of the direction of causal flow.) The distinction may also be applied to mental relations. Some, such as belief and appearance, have
202
Bernard W. Kobes
mind-to-world direction of fit, while others, such as intention and desire, have world-tomind direction of fit. A mental relation’s direction of fit is a matter of how we conceive the “responsibility” for success of a mental state or event involving that relation. In the case of belief and appearance, we hold fixed the state of the world, and it is the “responsibility” of the belief or appearance, or the processes that produce them, to effect a match between the mental content and the world. So the direction for belief and appearance is mind-to-( fit)-world. In the case of intention and desire, we hold fixed the mental state, and it is the “responsibility” of the world, or our actions on it, to effect a match between the mental content and the world. So the direction for intention and desire is world-to-( fit)-mind. Following Lloyd Humberstone, I will call relations like belief and appearance thetic, and relations like intention and desire telic (Humberstone 1992). I further assume that Putnam, Burge, and others have established that many of our thought contents are externally fixed. So it is possible to vary in imagination the thinker’s thought contents—even his de dicto thought contents—simply by varying his physical or social environment, while holding fixed his entire history of physical states of body and brain, “raw” phenomenal states (if such there be), and “narrow” behavioral and functional states.2 But I will remain neutral among a variety of more specific theories about how mental content may be externally fixed. So I will not try to adjudicate among, for example, causal or historical chain theories, teleological theories, or covariation theories of mental content. Despite the popularity of externalist views of mental content, philosophers are by and large impressed by the immediacy and authority of our knowledge of our own current conscious mental contents. If S currently and consciously thinks “Snow is an unnecessary freezing of water,” then S knows the intentional content of his thought immediately—that is, without the aid of inference or observation—and with an authority that no one else has with respect to S’s thoughts. There is a prima facie appearance of tension here. For if S’s thought contents are fixed partly by conditions external to S, would it not be incumbent on S to investigate those external conditions in order to know what his own thought contents are? And if self-knowledge requires investigation of external conditions, then S’s knowledge of his own thought contents is neither more immediate nor more authoritative than anyone else’s knowledge of S’s thought contents (see Woodfield 1982, viii). The by now familiar reconciling response to the tension between content externalism and first-person authority is, very briefly, that the very external conditions that help fix the contents of S’s first-order thoughts also, and simultaneously, help fix the contents of S’s thoughts about his thoughts. In cases of what Tyler Burge calls basic self-knowledge—as when I judge that I am thinking, with this very thought, that water is more common than mercury—I use the same content in the higher-order thought about my thought as I use
Mental Content and Hot Self-Knowledge
203
in the lower-order thought about the world. The higher-order thought is self-referential and self-verifying; thinking it ensures its own truth. The external conditions on my thought content affect the first-order thought and the higher-order thought equally. The thinker need not know in a discursive manner what these conditions are in order to have a knowledgeable higher-order thought about his first-order thought. These observations successfully block, I think, the simplest line of thought from externalism to our not having immediate and authoritative knowledge of occurrent, conscious mental contents. Let me call this the Burge—Heil reconciliation story, after two philosophers who presented early versions of it.3 Doubts and Discontents I want now to collect and briefly review from some recent literature a number of doubts and discontents about the Burge–Heil reconciliation story. I will then propose a revised and extended version of the account, with the goal of answering these doubts and discontents. Answering the doubts and discontents will put externalism in a somewhat broader framework in the philosophy of mind, which, though controversial, will prove to have significant theoretical attractions. The first source of doubt and discontent that I will consider is inspired by an argument of Brian Loar (1994).4 Consider S’s reflexive judgment, “I am now thinking that Socrates drank some hemlock.” Suppose S takes as given the existence of Socrates and hemlock. Still, S can think that Socrates drank some hemlock only if he stands in certain externally determined causal or historical relations to those objects. This reflection may cause S to suspect that his knowledge of his own mental content, even given the existence of Socrates and hemlock, is merely empirical. For, S reasons, it is open to doubt on empirical grounds that he himself stands in the appropriate external causal or historical relation to Socrates and to hemlock. So the issue, as Loar sets it up, is how S could reassure himself of the apriority of an inference from the existence presuppositions, that Socrates and hemlock exist, to the reflexive judgment “I am now thinking that Socrates drank some hemlock.” Now let us suppose that S knows that such reflexive thoughts are always true, given the existence presuppositions, since the content of the first-order thought and (part of) the content of the higher-order thought must be identical. S knows the Burge–Heil reconciliation story. The inference from existence presuppositions to reflexive judgment is, we may say, pragmatically valid. That is, if a thinking of the premise is true in a certain context, then a thinking of the conclusion would be true in the same context, even though the propositions expressed by premise and conclusion are related only contingently. But how can S assure himself, by way of such considerations, that the inference does not depend in some subtle way on his knowledge of external causal or historical relations that he bears to
204
Bernard W. Kobes
Socrates and to hemlock? What S knows is that all thoughts of a certain form—including the thought “I am now thinking that Socrates drank some hemlock”—are true. But it would seem that in order to use that information to resolve his empiricist doubt, S would already have to know that he is thinking a thought of that reflexive form! And that knowledge may, for all we have said, depend on empirical knowledge of external causal or historical relations. The difficulty, then, is that S cannot use his knowledge of the Burge–Heil reconciliation story in a wholly a priori demonstration, given the existence assumptions, that he is thinking that Socrates drank some hemlock. Loar imagines the thinker reasoning as follows: If I am now thinking that I am now thinking that Socrates drank some hemlock, that reflexive thought must be true, and so it would be the case that I am now thinking that Socrates drank some hemlock. Indeed. But this merely pushes the question back a stage. . . . [Given the coherence of the doubt that I stand in the appropriate relations to Socrates and to hemlock, how] am I to decide without further external information whether I am now thinking that I am now thinking that Socrates drank some hemlock? (Loar 1994, p. 64)
Extrapolating from Loar’s argument, it would appear that if the Burge-Heil reconciliation story is all S has to work with, then S would be launched on an infinite regress if he tried to construct a wholly a priori justification for his inference from the existence of Socrates and hemlock to the conclusion that he is thinking that Socrates drank some hemlock. To demonstrate to himself a priori that he is thinking that Socrates drank some hemlock, he would first have to know that he is thinking that he is thinking that Socrates drank some hemlock. And presumably, to demonstrate to himself a priori that he is thinking this, he would first have to know that he is thinking that he is thinking that he is thinking that Socrates drank some hemlock. And so on. This is an unacceptable result; for surely if S can make this inference a priori, and if he has the requisite reasoning ability, he should be able to exhibit the inference to himself as a priori without regress. Moreover, the same should be true if we substitute for ‘a priori’, ‘immediate[ly] and authoritative[ly]’.5 A second source of doubt and discontent concerns the scope of the reconciliation story. Burge discusses only basic self-knowledge, as when I judge that I am explicitly thinking, with this very thought, that water is more common than mercury. But surely our immediate and authoritative self-knowledge goes beyond that provided in this paradigm. For one thing, it seems that we have self-knowledge of what we are thinking even when we are thinking ordinary conscious thoughts about the world. In such cases we are not thinking explicitly reflexive thoughts. Yet we know what we are thinking. The explicitly reflexive thought of basic self-knowledge may be such that, as Burge says, thinking it ensures its own truth. But how does this help to explain our knowledge of our current conscious
Mental Content and Hot Self-Knowledge
205
thought contents when we are not thinking conscious reflexive thoughts in the form of Burge’s paradigm? Paul Boghossian has pressed a variety of other doubts about the extendibility of the reconciliation story (Boghossian 1989). Boghossian argues that the account does not explain our knowledge of the mental relations of our own mental states. He notes that “[s]elfregarding judgments about what I occurrently desire or fear, for example, are manifestly not self-verifying, in that I need not actually desire or fear any particular thing in order to judge that I do. Thus it may be that I judge: I fear that writing requires concentration, without actually fearing that it does. The judgment is not self-verifying” (Boghossian 1989, p. 21). More generally, although a thinking of the first-order thought content is entailed by a thinking of the higher-order thought content, the relational component of the firstorder thought is not entailed by the thinking of the higher-order thought. So this aspect of immediate and authoritative self-knowledge remains unexplained. Other doubts about extendibility loom. It seems we have authoritative self-knowledge with respect to the contents of immediately past mental events. How could the Burge–Heil reconciliation story account for this? As Boghossian points out, a memory-based judgment about what I was thinking just now is not self-verifying. The thinking of the first-order content can be a part of the thinking of the second-order content only if the two contents are absolutely coincident (Boghossian 1989, p. 21). Moreover, the Burge–Heil reconciliation story appears not to explain our authoritative knowledge of our standing mental states. If I judge that I believe that writing requires concentration, my judgment is authoritative. But the standing state of which I judge is extended in time, so it is difficult to see how my judgment could be self-verifying or how the first-order standing state could be a part of the second-order judgment (Boghossian 1989, pp. 21–22). So the second source of doubt about the Burge–Heil reconciliation story is really a family of concerns about its extendibility beyond the paradigm of basic self-knowledge. Boghossian also charges—and this is my third source of doubt—that the reconciliation story renders our self-knowledge “cognitively insubstantial” when in fact self-knowledge is a genuine cognitive achievement. Self-knowledge varies in quantity and quality depending on the effort, attention, and skill that S puts into it. But if we have self-knowledge simply in virtue of the self-verifying nature of Burge’s paradigm of basic self-knowledge, in which the content of the first-order thought is a part of the content of the higher-order thought, then self-knowledge requires neither observation nor inference based on observation. Yet it concerns contingent matters of fact. In these respects self-knowledge would be like the knowledge expressed by the sentence ‘I am here now’. How, then, can selfknowledge be subject to cultivation or neglect? Given the Burge–Heil reconciliation story,
206
Bernard W. Kobes
self-knowledge becomes problematic because it is in fact so imperfect, not because it is so perfect (Boghossian 1989, pp. 17–19). Finally, there are puzzles that arise from reflection on so-called slow-switching Twin Earth cases. On Twin Earth they have no water; instead of H2O, they have some other, superficially similar liquid with chemical composition XYZ, which they refer to by the word-form ‘water’, but which we on Earth may call twater. Suppose S is surreptitiously transported from Earth to Twin Earth. After a sufficiently long period of acculturation— say three decades of verbal interaction with his new cohorts, drinking and bathing in the liquid they call ‘water’, and the like—S will have acquired the relevant Twin Earth concept, which of course he now also expresses with the word-form ‘water’. Thinking back to an occasion of over three decades ago, S remembers thinking a thought that he then expressed using ‘water’. Now S does not know what he thought back then. Yet he has forgotten nothing! How can this be? Boghossian suggests an answer: perhaps S never genuinely knew what he was thinking (Boghossian 1989, pp. 13–15, 22–23).6 So reflection on slow-switching scenarios supplies a fourth source of doubt about the Burge–Heil reconciliation story. The Burge–Heil story answers a certain kind of very direct challenge to the compatibility of content-externalism and first-person authority. But as a positive account of selfknowledge it is rather thin, and that makes it difficult to see how to extend the account beyond the paradigm of basic self-knowledge. My goal here is first to set the reconciliation story within an account of the thinker as cognitive agent, and then to argue that the result gives us resources to respond to the four sources of doubt and discontent. Cognitive Agency To some philosophers it has seemed that we are doxastic agents, that what we think is up to us in certain ways. Others have seen our mental lives as having a more passive character. This contrast is vividly manifested in two views of the relationship of the thinker to his beliefs. Descartes, for example, held that belief involves assent by the will. Descartes seems even to have held that the will is perfectly free to assent to any proposition it chooses, with the possible exception of those rare occasions when it perceives clearly and distinctly. He writes: The will may be termed infinite; for we never observe any possible object of another will (even the immeasurable will of God) that does not also fall within the range of our own will. (Descartes 1971, p. 187) The existence of freedom in our will, and our power in many cases to assent or dissent at our pleasure, is so clear that it must be counted among the first and most axiomatic of our innate notions. (Descartes 1971, p. 188)
Mental Content and Hot Self-Knowledge
207
For Hume, on the other hand, belief is a passive phenomenon. Hume writes: [B]elief consists merely in a certain feeling or sentiment; in something, that depends not on the will, but must arise from certain determinate causes and principles, of which we are not masters. (Hume 1888, p. 624)
Each view finds support in certain elements of a reflective first-person perspective. On Hume’s behalf, it must be said that the phenomenology of belief powerfully suggests that I cannot believe, just like that, anything I want to. On the other hand, my beliefs don’t feel like they simply “happen” to me. Rather, the world often feels like it simply happens to me, and this may be sufficient to explain my sense, to the extent that I have it, of being passive with respect to belief. In fact, when I deliberate about what is the case, I seem to construct my beliefs about the world. Often enough it is a project that engages me as a person. So Descartes’s view is plausible in at least these respects: we are doxastic agents, and in learning, and in changing and updating our beliefs, we act on our store of beliefs. Taking cognitive agency seriously pays theoretical dividends: it can help explain some otherwise puzzling features of our self-knowledge. For under the right conditions, the commitment involved in agency allows for a spontaneous justified belief on the part of the agent about what he is doing. If S’s belief and desire together cause an action that they jointly make rational, then S need not, as a matter of conceptual necessity, be aware of the rational connection between the belief and desire, on the one hand, and the action, on the other. There is no requirement from the very nature of rational causation that S be aware of the rational connection that links his belief and desire to his action. However, S may in fact be aware of his belief and desire and the rational connection between them and his action, and this awareness of a rational connection may be efficacious in the production of the action. Let us use the term higher-order rational causation for such cases. Even in higher-order rational causation, S need not be aware of the rational connection between his beliefs and desires, including now his second-order beliefs about first-order rational connections, on the one hand, and the action, on the other. The point here is again to warn against a view of rational mental causation as requiring S always to be aware of the rational connections between the causal relata, for this view seems to generate a vicious regress. At some level rational causation, even higher-order rational causation, is “tropistic,” in that it ultimately involves causal processes not under the rational supervision of the agent.7 Yet it may be an important fact that conscious agents engage in higher-order rational causation, even if this is not required by the very idea of the rational causation of action. Higher-order rational causation may be typical for us; it may be even be essential for conscious action.
208
Bernard W. Kobes
Moreover, action may be turned inward; we can change not only the world but also our minds. Sydney Shoemaker has argued that the rational updating of beliefs and intentions in light of new experience requires that the thinker have mental access to the contents of his current beliefs and intentions.8 And there may be advantages, from the standpoint of efficient design and engineering, for a rational device to have information about its mental states and events and their logical and evidential relations to one another. It would of course be a mistake to portray our mental lives as always deliberative. Many, perhaps most of our thoughts simply occur to us without deliberation. But most of our ordinary bodily actions are likewise unaccompanied by deliberation. Bodily actions may be the products of habit, or alternatively, whimsy or caprice. In a choice between several equally attractive alternatives we are capable of making what seem to us random selections, and many of us vary our routines now and then in spontaneous ways. Yet, tics and involuntary tremors aside, all such bodily actions still seem up to us; we know what we are doing, and we maintain a sense of being their author. The same is true of cognitive agency. Though a thought may simply occur to me, it does not simply occur in me. I know what I am thinking, and obsessive-compulsive and hallucinatory thoughts aside, I have a sense of being not merely the possessor of my thoughts but their author. A thought occurs to me: “Oh, there’s Mark Richard in a red sweater.” I am certainly not responsible for the fact that I thereby represent; I am not even, in the relevant sense, the author of the red visual sensation that triggers and accompanies my thought. But I am the author of the thought, and I know what its intentional content is because I get to say what it is. The naturalness of these locutions, even for a casual thought in passing, marks the sense of cognitive agency with which our explanation of authority begins. Commitment and Spontaneous Beliefs Stuart Hampshire argues in his book Freedom of the Individual that in having an intention to A an agent S often has a spontaneous belief that he will A (Hampshire 1975).9 A spontaneous belief is a belief that is not derived from or causally based on anything else that the believer takes as a premise for the belief. If, for example, I am contemplating whether to buy some flowers on my way home from work, and then form the intention to do so, I have a spontaneous belief that I will buy flowers on my way home from work— a belief that is not derived from or causally based on anything else that I take as a premise for the belief. And such beliefs are often justified. Generally, in a case in which I believe with justification that A-ing is fully within my power, and in which I intend to A, and possess no special reason to doubt that I will carry out my intention, I have a spontaneous justified belief that I will A.
Mental Content and Hot Self-Knowledge
209
For me to have an intention to A, in a case in which I think A-ing is within my power, I have to think of it as up to me, as something I make happen. An intention is a commitment to act, and typically as soon as I am so committed I have the belief that I will A. If in having the putative intention to A I do not eo ipso believe that I’ll A—in a case in which I think it fully within my power and in which I possess no special reason to doubt that I will carry out the intention—then I am not exercising my commitment in the way that is paradigmatic for a genuine intention. So I can and typically do believe that I will A just in virtue of exercising my commitment to A-ing, and not in virtue of a belief about the existence of my intention to A together with a general belief about my likelihood of success in intentions of this kind. Granted, I might also reason that I will A based on my knowledge of the existence of my intention, together with the generalization that I always or almost always fulfill such intentions. I am typically in a position to step back from my practical commitment, acknowledge it thetically, and argue from the existence of the commitment, together with a general premise about how frequently I follow through on such intentions, to the conclusion that I will A. Such arguments, however, only bolster my justification for believing something that I may already know by a different channel. For if I argue from the existence of my intention, together with a general premise about my record of following through on intentions of this kind, then my intention must exist prior to and independently of my carrying out this train of thought. Moreover, my auxiliary premise about my record of following through on intentions of “this kind” would be sensitive to my exercise of the intention itself, the flexing of my resolve. And frequently my belief that I will A is coeval with the sheer exercise of my intention, prior to my knowledge of its existence. The proximal rational cause of my belief that I will A is then a telic state, in Humberstone’s terminology, a state with world-to-mind direction of fit, rather than a thetic state, a state with mind-to-world direction of fit. Let us call the first kind of knowledge that I will A, resulting proximally from the exercise of my commitment to A-ing, hot selfknowledge, and the second kind of knowledge that I will A, resulting from knowledge of the existence of my intention to A, together with general premises about my likelihood of success in intentions of this kind, cold self-knowledge. In the fullness of time, if all goes well, an intention to act is translated into an intentional act, and the belief that I will A becomes the belief that I am A-ing. The earlier belief that I will A is dynamically continuous with my current belief that I am A-ing. If the earlier belief was hot, then typically so is my current belief. That is to say, my current belief that I am A-ing has among its proximal rational causes a telic state, a state with world-to-mind direction of fit, namely, my intention to now A, or to now be A-ing. My belief that I am A-ing need not and typically does not derive from a belief about the existence of my current intention together with a belief about its succeeding.
210
Bernard W. Kobes
Certainly my belief that I am A-ing depends on complex empirical feedback. My belief that I am now buying flowers on my way home from work depends on perceptual feedback on my progress: the sight of the flower shop, the proprioceptive sense of walking into the store, the smell of the flowers, the sounds of the verbal transaction, the heft of the bouquet. Nevertheless, my belief that I am buying flowers remains a hot belief. The perceptual feedback serves a crucial auxiliary function. It enables me to check on my progress. But my buying flowers is a larger event than any of these perceptual atoms by itself, or the particular state or event thus perceived, or even the sum of perceived states or events up to any given time prior to the completion of the project. How do I know that I am buying flowers? Not, I claim, by an inductive extrapolation from the sum of the perceived states or events up to now, nor even typically from perceptual feedback up to now plus knowledge of my intention plus a general premise about my likelihood of success, but rather from perceptual feedback up to now plus the very exercise of my intention. Since its proximal rational causes include a state with world-to-mind direction of fit, my belief that I am now buying flowers is hot. The view then is that hot belief of what I am doing or will do precedes any cold belief I might have about what I am doing or will do. The hot belief is prima facie warranted in the absence of compelling reason to consider the matter coldly. Moreover, it is plausible that my epistemic entitlement to the hot belief does not depend on any disposition I might have to justify it coldly. Indeed, a particular believer might not have the mental capacity coldly to justify his belief that he will A. It is true that if I have reason to think or ought to think that I often change or abandon intentions of this kind, or that circumstances are likely to arise in this case that will cause me to abandon or fail in my intention, then I cannot be said to be justified in my belief that I will A. If my hot belief could not withstand cold scrutiny then it is unjustified. But these are only “defeating” conditions. The relevant necessary condition for my hot belief to be justified is simply the absence of such defeating conditions, not any disposition of mine to check or know that they are absent. Now in belief formation, I am passive with respect to those external facts that are beyond my control, yet it is a project of mine—it is “up to me”—to form beliefs that are faithful to them. Something roughly analogous is plausible about the linguistically expressible concepts that I exercise in thought. It is not up to me which concepts my sociolinguistic environment makes available to me, but given that a certain stock is made available, it is up to me which of those I employ in a given thought. The intuitive point is not so much that, when I exercise a concept, it’s my concept, but rather that it is my exercise of some concept or other, and that I get to say which. This supplies a rough sense in which it is up to me what the content of a given thought is, and interpretation is accordingly constrained. It is possible to explicate this sense in which it is “up to me” which concept I am exercising. Think of the mental relation to the content of self-knowledge—ÈI think that p˘—
Mental Content and Hot Self-Knowledge
211
first of all as telic. My mental relation to the content ÈI think that p˘ has world-to-mind direction of fit—though now, of course, the relevant “world” is itself my mind, or some part or aspect of my mind. The content of self-knowledge is something that I thereby make true. My bearing a telic relation to the content ÈI think that p˘ explains my thinking that p. I have argued elsewhere that, in the case of the belief relation, the representation in the higher-order thought of p as something believed often explains or accounts for the representation-of-p’s playing the functional role of belief (Kobes 1995). Similarly, I now want to suggest, at least in the case of linguistically expressible thoughts, a first-order thought’s playing the role of a constituent in a self-regarding higher-order thought to which S is telically related helps explain or account for its having the first-order content that it has. For in virtue of S’s telic relation to the content ÈI think that p˘, S implicitly forges links between his current thought and thoughts that S attributes or could attribute to others, and thoughts that S remembers or could remember thinking. These links are broadly syntactic in nature. They do not take the form of a revisable hypothesis, as in thinking that the words ‘heather’ and ‘gorse’ express or refer to the same property. Rather, they supply a formal constraint on the interpretation of S’s concepts, via a link resembling the syntactic link of two word-tokens that are counted as belonging to the same type. In this way S implicitly makes it the case that he is tokening a thought of a certain intentional type. For facts about S’s dispositions, given his telic relation to the content ÈI think that p˘, to make such links to the thoughts of others and to his own past thoughts, are among the facts in virtue of which his current thought has the intentional content that it has. A thinker’s ability to think a given thought content will in many cases depend on his ability to use a natural language. Even a thinker who cannot access the relevant wordforms, owing to forgetfulness or to aphasia, may think thoughts that he would not have been able to think had he not once used a natural language. Externalist accounts of intentional mental content, such as Burge’s, key on such linguistically mediated thoughts. The truth conditions of S’s natural-language sentences help to fix the truth conditions of his thoughts, even if S’s mastery of those truth conditions is flawed or incomplete. An account of how the public intentional content of a natural-language sentence can fix the intentional content of a thought must avoid the trap of making the relevant thought or mental act metalinguistic; it is not the thought or the bet that a certain sentence is true. I suggest that we can deepen our understanding of these matters by treating S’s relation to a self-attribution as telic. S participates in public practices of assertion and attribution of belief. Another person’s guileless assertion is taken in this practice to be a sufficient criterion for an attribution of belief to him. Such attributions of beliefs to others have publicly available truth conditions in virtue of being couched in a common natural language.
212
Bernard W. Kobes
Each is in a position to attribute to himself the very thought that he can attribute to others as part of the public practice. S is capable of thinking, ÈR thinks that p˘; plugging himself into the subject position, he can think ÈI think that p˘. S bears a telic relation to that self-attributive content; it has a performative character. S thereby makes it the case that he thinks that p. P. F. Strawson and Gareth Evans have been concerned to argue that the occurrence of the predicate in a self-attribution must be seen as univocal with its occurrence in a corresponding third-person attribution.10 For example, Evans (1982) has argued, by way of his “Generality Constraint,” that the self-attributing thinker must be capable of understanding his attribution as having been made to a third person. Even given this point, it might not be wholly clear why we should not individuate first-order thoughts individualistically and treat all attributions as nonindividualistically typed but the best we have for purposes at hand, or as serving some purpose that cuts across psychological explanation, for example, the transmittal of information.11 On the conception I have been advocating, mental content is “inherited” from the self-attribution to the first-order psychological state. By committing himself to a mental content, S applies to himself a public practice of content attribution that he already engages in with respect to others. The self-attribution has a performative character in virtue of S’s telic relation to its content. Thus the public truthconditions of the attribution get a purchase on the individuation of his own psychological state. Some aspects of a linguistically expressible concept’s causal role help fix it as belonging to a certain intentional type, whereas other aspects of its causal role derive from S’s ability to use the concept properly. (I set aside issues about error and misuse.) A concept has a causal role only in contexts of full thoughts. The current proposal is, in effect, that certain aspects of a concept’s causal role in virtue of the telic higher-order thought—dispositional links to attributions to others and to one’s past self—establish or at least constrain its intentional type. The dispositional links help to fix the concept’s intentional type, and underwrite interpersonal comparability of concepts, and intrapersonal comparability of concepts over time. Intentional type in turn fixes the proper causal roles of first-order thoughts and concepts—and derivatively, their actual roles, insofar as the thinker is able to use his concepts properly. Moreover, by an internal analogue of Hampshire’s thesis about spontaneous knowledge, S knows that his thought has the content p, in the way that one knows things one intends to do or is intentionally doing. The telic relation S bears to the content ÈI think that p˘ is a form of commitment, and in being so committed S acquires a spontaneous belief with the same content, namely, that he does think that p. His self-knowledge about the content of his belief is nonobservational and noninferential, and in particular does not depend on his knowing about his own telic higher-order thought. Thus an analogue of Hampshire’s
Mental Content and Hot Self-Knowledge
213
thesis gives S authoritative knowledge about that over which S has a certain kind of control. Authoritative self-knowledge of the relevant kind is the product of the performative or telic character of the higher-order thought, together with an internal analogue of Hampshire-style spontaneous belief. Authoritative self-knowledge derives, then, not from S’s being well-positioned to observe his own thoughts, and even less from a polite convention of attribution, but from S’s being a cognitive agent, together with the spontaneous knowledge that characteristically accompanies agency. Authoritative self-knowledge in this sense may be only one species of immediate or privileged self-knowledge. There may be kinds of immediate or privileged self-knowledge—for example, of S’s sensations—that are not in the same sense authoritative. For they do not derive from the authority invested in S as a cognitive agent, but only from S’s immediacy or privilege in being the one in whom the sensation is “broadcast.” Smoldering Self-knowledge Let us return to one of Boghossian’s challenges to the extendibility of the Burge–Heil reconciliation story. Boghossian notes that in Burge’s paradigm of basic self-knowledge the higher-order thought must be absolutely coincident with the thought it is about. But, he argues, this is a very special condition; we also have direct knowledge about our immediately past thoughts, and Burge’s account does nothing to explain how that is possible. If, for example, at t1 I think that writing requires concentration, then at t2, a very short time later, I can with authority judge that I just now thought that writing requires concentration. But my judgment at t2 is not self-verifying; it is conceivable that I make the t2 judgment without having made the t1 judgment. The first-order thought at t1 is not a part of the second-order thought at t2. Yet I know what I just now thought noninferentially and authoritatively. So, it appears, Burge’s proposal is incapable of explaining some paradigm cases of direct self-knowledge (Boghossian 1989, pp. 21–22). Let us assume that self-knowledge of past thoughts can be direct, in the sense that it need not and often does not rely on any kind of observation of a memory trace, or inference based on observation. This assumption of course sharpens Boghossian’s challenge, since his point is precisely that the Burge–Heil reconciliation story cannot explain the directness and authority of some of our knowledge of our own past thoughts. Often the thinker simply makes a memory-based judgment of what he thought without making his memory or memory-trace the subject of any observation or inference, and he is epistemically justified in doing so. But what is the intentional content of the memory? It is not simply the content of the first-order thought at t1; that is, it is not simply: Writing requires concentration. For that
214
Bernard W. Kobes
would constitute remembering a general fact, that writing requires concentration, whereas what the thinker remembers is a specific event, namely, that I just now thought that writing requires concentration. So the memory trace that was formed at t1 and comes to consciousness at t2 has the content of a higher-order thought. But how did a memory trace with this content come about? My suggestion is that it is the smoldering remains of a hot self-belief at t1. It came about at t1 as a spontaneous concomitant of the act of judging that writing requires concentration. It was stored in the way that memories are stored, and came to consciousness at t2. (Or perhaps it was conscious all along, active in “working memory”; Boghossian’s case does not specify this one way or the other.) So although self-knowledge at t2 is not, strictly, hot self-knowledge, it nevertheless depends on prior hot self-knowledge, preserved in memory. We do typically remember ordinary conscious but non-self-conscious thoughts. We not only remember their first-order intentional contents, but we remember thinking them. In Burge’s paradigm of basic self-knowledge, as when I judge that I am now thinking, with this very thought, that water is more common than mercury, the higher-order thought is presumably itself a conscious thought. But one theoretically attractive way to extend Burge’s paradigm of basic self-knowledge to conscious but non-self-conscious thought is to suppose that a necessary condition for a mental state or event m’s being conscious is that it be accompanied by a perhaps unconscious occurrent thought to the effect that one is in that very state m.12 When a bit later one remembers that one just now thought that p, that is the coming to consciousness of a formerly unconscious higher-order thought. Thus we would have an explanation of the apparent immediacy of some of our knowledge of past conscious but non-self-conscious thoughts. The idea of smoldering self-knowledge illustrates how the paradigm of basic selfknowledge may be extended to account for the authority, such as it is, of our knowledge of our own past thoughts. Memory at t2 of past thoughts is in a derivative sense authoritative, for it depends on the authority of S’s hot self-knowledge at t1. At the same time, all memory, including memory of past mental events, is a fallible process, and subject to cultivation and neglect. (Trying to recall the thought that caused me to be vaguely depressed a few moments ago, I may think it was the thought of my friend’s misfortune, when in fact it was the thought that a similar misfortune might befall me.) Even when I identify a remembered thought not by its causal relations but by its content, I may simply misremember that content. But when I do correctly remember the content (I recall, “Yesterday I thought: hypochondria is the only disease I haven’t got”), my current thought, being the trace of earlier hot self-knowledge (the day-old trace, that is, of yesterday’s thought, “I think: hypochondria is the only disease I haven’t got”), partakes of its authority. More generally, note that there is an uncomfortable tension between, on the one hand, Boghossian’s objection that the Burge–Heil reconciliation story makes self-knowledge too
Mental Content and Hot Self-Knowledge
215
easy, hence cognitively insubstantial, and, on the other hand, his charge that the account does not explain the range and variety of direct, authoritative self-knowledge. The former objection depends on our self-knowledge being error-prone and fallible, while the latter objection depends on our self-knowledge being direct and authoritative in a wide range of cases. The general strategy of my response exploits this tension and is illustrated by the idea of smoldering self-knowledge. Though I will not carry out the strategy in full detail here, I will outline it briefly. Basic self-knowledge is the central phenomenon that a variety of extended kinds of selfknowledge exploit. Basic self-knowledge is indeed cognitively insubstantial in several senses: it comes easily and errors are difficult or impossible to imagine; it does not depend on observation or inference based on observation; and it is not sensitive to the degree of the subject’s attention, training, self-honesty, or skill. The case of conscious but nonself-conscious occurrent thought can be subsumed under the same rubric by postulating unconscious occurrent higher-order thoughts. The three extensions that Boghossian discusses—knowledge of the mental relation, knowledge of past mental states, and knowledge of standing mental states—build on a foundation of basic self-knowledge but extend it in certain ways. The extensions introduce occasions for error, and occasions for varying degrees and quality of self-knowledge, depending on attention, training, self-honesty, and skill. So extended self-knowledge, unlike basic self-knowledge, is cognitively substantial in several senses, while deriving a partial authority from the basic case. “Slow-Switching” Twin Earth Puzzles Bertrand Russell once suggested that a good test of a philosophical theory is how well it handles puzzle cases. In that spirit, let us suppose that at t1 S, an ordinary Earthling, thinks “I am thinking that water is a liquid.” His self-knowledge is direct and authoritative. Now suppose that he is surreptitiously switched to Twin Earth. He interacts with the speakers there, and with the twater there. Eventually he acquires the concept that Twin Earthlings express by the word ‘water’, that is, the concept twater. But now, at t2, he can still think back to t1. Although he has forgotten nothing, he no longer knows what he thought at t1. Boghossian writes: No self-verifying judgment concerning his thought at t1 will be available to him then. Nor, it is perfectly clear, can he know by any other non-inferential means. . . . But there is a mystery here. For the following would appear to be a platitude about memory and knowledge: if S knows that p at t1, and if at (some later time) t2, S remembers everything S knew at t1, then S knows that p at t2. Now, let us ask: why does S not know today whether yesterday’s thought was a water thought or a twater thought? The platitude insists that there are only two possible explanations: either S has forgotten or he never knew. But surely memory failure is not to the point. . . . It is not as if thoughts with
216
Bernard W. Kobes
widely individuated contents might be easily known but difficult to remember. The only explanation, I venture to suggest, for why S will not know tomorrow what he is said to know today, is not that he has forgotten but that he never knew. Burge’s self-verifying judgments do not constitute genuine knowledge. (Boghossian 1989, p. 23)
Note the purely hypothetical role of the Twin Earth switching scenario in this argument. The argument exploits only the bare logical possibility of Twin Earth switching. If the argument is a good one, it shows that Burge’s self-verifying judgements do not constitute genuine knowledge even if we are justifiably certain that no Twin Earth switching actually occurs. First, let us consider the puzzle in the following version: we stipulate that S knows at t2 that he has at some past time been subject to Twin Earth switching. Call this knowledgeable slow switching, since S knows that he has been switched. (Actually the case depends only on S’s believing or suspecting that he has been slow-switched, or doubting that he has not been slow-switched.) But S simply does not know if he is now on the same planet that he was on at t1 or not. In particular, he does not know whether he was switched before or after t1. He uses both ‘water’ and ‘twater’. But he remembers the t1 event, and in fact he wonders, “Was I thinking at t1 about water or twater?” This seems to be the version of the puzzle that Boghossian has in mind. But S does have available to him the same concept that was available to him at t1. He is perfectly free to exercise the same concept at t2. At t2, thinking back to t1, he may invent a new word for the substance he remembers so well, say, ‘mwater’. He knows that mwater is either water or twater. He does not know which, but this is no defect in the concept mwater. In fact, there are a great many things he knows about mwater. For example, he knows that around the time of t1 it often rained mwater. And he knows that mwater is a clear colorless liquid that quenches thirst. Are we to deny him this knowledge just because he does not know how to re-express it in terms of the words ‘water’ or ‘twater’? That would be highly uncharitable; it seems to be perfectly good knowledge of the external world. And he knows one more thing: he knows that at t1 he thought that mwater was a liquid. He knows this because he has a memory trace that results from the hot self-knowledge he had at t1. This too is perfectly good smoldering self-knowledge. Even at t1, it may interact with other things he believes about the world in inference and practical reason. Though he does not know how to re-express this bit of self-knowledge in terms of ‘water’ or ‘twater’, it should nevertheless be reckoned among his cognitive achievements. Let us suppose, then, that in Boghossian’s puzzle S never finds out that he was switched, and does not even suspect that he might have been switched. Call this, in contrast with the first case, ignorant slow switching. At t1 S had the concept water; at t2, he has the concept twater. At t2, S thinks a thought that he would express as, “At t1, I thought that water is a liquid.” It seems to us that he does not really know at t2 what he thought at t1.
Mental Content and Hot Self-Knowledge
217
Yet he has forgotten nothing. Does Boghossian’s argument, concluding that his thought at t1 did not constitute genuine knowledge, go through now? No. For one thing, Boghossian’s “platitude” is plainly false. According to the platitude, quoted above, if S knows that p at t1, and if at (some later time) t2 S remembers everything S knew at t1, then S knows that p at t2. But S might acquire a new belief q that causes him to lose his old knowledge that p, even though S forgets nothing. Suppose that q constitutes evidence for the proposition ÿp, evidence that is misleading, as it happens, since p is true. Sherlock Holmes might have sufficiently strong evidence to know that the countess murdered the earl with a knife. Then he comes to believe that the countess was on the train to London at the time the earl was stabbed. Now Holmes no longer knows that the countess is the murderer, but he has forgotten nothing. Note that the counterexample is compatible with the new belief q’s being itself true. Even if q is true, its misleading evidential bearing on ÿp can undermine the status of S’s belief that p as knowledge. The countess was indeed on the train. Holmes’s coming to learn this undermines his justification for his belief that the countess is the murderer. Moreover, the counterexample is compatible with S’s continuing to believe that p. Holmes irresponsibly persists in his belief that the countess committed the murder. Luckily for Holmes, he is right: the countess had set up a fiendish knife-throwing contraption hooked up to a timing device. But Holmes’s belief has lost its status as knowledge. So Boghossian’s “platitude” is quite false. In slow switching, we are not forced to choose between S’s having forgotten something he once knew and S’s never having had genuine knowledge. The lesson of our reflections on Holmes is that, if S no longer knows the content of what he once thought, this may be because some intervening event, such as ignorant slow switching, has destroyed the status as knowledge of some once-genuine knowledge that is still remembered. But there is more to be said. At t2, S has a self-belief that he could express by saying, “At t1, I thought that water is a liquid.” Does the thought S would express in this way constitute memory-based self-knowledge? On the one hand, we are tempted to say that it does, on the grounds that it results from a veridical memory trace of earlier self-knowledge. On the other hand, we are tempted to say that it does not, on the grounds that at t2 S is exercising the concept twater, a concept he did not possess at t1. The answer depends, I think, on recognizing that S’s thought at t2, which he would express by saying “At t1, I thought that water is a liquid,” is equivocal. Unbeknownst to S, his thought at t2 has two propositional contents simultaneously. The thought at t2 is a result of S’s cognitive agency at t2 and therefore cannot escape including an exercise of the Twin Earth concept twater. At the same time, S’s exercise at t2 of the concept he expresses by ‘water’ was triggered by a memory event that is rooted in the Earth concept water. S’s concept has historical roots in two speech communities, and in distinct kinds
218
Bernard W. Kobes
of clear colorless drinkable liquid on two planets. So it is an equivocal concept, no matter how much twater has flowed under the bridge.13 We may represent S’s total corpus of beliefs at t2 by way of a fragmentation strategy (cf. Lewis 1982). S’s total belief corpus at t2 is broken into two large overlapping fragments. It appears to S as only one coherent belief corpus, because the concepts in terms of which he represents the corpus to himself include the same equivocal concept as the (first-order) corpus itself. If S had n concepts that were independently two-ways equivocal, perhaps through a series of n independent surreptitious switches to different linguistic environments, then S’s belief corpus would be broken into 2n distinct fragments. When a belief corpus is fragmented in this way, we may say that S has weak selfknowledge (i.e., self-knowledge in the weak sense) if he has self-knowledge according to at least one fragment, and that S has strong self-knowledge if he has self-knowledge according to all fragments. Then in the case at hand we can say that S has weak selfknowledge at t2 about his belief at t1. For on one fragment he has a belief that at t1 he thought that water is a liquid, which is true, but on another fragment he has a belief that at t1 he thought that twater is a liquid, which is false. In this way we can account for and honor two intuitions about Boghossian’s puzzle case: that S holds an erroneous belief at t2 about his thought at t1, and that he in some sense still knows what he thought at t1. But the conclusion that S had at t1 no genuine self-knowledge is undermined. Note that on this account, S has strong self-knowledge at t2 about his thought at t2, selfknowledge that he might express by saying, “I believe that I judge that at t1, I thought water is a liquid.” For on one fragment he believes that he judges that at t1 he thought that water is a liquid, which is true; and on another fragment he believes that he judges that at t1 he thought that twater is a liquid, which is also true. More generally, the exercise of an equivocal concept is no barrier to strong, hot selfknowledge. Some time after a surreptitious switch from Earth to Twin Earth, given acculturation, S will have both the concept water and the concept twater, and will unwittingly exercise both simultaneously whenever he thinks a thought that he would express using the word ‘water’. A cognitive agent in this situation has strong, hot self-knowledge. Of course S fails to know that his thought is equivocal, and this is a significant limitation on his discursive self-knowledge. But this is just another instance of the general point that a thinker can lack knowledge of the individuating conditions of his thought, even while those individuating conditions help to fix the identities of both his first- and second-order thoughts.14 My account of the two versions—“knowledgeable” and “ignorant”—of the slowswitching scenarios illustrates and supports the theory presented earlier of authoritative self-knowledge of mental content. According to that theory, the higher-order thought involves a telic relation to its content, and works by implicitly establishing links between
Mental Content and Hot Self-Knowledge
219
the self-attribution and attributions to others and to one’s past self. Now in the case of knowledgeable slow switching, S establishes a link between his current t2 thought (about his past t1 thought) and the remembered t1 higher-order thought itself. The concept, which S labels ‘mwater’, has the same intentional content as our concept water, and (unbeknownst to S) it is true of all and only instances of H2O. Moreover, S takes his currently exercised concept to have its intentional type entirely fixed by that link, and S excludes links to the concept that he ordinarily, these days, expresses by the word-form ‘water’. And it is precisely because of the performative character of S’s higher-order thought, in virtue of his telic relation to its content, that he can exercise a concept, like mwater, thus selectively linked. On the other hand, in the case of ignorant slow switching, S establishes links between his current t2 thought (about his past t1 thought) and both the remembered t1 higher-order thought itself, and other current t2 thoughts. Because he has the authority to do so, his current thought becomes equivocal; it has both contents simultaneously. It is precisely because of the performative character of S’s self-attribution, in virtue of his telic relation to its content, that we are in no position to discount one set of links as mistaken. As reasonable interpreters of S we must therefore acknowledge both, and ascribe to him an equivocal concept. The performative character of self-attributions comes through with unusual clarity, I venture to suggest, in the unusual contexts of slow switching scenarios. Boghossian imagines a slow-switching scenario in which Peter, hiking in northern New Zealand, encounters the tenor Luciano Pavarotti floating on the pristine waters of Lake Taupo. They chat briefly. Some years later Peter is switched to Twin Earth, where the word-form ‘water’ refers to XYZ, and ‘Pavarotti’ and ‘Lake Taupo’ refer to the twins of Pavarotti and Lake Taupo respectively. Peter is unaware of the switch, so this is a case of what I have called “ignorant” slow switching. Over decades he is insensibly acculturated to Twin Earth words and concepts, but his memorable encounter with Pavarotti stays with him. He reads and understands newspaper reports using the name ‘Pavarotti’ [which are about Twin Pavarotti], but his vivid and accurate memories of the long-ago encounter are about Pavarotti floating on water (Boghossian 1994, pp. 36–39; 44–45). It would seem, then, that some of Peter’s tokens of ‘Pavarotti’ refer to Pavarotti, while others refer to Twin Pavarotti, and some of his tokens of ‘water’ refer to H2O, while others refer to XYZ. Yet Peter himself is unaware of this, and could become aware of it only by empirical investigation, and not by any a priori reflection on his mental life no matter how rational and thorough. Or so Boghossian interprets the case. The case shows, he claims, that externalist accounts of mental content conflict with a highly plausible principle of “transparency of difference,” namely, that if two of a thinker’s thoughts possess distinct intentional contents, then the thinker must be in a position to know a priori that they do. Worse, the case seems to show that externalism about mental content allows for the
220
Bernard W. Kobes
possibility that two of a thinker’s thoughts of the same syntactic type might have distinct contents, without the thinker being in a position to know a priori that they do.15 This would be an intolerable result for externalism. It allows, as Boghossian notes, the possibility of a perfectly rational thinker committing himself to logically invalid arguments. Reflecting on the properties of the stuff he calls ‘water’, Peter thinks, “Whoever floats on water, gets wet” (P1). As a general quantified thought by a naturalized Twin Earth resident, this true thought employs a concept referring to XYZ. Peter also thinks, “Pavarotti once floated on water” (P2). Based as it is on a vivid and veridical memory of his distant encounter with Pavarotti, this true thought employs a concept referring to H2O. Peter concludes, “Pavarotti once got wet” (C). The argument is invalid, since the two premises equivocate on a key concept. But it will seem valid to Peter, and no amount a priori reflection could reveal to him that it is not. Our account of how to attribute contents in cases of ignorant slow switching blocks this intolerable result in a plausible and externalistically acceptable fashion. Peter’s thoughts, unbeknownst to him, have equivocal intentional contents. His thought “Whoever floats on water, gets wet” expresses two propositions at once, one containing the relational property floats on twater, the other the relational property floats on water. His thought “Pavarotti once floated on water” expresses two propositions at once, one involving Pavarotti, the other involving the tenor’s twin. “The” argument is really two arguments, both valid (and probably both sound as well, if we may assume that Twin Pavarotti once floated on XYZ). Externalism is therefore not committed to Peter’s being disposed to reason invalidly. This interpretation of Peter as thinking equivocal thoughts is not an ad hoc defense of externalism. Although Peter’s thought “Pavarotti once floated on water” is memory based, its subject concept is Peter’s, and he gets to say whether he is exercising the same concept as he did when reading yesterday’s newspaper report about the man it called ‘Pavarotti’ (i.e., our tenor’s twin).16 Our theory articulates this plausible and intuitive idea as follows: In making his memory-based judgment Peter is a cognitive agent, and therefore bears a telic relation to the content “I am thinking that Pavarotti once floated on water.” This thought of Peter’s derives its content at least partly from Peter’s dispositions to link it to attributions that he might make to others, and to himself at other times, and not solely from the temporally distant man and liquid that are the causal provenances of his firstorder memory trace. The telic higher-order thought has a performative character, and makes it the case that Peter’s first-order thought has the content—or, in this case, contents—that it does. Given the phenomenon of spontaneous belief attendant on telic relations, Peter will know the contents of his thoughts, but this “strong” self-knowledge will also be equivocal in a way that masks from rational view the equivocation of his firstorder thoughts.
Mental Content and Hot Self-Knowledge
221
There is, I think, no clear sense to be made of the idea that two of a thinker’s (nonindexical, nondemonstrative) thoughts or concepts might be of the same syntactic type but have distinct contents. For the clearest notion we have of two thoughts or concepts being of the same syntactic type derives from the thinker’s dispositions to link them in characteristic ways in higher-order thoughts in which they appear, and such links will constitute the thoughts or concepts as having the same intentional content. One manifestation of this is a thinker’s disposition to link concepts in deductive reasoning. Consider Peter’s linking, in the argument above, the ‘water’ concept of (P1) with the ‘water’ concept of (P2). Intuitively, Peter’s linking them in the argument manifests not merely his belief that he is exercising concepts with the same content, but his intention to do so. Of course, the same might be said about all equivocation in argument. One who thinks, “I have a duty to do what’s right; I have a right to offer my frank opinion of your hat; therefore I have a duty to offer my frank opinion of your hat,” in some sense intends the tokens of his concept right in the two premises to have the same intentional content. In many such cases the thinker can be brought to see his error by calling his attention to other examples that make his equivocation in this example salient to him. So a particular intention to exercise the same concept in two premises of an argument is not sufficient to constitute them either as belonging to the same syntactic type or as having the same intentional content. In Boghossian’s case, however, where it is agreed that the syntactic type of the two ‘water’ tokens is the same, Peter’s linking the ‘water’ concepts in his argument about Pavarotti is but one manifestation of a complete set of relevant dispositions. For the set of attributions to which Peter is disposed to link the higher-order thought corresponding to (P1), as containing a ‘water’ concept of the same type, is identical to the set of attributions to which he is disposed to link the higher-order thought corresponding to (P2). The performative characters of these telic higher-order thoughts therefore constitute the contents of the ‘water’ concepts of (P1) and (P2) identically. Loar’s Objection Brian Loar objects to the Burge–Heil reconciliation story on the grounds that it cannot be used by S to construct a non-question-begging, wholly a priori justification of the pragmatic inference from the (presupposed) existence of Socrates and hemlock to S’s now thinking that Socrates drank some hemlock. To summarize briefly our earlier discussion: As a precondition of using the Burge–Heil reconciliation story, which assures S that all thoughts of a certain form are self-verifying, S would first have to know that he is thinking a thought of the relevant self-verifying form. In order to know, by way of applying the reconciliation story, that he is thinking “Socrates drank some hemlock,” S would first have to know that he is thinking “I am thinking that Socrates drank some hemlock.” But
222
Bernard W. Kobes
given the coherence of the doubt that he stands in the appropriate external relations to Socrates and to hemlock, S would beg the question if he were simply to assume that he knows this a priori, without reliance on empirical observation of the world. I extended Loar’s objection to include the charge that S would be launched on an infinite regress were he to attempt to construct a wholly a priori demonstration that he is thinking “I am thinking that Socrates drank some hemlock.” And I suggested that the issues are substantially the same if we substitute for Loar’s ‘a priori’ my favored terms ‘direct[ly]’ and ‘authoritative[ly]’. In responding to Loar, we should first distinguish two ways in which the Burge–Heil reconciliation story may be used. The reconciliation story may be used by theorists in constructing an account of S’s epistemic entitlement to his belief about his own thought. Alternatively, the reconciliation story may be used by S himself, as part of his epistemic justification of his belief about his own thought. In the latter case, S may actually run through a bit of explicit reasoning that includes the reconciliation story, or he may merely be disposed to use the reconciliation story in that way, and be epistemically justified in his self-knowledge in virtue of that disposition. But a correct account of S’s entitlement may be such that S is not even disposed to run it through; it may not be accessible to S, and if it is accessible, he may not agree with it.17 In what I shall call the base or level-0 case, S does not use the Burge–Heil reconciliation story, nor need he be disposed to do so. Rather, that story is used by the theorist to explain S’s entitlement to his self-knowledge. S simply thinks a thought with content p. In so doing, S is a cognitive agent; his thinking that p is relevantly like an intentional act, and not merely something that happens to him. Thus S bears a telic relation to the content ÈI think that p˘. This spontaneously (in Hampshire’s sense) yields a thetic relation to the content ÈI think that p˘, and this thetic relation is S’s authoritative self-knowledge. Because it arises spontaneously, it is noninferential. Yet S is entitled to it. The Burge–Heil reconciliation story, augmented by the above account of first-person authority as the product of the performative character of telic higher-order thoughts and Hampshire-style spontaneous belief, explains S’s entitlement. If S realizes that he has self-knowledge, it may occur to him to wonder about its nature; he may wonder how his self-knowledge could be authoritative. This gives rise to the level1 case, in what will turn out to be a ladder of justifications. S may read Burge’s and Heil’s articles and may want to construct an explicit justification for his own self-knowledge. So, S proceeds: He thinks a thought with content, ÈI think that p˘. So far we do not understand how S could have the basis for straightforwardly applying the Burge–Heil reconciliation story. For all S has is the thought, that he, S, thinks that p, and this higher-order thought, even if it is a belief, is not yet presented as something that S can think about. He can think with it, and use it as a premise in reasoning, but as a
Mental Content and Hot Self-Knowledge
223
premise in reasoning it does not interact with the Burge–Heil observation that thoughts of a certain form are self-verifying and are therefore true, since the thought that S’s thought is about only has the form p. From the thought ÈI think that p˘, S is not in a position straightforwardly to infer that he thinks any sort of reflexive or higher-order thought, and that is what he needs. Yet intuitively it seems that S does have some sort of basis for applying the Burge–Heil reconciliation story, and his thought ÈI think that p˘ is key to the application. But that thought will not serve usefully here as a premise in reasoning. Rather, it is a performance, just as the thought that p was a performance in the level-0 case, and it is in some nonobservational sense on display, or perhaps better, it is the subject matter of reasoning, rather than a premise in reasoning. The reader may perhaps have anticipated my account of the level-1 case: In thinking ÈI think that p˘, S is a cognitive agent. Thus S bears a telic relation to the content ÈI think that I think that p˘. This spontaneously (in Hampshire’s sense) yields a thetic relation to the content ÈI think that I think that p˘. This is the relevant premise that figures in S’s application of the Burge–Heil reconciliation story. This premise constitutes authoritative noninferential self-knowledge, and we as theorists have explained S’s entitlement to it. From it, together with the Burge–Heil reconciliation story, S infers that his own reflexive thought is true. That is, S infers, ÈMy thought, I think that p, is true˘; or, more simply, ÈI think that p˘. And this is the self-knowledge of first-order thought for which S has now supplied an explicit justification. There is still, as always, an opening for a new philosophical anxiety. For S may realize, under the influence perhaps of reading Loar’s article, especially the part quoted above, that his carefully constructed justification of his self-knowledge depends on his belief ÈI think that I think that p˘. Does this belief, perhaps, depend on some kind of nonauthoritative observational process? As theorists we have constructed an account of S’s entitlement, but of course S does not know that. So now S may wish to construct an explicit justification of this belief. Call this the level-2 case. So, S proceeds: He thinks a thought with content, ÈI think that I think that p˘. Again, S is in no position simply to use this current thought as a straightforward premise in explicit reasoning with the Burge–Heil account. If S’s goal is as stated, namely, to construct an explicit justification for the premise of the explicit reasoning of level-1, this current thought is of the wrong form to interact with the Burge–Heil account. If his (revised) goal is to take a short-cut, and reason from the current thought together with the Burge–Heil account to ÈI think that p˘, then he lacks a justification of his starting point, and we as theorists lack an account of his entitlement to it. As before, we must think of the current thought as a performance, and not as a premise. Intentionally thinking the current thought, S bears a telic relation to ÈI think that I think
224
Bernard W. Kobes
that I think that p˘, and spontaneously he bears a thetic relation to the same content. This is authoritative, noninferential self-knowledge, and we as theorists can explain his entitlement to it. S uses it as a premise, together with the Burge–Heil reconciliation account, to infer ÈI think that I think that p˘—the premise of the level-1 case. A second application of the Burge–Heil account allows S to infer ÈI think that p˘, which constitutes selfknowledge of first-order thought content. If all that a reflective and epistemically cautious thinker S has at his command is the unadorned Burge–Heil observation that thoughts of a certain form are self-verifying, then he can never fully reassure himself of the directness and authority of his self-knowledge. For every time he climbs a rung of the ladder of explicit justifications, he lays himself open to a new Loar-inspired anxiety, focusing on the status of his new noninferential starting point, which will spur him to climb the ladder one more step, and so on ad infinitum. As theorists we can construct an account of S’s entitlement to his starting point at any level, including level-0 where S engages in no justificatory reasoning at all. So an unreflective thinker S does indeed have direct and authoritative self-knowledge, and so does a reflective thinker if his epistemic scruples are not so great as to cause him to renounce his birthright, so to speak, to direct and authoritative self-knowledge. But S can never fully reassure himself, by way of the unadorned Burge–Heil observation, that he has such direct and authoritative self-knowledge, for he will always anticipate a doubt attendant on the next rung of the ladder of justifications. Nor will a reflective and cautious thinker be able to fully reassure himself if he simply adds to the unadorned Burge–Heil observation—unsupplemented by our current account— the distinction between entitlement and justification. If S is trying to reassure himself, then he is at once both thinker and theorist, and this blurs the distinction between entitlement and justification. What S requires is something in the content of the account of his entitlement, and not merely in the fact of its being an account of entitlement, as opposed to a justification, that will permit him, scrupulous as he is, to stop climbing the ladder of justifications. The larger picture presented here, however, should give S the epistemic confidence he needs to stop climbing the ladder of justifications. He will realize that however high he climbs, he will inevitably have to start with a mental act. A mental act per se requires no epistemic justification, if its role in the argument is to supply the subject matter of thought, rather than a premise in reasoning. S will realize too that his thought about what his mental act thus supplies is authoritative because of both the performative character of his telic higher-order thought—what concept he exercises on an occasion is up to him—and the spontaneous belief he has of contents to which he stands in telic relation. So if S thinks of himself as a cognitive agent, and as entitled to a mental act as a starting point, he will see that further climbing of the ladder of justifications is entirely optional.
Mental Content and Hot Self-Knowledge
225
Acknowledgments This paper first appeared in Philosophical Topics, vol. 24, no. 1 (spring 1996), pp. 71–99. Permission to reprint is hereby gratefully acknowledged. For helpful comments on some of these ideas I thank Brad Armendt, Tyler Burge, Stewart Cohen, Brian Loar, and Steven Reynolds. Parts of this work were presented at a conference in October 1993 at Simon Fraser University on the work of Tyler Burge, at the University of California at Davis, and at the 1996 meeting of the American Philosophical Association, Central Division. I profited from the discussion on those occasions. This work was supported by a Faculty Grant-in-Aid from Arizona State University. Notes 1. The representation that p may be thought of as a mental event of thinking that p, individuated by its intentional properties. I do not mean to be committed to a language-of-thought hypothesis, according to which representations are individuated by physical or syntactic features that are constituted prior to or independently of intentional properties. 2. See Putnam (1975), pp. 215–271, and Burge (1979), pp. 73–121. 3. See Burge (1988), pp. 649–663, and Heil (1988), pp. 238–251. 4. It should be noted that Loar puts knowledge of one’s own references at the center of the topic of epistemic authority. He is primarily concerned with the question of how a thinker can make a priori inferences like: Socrates exists ‘Socrates’, as I use it, names Socrates. This inference is, as Loar shows, pragmatically valid. That is, if an utterance of the premise is true in a certain context, then an utterance of the conclusion would be true in the same context, even though the propositions expressed by the premise and conclusion are related only contingently. Yet, Loar argues, a thinker who knows that this inference is pragmatically valid cannot use that knowledge to reassure himself that he can make the inference a priori. But rather than present Loar’s argument on this point, I turn directly to a variant of the argument that Loar presents (on pp. 63–64) for our topic, the case of knowledge of one’s own mental contents. 5. Loar is concerned with the apriority of a certain inference from existence premises; my focus is on the directness and authority of self-knowledge, which I see as typically noninferential. Moreover, as noted earlier, Loar’s main concern is with knowledge of one’s own references, which he sees as a more basic problem than knowledge of mental content; his argument against the Burge–Heil account is presented as a corollary consideration. Loar’s own solution to the puzzles he raises requires that ordinary object-level concepts have reflexive implications. Space prohibits a critical discussion of Loar’s views, but I take it to be an advantage of my solution to Loar’s puzzle, presented below, that it does not treat ordinary object-level concepts as about themselves in any sense. 6. See also Boghossian (1994), pp. 33–50. 7. For the term ‘tropistic’ and related discussion, see Mark Johnston (1988), pp. 63–91. 8. See Shoemaker (1988), pp. 183–209; (1990), pp. 187–214; (1991), pp. 127–149. 9. Bas C. van Fraassen calls attention to Hampshire’s view in van Fraassen (1995). 10. See P. F. Strawson (1959), p. 99, and Evans (1982), pp. 224–235. 11. For a defense of something like this view, see Loar (1988), pp. 99–110. 12. See my “Telic Higher-Order Thoughts and Moore’s Paradox” (1995) for a critical discussion of David Rosenthal’s HOT theory of consciousness. Note that for present purposes I require an occurrent unconscious
226
Bernard W. Kobes
HOT only as a necessary condition, not as an explanation or analysis, of ordinary non-self-conscious conscious thought. 13. Twin Earth switching is not necessary to generate equivocal thoughts. Slow switching between America and England may suffice. Suppose S acquires the term ‘endive’ while growing up in America and subsequently moves to England, where ‘endive’ refers to a different vegetable (or so let us suppose; see Stich 1983, p. 63). Acculturation may occur over a period of decades without S’s learning anything about what the English called ‘endive’ that would distinguish it from what Americans call ‘endive’. In that case, S’s thoughts containing his concept endive would be equivocal; they would have two intentional contents at once, without S’s knowing it—indeed, without S being in any position to know it a priori. Moreover, in my view, there are many actual cases of equivocal thoughts that are generated by a thinker’s simultaneous allegiance to both a scientific and a nonscientific linguistic community, cases where the thinker does not realize that the meanings of the relevant term or concept differ in the two communities. Think of someone who uses ‘fruit’ (do tomatoes count?) or ‘nut’ (do peanuts count?) and who maintains equal and simultaneous linguistic allegiance to the speech communities of both botanists and grocers. Such a thinker may think thoughts that express two propositions at once, without his being in any position to know this a priori. 14. In fact, I think, all hot self-knowledge is strong. Moreover, all smoldering self-knowledge at t2 of what is in fact an ambiguous thought at (some earlier time) t1 is also strong, unless further ambiguity in S’s thought has been introduced in the intervening period. Smoldering self-knowledge can be weakened only by the introduction of new conceptual ambiguity in the period between t1 and t2. 15. Boghossian seems to think that this is just what a violation of the principle of transparency of difference would have to consist in. This is a mistake; simpler violations are imaginable. Suppose S is fairly confident that ‘chicory’ and ‘endive’ are synonyms, that they refer to the same herb and express the same property. In fact they refer to different herbs, and on externalist views of mental content, S’s concept chicory has an intentional content distinct from that of his concept endive. No matter how rational and reflective S might be, he is in no position to know a priori that his concepts have distinct intentional contents. So much the worse, an externalist might say—ought to say, I think—for Boghossian’s principle of transparency of difference. (I likewise reject Boghossian’s principle of “transparency of sameness,” which turns out to be equivalent to transparency of difference given some plausible assumptions. These matters deserve more discussion than I am able to give them here.) But the Pavarotti case mounts a serious challenge to externalism, involving as it does thought tokens of the same syntactic type. 16. Whether he does so say, in the relevant sense, is not a simple matter of asking him; see the discussion below about equivocation in argument. 17. For more on entitlement versus justification, see Burge (1993), pp. 457–488.
References Anscombe, G. E. M. 1957. Intention. Oxford: Blackwell. Boghossian, Paul Artin. 1989. Content and Self-Knowledge. Philosophical Topics 17: 5–26. ———. 1994. The Transparency of Mental Content. Philosophical Perspectives 8: 33–50. Burge, Tyler. 1979. Individualism and the Mental. Midwest Studies in Philosophy. 4: 73–121. ———. 1988. Individualism and Self-Knowledge. Journal of Philosophy 85(11): 649–663. ———. 1993. Content Preservation. Philosophical Review 102: 457–488. Descartes, René. 1971. The Principles of Philosophy. In Descartes: Philosophical Writings, G. E. M. Anscombe and P. T. Geach (eds.). New York: Bobbs-Merrill. Evans, Gareth. 1982. The Varieties of Reference. New York: Oxford University Press. Hampshire, Stuart. 1975. Freedom of the Individual. Princeton, NJ: Princeton University Press. Heil, John. 1988. Privileged Access. Mind 97: 238–251. Humberstone, Lloyd. 1992. Direction of Fit. Mind 101: 59–83.
Mental Content and Hot Self-Knowledge
227
Hume, David. 1888. A Treatise of Human Nature. L. A. Selby-Bigge (ed.). Oxford: Oxford University Press. Johnston, Mark. 1988. Self-Deception and the Nature of Mind. In Perspectives on Self-Deception, Brian P. McLaughlin and Amelie Oksenberg Rorty (eds.). Berkeley: University of California Press. Kobes, Bernard W. 1995. Telic Higher-Order Thoughts and Moore’s Paradox. Philosophical Perspectives 9: 291–312. Lewis, David. 1982. Logic for Equivocators. Noûs 16: 431–441. Loar, Brian. 1988. Social Content and Psychological Content. In Contents of Thought, Robert Grimm and Daniel Merrill (eds.). Tucson: University of Arizona Press. ———. 1994. Self-Interpretation and the Constitution of Reference. Philosophical Perspectives 8: 51–74. Putnam, Hilary. 1975. The Meaning of ‘Meaning’. Mind, Language, and Reality: Philosophical Papers, volume II. Cambridge: Cambridge University Press. Shoemaker, Sydney. 1988. On Knowing One’s Own Mind. Philosophical Perspectives 2: 183–209. ———. 1990. First-Person Access. Philosophical Perspectives 4: 187–214. ———. 1991. Rationality and Self-Consciousness. In The Opened Curtain: A U.S.–Soviet Philosophy Summit, Keith Lehrer and Ernest Sosa (eds.). Boulder: Westview Press. Stich, Stephen. 1983. From Folk Psychology to Cognitive Science. Cambridge, MA: MIT Press. Strawson, P. F. 1959. Individuals. London: Methuen. van Fraassen, Bas C. 1995. Belief and the Problem of Ulysses and the Sirens. Philosophical Studies 77: 7–37. Woodfield, Andrew (ed.). 1982. Thought and Content. New York: Oxford University Press.
11
Phenomenal Intentionality as the Basis of Mental Content
Brian Loar The mental or psychological content of a thought is a matter of how it conceives things; that is what we hope to grasp, at least approximately and in part, when we try to understand another person. We want to know not merely what her thoughts represent as it were impersonally, but also how they represent things to her. A person’s thoughts represent things to her—conceive things—in many ways: perceptually, memory-wise, descriptively, by naming, by analogy, by intuitive sorting, theoretically, abstractly, implicitly and explicitly. These various manners of conceiving have something in common: They have intentional properties, and they have them essentially. The conceiving cannot be pulled away from the intentional properties, in our ordinary reflexive understanding of them. But this creates a problem. It is not unnatural to suppose that conceivings are in the head. So if the intentional properties of conceivings are essential to them, intentionality must be in the head as well. The problem is that there are fairly compelling externalist reasons to the opposite conclusion. Yet it seems to me that there must be something right about the internalist thesis and the intuition that backs it, something quite basic to our understanding of the mental. This is what I will try to make coherent. Mental content has often been supposed to be what “oblique” that-clauses capture. That would lead directly to a considerable difficulty in the idea of internal intentionality. For that-clauses capture references, and the references of our outward-directed thoughts are— according to the most believable theories1—determined by external relations. There is a quick way to deal with this difficulty, and I hereby adopt it. Mental content is in fact individuated independently of that-clauses. This seems to me to follow from the semantic behavior of that-clauses together with a basic constraint on mental content or ways of conceiving—what is often called “Frege’s constraint.” Something counts as a judgment’s content only if we cannot make sense of a person’s judging both it and its negation— unless she in some way compartmentalizes those judgments. You can make sense of a person’s judging that Paderewski plays well and at the same time judging that Paderewski does not play well even though the two beliefs have the same reference and draw on the same public name. Nothing semantic distinguishes the ordinary meanings of those thatclauses except the negation. Given Frege’s constraint, this means that mental content is individuated more fine-grainedly than the interpersonally shared “oblique” content of certain that-clauses. I rather think the phenomenon is all-pervasive, that for virtually any that-clause a similar underspecification of content can be shown.2 A closely related point is this. Consider any perceptually nuanced conception of mine. I can invent a neologism to express that conception, and use it in self-ascribing that-clauses. But the that-clauses are then secondary: What matters is my reflexive grasp of the perceptual concept, its psychological content. That-clauses as they are standardly used apparently capture too little
230
Brian Loar
information, even on oblique interpretations, and that information is not of the right sort: That-clauses are more about socially shared concepts and their referents than about the various perceptually based and other ways in which thoughts conceive their referents. They are not especially psychologically informative. If mental content is accessible and is not literally expressed by that-clauses, how does it get conveyed? Typically in the gaps between the words. Suppose you say that Guido thinks that the woman over there resembles Greta Garbo, and you say this while he has the woman in full view. I understand you to mean that Guido’s thought picks her out visually. That visual mode of presentation is a constituent of (what I mean by) the mental content of his thought. That we might invent a word to capture just that highly specific visual mode of presentation, and insert it in a that-clause, is not interesting—nor is it even particularly interesting that we can say ‘the person he is looking at’. Guido’s thought involves, among other factors, a visual mode of presentation, and we conceive it independently of what is mentioned in that-clauses. But this is neutral between internalist and externalist views of mental content. For it is compatible with “neo-Fregeanism,”3 the idea that for example perceptual modes of presentation are to be individuated object-dependently and property-dependently. The present point, though, is simply to put space between mental content and that-clauses. Our conceptions of mental content have a life of their own apart from that-clauses—there are for example perceptually based demonstrative concepts, as we intuitively understand when we think about Guido.4 Conceptions of mental content in the analytic tradition have tended to be phenomenologically impoverished, largely because of the emphasis on language and reference. And when we turn to the phenomenology, as I will try to show, we do get a grip on internal intentionality. A compelling intuition about mental life sees it as a stream of conscious thoughts, feelings, and perceptions. This is not all or even perhaps the larger part of the mind. But it is central to our founding conception of the mental. When we conceive these various conscious states, moreover, we conceive them as intentional. The stream of conscious thoughts, memories, and perceptions seems to have a life of its own that is constituted independently of its external environment. This is intuitively supported by an obvious thought experiment. Apparently I can imagine what it is like to be an isolated brain that is a physical duplicate of my own brain. What I imagine includes not just that brain’s nonintentional phenomenal states, its flutters and pains, but also states and events that correspond to my own outward-directed thoughts and perceptions. I imagine my isolated twin’s states and events as subjectively representing things in the same manner as those thoughts and perceptions of mine. The intuition supports the view that my own mental stream’s intentional features—even those of its outward-directed thoughts—are constituted independently of my actual situation in the world. (Note well that I have said ‘inten-
Phenomenal Intentionality as the Basis of Mental Content
231
tional features’ and not ‘references’.) This is not to say that the seeming imagining of the isolated brain’s intentional states proves there is such a thing as internal intentionality. But it surely makes one wonder if we can make sense of the idea, make a case for its coherence. The reader will reasonably want to know what is meant by ‘intentional’ if not ‘referential’. Let me say for the time being that the internal intentionality of perceptions and thoughts consists in their apparent directedness, in their purporting subjectively to refer in various complex ways. This is, according to what follows, an ineliminably phenomenal feature that is shared by my and my isolated twin’s states as I imagine them. Why care if a phenomenological conception of internal intentionality can be made sense of? It is there for the noticing; and we have a wrong philosophical view of our intuitive conception of the mind if we persuade ourselves in the abstract that internal intentionality cannot be there. Does this matter to commonsense psychological explanation? Yes of course. There have been strenuous efforts to explain how causal and social relations to distal objects can be essential to psychological explanations of behavior; and the resulting theories are, in my view, more than a little strained. A consequence of making sense of internal intentionality is to vindicate a classical internalist view of commonsense psychological explanation, or at least to make it coherent. Still the main question is more basic than the explanation of behavior. It concerns whether mental properties as they are in themselves have merely contingent connections with behavior and environment. That is hardly a small matter if we are interested in what we are: We have inner mental lives. The intuitive idea that the intentionality of outward-directed thoughts can be internally determined has run into serious trouble, for a certain externalist conception of intentionality has considerable intuitive force. Tyler Burge, as much as any other philosopher, has made a powerful case for externalism about mental content.5 This he has done by arguing that—to put it in a way more abstract than his—the semantic resources of the analytic tradition, whereby intentionality consists in the truth-conditions and satisfaction-conditions of thoughts, cannot support internalism about intentionality. I agree with this. But I draw a different conclusion: What matters to intentional internalism does not depend on those classical truth-conditional factors. Something theoretically novel (though familiar in experience) needs acknowledging. My homage to Burge in this volume will be expressed by my being driven to extremes.6 While the internalist intuition appears to me correct, the core of current externalist theory also appears correct. So the core of externalist theory must be compatible with intentionality’s being an internally constituted feature of mental states. Externalists are right about the reference and truth-conditions of thoughts. But despite vivid appearances to the contrary, intentionality does not presuppose reference and it is not externally determined. That is the idea I will try to make sense of.
232
1
Brian Loar
Externalism about Intentionality
Externalists about intentional mental content regard the opposing position as conceptually incoherent.7 Here is the externalist reasoning, as I try to put it straightforwardly to myself. (It is worth noticing that this externalist line of reasoning does not presuppose that mental content is as that-clauses capture it. The earlier point in denying this was not to confute externalism directly but to open up mental content as it were to phenomenological access.) The externalist’s first premise is that thoughts can be intentional or “directed,” can “purport to refer,” only by presupposing actual references. The externalist may grant that a thought can purport to refer to something external even if it does not succeed in referring, even if there is so to speak nothing there. Even so it must represent what it purports to refer to as such and such, as having some property F. And so it must succeed in referring to that property. The second premise is that such reference is constituted by externally determined—causal, social, and so on—relations. If the premises were both correct, no sense could be made of the isolated brain’s having the same intentional states as me; intentional mental content could not be internally determined. Almost everyone agrees that any singular concept may fail to refer and still have the intentional content it would have if it had succeeded. This holds in the most obvious way of ordinary definite descriptions. ‘The oldest dolphin in Andorra’ purports to refer, and for all I know it fails. It also holds of perceptual-demonstrative concepts: I may exercise a visual-demonstrative concept—‘that horrifying animal’—and yet be hallucinating. Despite their failing to refer, both concepts intuitively have full intentional content: Each purports to point, quite specifically, to an object, even though the world does not put an object in its way. But the presupposition thesis is satisfied because, on the face of it, the intentionality of singular concepts depends on what those general concepts refer to: ‘dolphin’ and ‘animal’ refer to kinds or properties, and according to the externalist this constitutes, at least in part, the singular concept’s intentionality. The externalist may of course allow that a property-concept can itself fail to refer, even while fully purporting to pick out a property. But this must be grounded in further concepts that actually refer and that hence stand in externally determined relations to externally constituted properties. Another way to put the externalist’s (often implicit) point about intentionality: Thoughts cannot purport to refer unless they impose success-conditions, or satisfaction-conditions; and these depend, however indirectly, on reference to objective properties. Externalism about intentionality assumes, on this account, that intentionality presupposes reference. The externalist’s second premise is that referring to and connoting external properties consist in externally determined relations between concepts and properties, at least for concepts that purport to be outward directed. Externalist positions about reference may diverge: Some regard all basic reference relations as non–socially mediated causal rela-
Phenomenal Intentionality as the Basis of Mental Content
233
tions to things, and others, Tyler Burge famously, as including social relations to the usage of others. I accept Burge’s view to this extent: We cannot realistically deny the role of social relations in the mediation of much ordinary reference. How much farther we should go is not clear. Suppose one’s concept ‘animal’ derives its reference socially, from biologists’ conceptions. It is not clear whether their more basic concepts might ultimately be determined purely personally by nonsocial perceptual and conceptual-role relations. But this issue is beside the present point. In agreeing with externalism about reference I accept this: Basic property-references and property-connotations are constituted by relations that, at least in part, are externally determined, whether socially or not. In the standard debate, this concedes substantial ground to externalism about intentional content. But it does not imply it; for it does not imply that intentionality presupposes reference. On the externalist view of mental content, another brain’s perceptual states and thoughts can be intentionally equivalent to my externally directed perceptual states and thoughts only if we share at least some of my actual external references. The equivalence would require an overlap in property-reference. If the isolated brain’s perceptual states and other concepts do not pick out some of the same properties as my externally directed perceptual states and thoughts, then none of its mental states can have the same intentional content as my outward-directed mental states. This is the externalist premise to which we must reply, for we agree that property-reference consists in externally determined relations. To sum up the externalist line of reasoning: (i) Mental content is intentional; (ii) intentionality presupposes reference; (iii) reference, for outward-directed thoughts, consists in externally determined relations (especially to kinds and properties); therefore (iv) my outwardly directed thoughts do not have internally determined mental content or intentionality—that is, the mental content of my outward-directed concepts cannot be shared with an isolated brain. I must emphasize the distinction between externalism about reference and externalism about intentional mental content: I accept the former and deny the latter. 2
Conceptual Roles and Mental Content
What we seek is a conception of mental content that is available commonsensically and that is internally determined. This could seem to be easily delivered by an established idea, namely, that we implicitly individuate thoughts in terms of their conceptual roles. A thought’s conceptual role consists in its inferential and probabilistic connections with other thoughts, desires, and so on, and with perceptions. Conceptual role theory standardly avoids appeal to intentionality. That can seem to be a considerable advantage, for it permits an attractive clarity: The horizontal and vertical aspects of mental content are factored out,
234
Brian Loar
that is, internal-explanatory conceptual role and external-referential intentionality. Externalism about intentionality is acknowledged, but internalism about mental explanation is nevertheless defended. Now I am quite sure that conceptual role is central in individuating thought-contents. How a thought conceives things must consist in part in its conceptual commitments. That is essential to what I will propose. But conceptual role on its own seems to me inadequate to explain our ordinary understanding. Conceptual roles are too blank to constitute internal mental content as we conceive it. Thinking is something lively—there is something that it is like to engage in it. So phenomenological reflection on thinking hardly conceives its properties in purely dispositional terms. But perhaps we might add phenomenal states to conceptual roles—and would this not give us the internal liveliness? We might then think of perceptual states and other phenomenal states as among the realizers of conceptual roles, or somehow intimately connected with such realizers. The liveliness of thinking in general would stem from perceptual states, linguistic states, various forms of imagery, with conceptualizing supplied by their connections within an interlocking network of conceptual roles. And who knows what innate conceptual structures there might also be into which perceptual states could nicely fit.8 Although the picture thus vaguely put is doubtless on the right track, it does not seem to me to promise an internalist conception of mental content. For we apparently lack appropriate nonintentional conceptions of perceptual states. We can hardly peel the phenomenal aspects of vision away from its intentionality; we just do not have nonintentional conceptions of “visual fields” or the like. Or try as I might I cannot muster such conceptions. Visual perception is phenomenologically focused on objects, spaces, and their properties; there are no pure visual sensations that might add nonintentional life to conceptual roles. If the externalist is right about intentionality, a phenomenal elaboration of a conceptual role theory will not yield ordinary, intuitive conceptions of internal mental content. The externalist might in any event complain that the project would be futile even if we had purely phenomenal and nonintentional conceptions of perceptual states. Internal goings on would not on their own constitute a mental life: For they would, phenomenologically, not look out to external space. It would be in McDowell’s dramatic phrase “all darkness within.” I can envisage a spirited defense by the conceptual role theorist. Both of the foregoing objections ignore the availability of a deflationary notion of intentionality, that is, of reference and truth-conditions. It may well be that we cannot conceive visual qualia nonintentionally. But this could have the following explanation. “We cannot conceive ordinary visual experience unconceptualized—that is, unless it is minimally conceptualized by object-concepts, spatial-concepts, and so on. This conceptualization may be understood in terms of something like conceptual role or conceptual struc-
Phenomenal Intentionality as the Basis of Mental Content
235
tures as long as we also grant the conceptual role theorist what he is classically entitled to, namely, a ‘disquotational’ or deflationary notion of reference.9 To conceive a way of conceptualizing visual experience simulationally will then employ object-conceptions, and so on. And when we reflect on them we can hardly avoid, as it were, disquoting. They are our concepts and we can hardly think of them in nonsimulational objective terms.” This explanation is tantamount to proposing a deflationary internalist theory of intentionality, which it may be argued quite adequately accounts for the phenomenology. Let us have a look at such a theory. 3
An Internalist Theory of Intentionality Rejected
What we might call the standard internalist conception of intentional mental content denies that the reference of basic concepts is constituted by externally determined relations (the second externalist premise). At the same time, it does not question that intentionality presupposes reference (the first externalist premise). On this standard internalism, certain basic concepts express or pick out properties and relations independently of causal relations between the mind and those properties. This might for example be held of our concepts of shape and spatial relations, that is, that they express spatial properties and relations without externally determined referential mediation.10 And a similar view may be taken of concepts of categorial properties such as causation and physical-objecthood. Externalists are dismissive, and there are two rather different objections. The first is the thought that the externalist thought experiments of Kripke, Putnam, and Burge11 show that reference is environmentally and hence externally determined. The second objection is more basic, and harder to get a grip on: Such internalists are accused of holding implicitly a magical theory of reference (as Putnam called it), that the mind somehow grasps externally constituted properties independently of natural relations to those properties. Let us see how a reasonable internalist about reference might respond to each objection. As a preliminary we classify concepts according to the apparent role that wide contexts play in determining their references. By ‘reference’ I mean not only the reference of singular terms and predicates, but also the truth-conditional contributions of logical connectives, and more generally of all concepts. So we have: A. Wide concepts, whose reference is determined by externally determined relations to external contexts. These include singular indexicals and demonstratives, and kind terms such as ‘water’, ‘tiger’, ‘arthritis’. Whether a concept belongs to this type is decided by thought experiments that shift contexts between earth and twin earth, between arthritic twins, and the like.
236
Brian Loar
B. Narrow concepts, for which reference is context independent, that is, independent of contexts that transcend ‘internal conceptual role’ and the like. Paradigms are the logical connectives. If a connective has the conceptual role of ‘and’ or ‘all’ it eo ipso expresses conjunction or the universal quantifier. There are no Twin Earth reference-shifts for logical connectives. Presumably mathematical and modal concepts belong here as well. (It will be convenient, if somewhat inelegant, to include in group B indexicals that pick out internal states: ‘this sensation’, ‘this thought’; for their referring arguably does not consist in externally determined relations.) C. Debatable cases, predicates that do not uncontroversially or determinately belong to class A or to class B. These are the concepts already mentioned: ‘cause’, ‘physical object’, spatial concepts, and perhaps others. As I have found among philosophers in conversation, intuitions about such concepts are far more open than they are about ‘water’ and the like. A conservative internalist strategy—which I will describe but not endorse—proceeds as follows. Assign concepts in class C to class B. If for example a concept has the internal mental role (conceptual, imagistic) of our spatial concept ‘between’, then count it as by that very fact referring to spatial betweenness, regardless of context. On this proposal even isolated brains can think about space. Next, count such concepts as descriptive primitives. Then count kind- and property-concepts in class A as abbreviations for descriptions whose logical and nonlogical primitives are the B and C concepts. Perhaps with a few additions, logical, categorial, causal, spatial, and sensory concepts would then give us a basic working stock of primitives. This is a familiar description-theoretic internalist idea. Next, take ‘water’ to mean something like ‘the so and so stuff that causally grounds our use of “water” ’. The reference of the concept ‘water’ then varies with the reference of ‘our’. Reference-shifts and reference-constancies of familiar sorts are accommodated.12 If the description-theorist runs into circularity problems with terms of class A—explaining ‘parent’ via ‘child’, ‘Cicero’ via ‘Cataline’ and so on, there are generalized descriptive techniques for dealing with that. This assumes that we legitimately assign concepts of class C to class B and not to class A. But might there not be analogues of Twin Earth thought-experiments for those basic concepts? Can isolated brains conceive of spatial relations? The internalist I have in mind says that we are conflating cases that can easily be kept apart. The Kripke, Putnam, Burge thought experiments tell us something substantial concerning our intuitions about concepts in group A, but those intuitions need not extend to concepts of group C: the internal properties of our spatial experience determine which spatial properties and relations our thoughts are about.13 It would be wrong to charge our conservative internalist with magic. Perhaps some unreflective philosophers, and some dualists, do have a magical conception of reference,
Phenomenal Intentionality as the Basis of Mental Content
237
implicitly taking the mind somehow to reach out and grasp, nonnaturally, externally constituted properties. But there is a less dramatic way to deny that reference is essentially causally–socially determined. No magic underlies the intuition that ‘and’ stands for conjunction without mediation by a causal or other contingent natural relation. That standingfor relation can be explained prosaically. Adopt a deflationary conception of the reference of ‘and’ and of all expressions in class B. All there is to the reference of ‘and’ is captured in this schema: ‘P and Q’ is true iff P and Q. Suppose a connective of another’s language has the same conceptual role as a connective of our language. Then assign to their connective the deflationary truth-conditional interpretation of ours.14 This (projectivist) way of putting things is equivalent to the (nonprojectivist) idea that, for group B concepts, conceptual role determines reference, that is, contribution to truth-conditions, without the mediation of further contingent relations. Our notion of reference, we might suppose, simply takes the reference of such logical concepts to be thus minimally determined. The internalist theory of reference, then, as I am characterizing it, takes concepts of class B (including C) to refer in a minimal or deflationary sense. So denying externalism about basic predicate reference does not thereby commit internalists to magic. For in a plain sense internal conceptual role determines reference for concepts of class B cum C without mediation. What does ‘determines’ mean here? There is a conventionalist element in deflationary theories; on them, reference is not substantively determined. It is as if we conventionally assign certain references to certain basic conceptual roles.15 A projective-deflationist theory of reference captures that conventional assignment without its appearing arbitrary. This account seems to work well enough for ‘or’ and ‘all’. Does it work for spatial concepts? Keep in mind that magic is beside the point. The question yet remains whether reference to spatial properties is like reference to truth-functions or like reference to water. I do not find it so plausible to count isolated brains as capable of concepts that pick out spatial relations, and here is why. My own spatial concepts appear to have a crucially demonstrative element, pointing visually and tactually to certain relations and properties, at least vaguely. I of course cannot define ‘straight line’ by pointing. But this does not mean that what determines spatial reference is not in part demonstrative. By pointing to the sorts of relation and properties that are to count as curviness, betweenness, and so on, spatial perception apparently gives worldly content to otherwise purely abstract concepts. Without such diffuse pointing in visual and tactile experience, spatial concepts would, it seems to me, be empty. The internalist about reference may say: But brains in vats have visual and tactual experiences, and purport thereby to refer demonstratively. Quite so. But the perceptual factors of spatial concepts imply something about how their reference is determined: If those concepts are in part both visual and demonstrative, then their references will have to be determined in part in the manner of visual demonstrative concepts. If spatial concepts depend on concepts of the form ‘a relation of that general sort’, where
238
Brian Loar
‘that’ points visually, then the reference of ‘that’ depends on seeing, and seeing is an externally determined relation.16 It follows that spatial concepts are wide concepts: They belong to class A, where all outwardly directed perceptual-demonstrative concepts belong. But if spatial concepts are not in class B then the internalist about reference really has no hope. No description-theoretic explaining away of the apparent external determination of reference for concepts of class A will be viable. 4
Phenomenal Intentionality
How are phenomenal aspects of perceptual states related to their intentional properties? Several views are current. At one extreme there are pure qualia views. The qualitative aspects of (say) visual experience are in themselves nonintentional; those sensational aspects of visual experience are intrinsically as nonrepresentational as the blotched paint on a stucco wall. This is familiar from certain ways of construing color perception: The surface property of redness causes experiential red*-ness. Externalists comfortable with qualia might then regard perception as structured thus: A given visual experience has the property of red*-ness, and that property, although not itself intrinsically intentional, is a component of a perceptual representation whose intentionality lies in a causal relation to redness.17 At another extreme, phenomenal aspects of experience are held to be an illusion. Representationism18 holds that the only phenomenal qualities we can discern are the properties perceptions represent their (purported) objects as having: There is only redness and no red*-ness. Externalist representationism holds that visual representation is a matter of externally determined (e.g., causal) reference relations. The view is apparently widely held, and, interestingly enough, often on phenomenological grounds.20 These two very different externalist views of the relation between phenomenal qualities and intentionality provide useful contrasts with the internalist view of phenomenal intentionality that I find intuitive. On the one hand I rather think that we have a coherent conception of the felt aspects of perceptual experience; on the other hand I do not think these aspects are “purely” qualitative, that is, in themselves nonintentional. Let us begin with the latter point, for it is an important source of the representationist’s intuition. The idea of nonintentional visual qualia appears (to me) unmotivated. We cannot phenomenologically separate the pure visual experience from its purporting to pick out objects and their properties. It may seem that this makes sense for certain afterimages, phosphenes, and the like; even that strikes me as dubious, but I will not discuss it here. What seems to me obvious is that ordinary visual experience admits of no phenomenological bracketing of intentional properties: We simply cannot attend to the pure “visual
Phenomenal Intentionality as the Basis of Mental Content
239
field” and its nonintentional components. In some sense ordinary visual experience comes phenomenologically interpreted. But this does not imply representationism—although it seems often supposed to do so. It is compatible with, and in my view best explained by, a certain internalist view of intentionality that relies on the idea of phenomenal aspects of experience, in a broad sense. Let me first sketch the basic idea and then consider the representationist’s denial of qualitative aspects of experience. What I will call phenomenal intentionality is a phenomenologically accessible feature of virtually all perceptual experience and of perceptually based concepts, for example visual demonstrative concepts. The following will I hope convey the gist. Suppose some indistinguishable lemons are one after the other brought to my visual attention. The lighting, the position of my eyes and so on, are held constant. I am asked to think something about each lemon in turn, say ‘that’s yellow’. Afterwards I am told that some of the apparent lemons were hallucinations (that is what the wires were for). I am asked whether, despite this, my successive visual demonstrative thoughts all visually presented their objects in the same way. Surely a natural reply is yes, in a rather intuitive sense. This presents itself as sameness in an intentional feature. For those demonstrative concepts (both the ones that succeed in referring and the ones that do not) all purport to pick out some object visually. You cannot capture this common feature by generalizing over objects: ‘There is some object that the demonstrative concept visually presents’. And surely the content of those thoughts is not itself existentially quantified: “I am seeing some lemon or other and it is yellow.” The thoughts in question are demonstrative and they are not self-consciously reflexive. An apt way to put those concepts’ common feature seems to be this: Those visual demonstrative concepts, and the perceptions that underlie them, are all singularly visually directed. This is a nonrelational phenomenal feature, by which I mean something rather strong: We are aware of internally determined phenomenal features of visual experience, of their manifold felt aspects, and among those features— though not separable in imagination—is the directedness just mentioned. The feature presumably belongs primarily to a visual perception, and derivatively to a visual demonstrative concept that incorporates the perception. I will speak loosely of its being a feature sometimes of the perception and sometimes of the concept. Why call it intentional? I do this in the hope of engaging archaic intuitions. A natural way to capture the phenomenon is this: “the visual perception purports to refer,” “it is directed,” “it points.” When we considered whether conceptual-role properties, individuated “syntactically,” leave out something importantly representational about thoughts, we could surely have noted the relevance of phenomenal directedness. Does the idea of phenomenal directedness commit me to there being some mark—a little arrow—in the visual field? Well, the visual field would have to be packed with arrows, since virtually every one of its parts is directed on some bit of the passing scene.20 When I say that directedness is
240
Brian Loar
‘phenomenal’ I mean merely that I can identify it in experience. I apparently can tell that hallucinatory experiences have a ‘purporting to refer’ property that is also present when visual experiences pick out real objects in the normal way. Even if singular directedness is an internally constituted property of perceptual concepts, it does not on its own vindicate an internalist view of perceptual intentionality. For the externalist will surely object that a visual perception that fails of reference will nevertheless purport to refer to its (nonexistent) reference as having some property F. I earlier noted that this requires (in the simplest case) reference to the property F, which is in general externally determined. So even if a singular demonstrative has a phenomenal directedness that is independent of the demonstrative’s reference, internalism about intentionality is hardly thereby made coherent. The point is clearly correct. The present idea though is not that the singular phenomenal directedness of a visual demonstrative concept is sufficient for internalism about that concept’s intentionality. But it is a key step in constructing the notion of internal intentionality. The apparent intentional properties of a singular visual demonstrative concept are not exhausted by the references of its constituent kind terms (its ‘as F’ contents). The latter do not account for the intentionality of the visual demonstrative as a whole, for that apparently is an intentional property over and above the referential properties of the constituent qualifying concepts. So there is an intentional property, (singular) directedness, that does not consist in (singular) reference and is not explained merely by the reference of kind terms. Perhaps it will be said that this directedness is just a matter of conceptual role. That can be said, but it hardly neutralizes the phenomenology. Who knows whether what appears as directedness consists in some underlying factor that might aptly be called conceptual role? We apparently do not conceive it in terms of its place in some system of conceptual roles. We have a phenomenological take on it, and that is what I call attention to. Once phenomenal directedness is admitted, it is difficult not to admit also, as we will see below, something analogous for a crucial group of kind concepts. I will argue that this is all we need as a satisfactory basis for internally constituted intentionality for thought in general. 5
The How and the What of Intentionality: Mere Intentional Objects Considered
As I see it, phenomenal intentionality is a matter of how one’s perceptions and thoughts represent things if they succeed, rather than of what is thereby represented. The representationist says the latter. It may be replied that the ‘how’–‘what’ distinction is bogus. For to say how a visual perception purports to present something is to say what it presents it as. Even if no appropriate physical object is there, my perception presents something as a snake. And that presupposes reference to a property, at least the property of being a phys-
Phenomenal Intentionality as the Basis of Mental Content
241
ical object of a certain approximate size and shape. Right. But this misses my point. As we noted above, directedness is needed in addition to predication to explain the how of an empty visual demonstrative concept. The nonrelational intentionality of predicative concepts is yet to be discussed. But for now let us be content to say that the directedness of a visual perception is an aspect of how the perception (and the demonstrative concept that incorporates it) presents things. It is not a matter of the perception’s presenting something as F, but rather of its style or manner or mode of presentation. But the representationist has a reply. Representationists typically count a perception that fails to refer as yet having an intentional object—a mere intentional object so to speak. This suits the ‘what’ conception of intentionality: The intentional properties of even an empty thought or perception would consist in what it represents (as being such and such), namely, its intentional object.21 But this is a peculiar way for the representationist to put things. The representationist is a referentialist. In the lemon hallucination the only references to be noted are properties, that is, the references of the perceptual demonstrative’s as it were predicative factors. There is no further reference: The perception purports to refer and fails. And failing to refer is not a form of reference, however apt talk of “representation” may appear. Speaking of mere intentional objects is all right in its way; but it should not mislead one into claiming to have characterized phenomenal intentionality in purely referential terms. An intentional ‘how a perception presents things’ cannot easily be avoided. To pursue the point, here is a simple fact: In the veridical case there are not two intentional objects, the mere intentional object and the real object. What the hallucinatory perception has in common with the veridical perception is, then, not a ‘mere intentional object’. The two perceptions have something intentional in common, and it is that common feature which concerns me. The representationist may say that although they do not share having a mere intentional object, they do, however, share the property of having an intentional object. But it is hard to take this seriously. The veridical perception has an intentional object in a transparently relational sense: It refers. It could only be a fanciful Meinongianism that construes having a mere intentional object relationally. But suppose we went Meinongian. It is still far-fetched to suppose that we then end up with something that the veridical and hallucinatory perceptions have in common, for ‘having an intentional object’ would have to stand for two very different relations. The simple fact is this. What the two cases have in common is something phenomenological. We could call it ‘having-an-intentional-object’, with the hyphens marking a nonrelational reading. But it is less misleading to use an overtly nonrelational form, for example, ‘directedness’. And this clearly concerns the manner in which a perception or visual demonstrative concept presents things rather than what is represented. Why call directedness ‘intentional’ if it is nonrelational, if it is about the how rather than the what of perception? What else to
242
Brian Loar
call it? It seems to be the primitive basis of our intuitions of the phenomenal ‘aboutness’ of perception. The determined referentialist may pursue a different strategy. For there is still the language of ‘as if’, the language of appearances. We can describe the lemon hallucination by saying that it is as if I am seeing a lemon, or ‘it appears to me that I am seeing a lemon’. If one holds an externalist view of reference one will also hold an externalist view of such appearances: For the function of a that-clause (as in ‘appears that’) is to capture the references of the state or property thereby ascribed. If appearance-properties are then captured in that form, intentional qualia will turn out to involve relations to external objects and properties, and so cannot be regarded as entirely internal properties. Now this strategy rather overlooks the representationist’s commitment to phenomenology. Recall how Smart attempted to capture the experience of a yellowish-orange afterimage: Something is going on in me that is like what is going on in me when I am seeing an orange. The problem with this analysis of sensory experience is that it is phenomenologically blank: it does not imply that there is anything in particular that it is like to experience a yellowish-orange afterimage. The language of appearance is, unlike Smart’s locution, at least mental in its implications. But to say that it appears that . . . , or that it is as if . . . , is not to say how it phenomenally appears. And the point of the lemon case was that there is something phenomenologically in common among the various visual experiences, and that it included something that is phenomenologically intentional. Mere talk of appearance may point to that extrinsically, but it does not capture what it is like. This does not mean that ‘appears’ has no phenomenological role. For one can say to oneself ‘it appears thus’ and point in memory or imagination or present experience to a phenomenal type. But that of course does not help the externalist. 6
Is There Phenomenal Paint?
According to Gilbert Harman,22 when you turn your attention to your perceptual experiences, all you can discern phenomenologically are properties of the (apparent) object of the experience, that is, shape, color, and so on. This is in strong contrast with how pictures appear: We can attend to the paint in a picture, but, according to Harman, there is no phenomenal paint of which we are introspectively aware. Phenomenal paint fails to appear not only in unreflective experience but even on phenomenological reflection. Of course he does not deny that some perceptual experiences can lack real objects and yet have a fully phenomenal presence. But he is content to appeal to mere intentional objects, intentional objects that do not exist. An experience’s qualities in such a case consist in the properties it attributes to the object that isn’t there. The structure remains the same: We
Phenomenal Intentionality as the Basis of Mental Content
243
are aware not of the experience’s phenomenal qualities (for there are no such properties) but of the properties of the apparent object of experience. The argument of the last section, concerning the need for a nonrelational commonality between veridical and nonveridical perceptions, does not, apparently, touch Harman’s point. For it does not imply that the phenomenology delivers any highly specific qualitative aspects of experience, as opposed to highly specific properties of the apparent objects of experience. Perceptual experience is phenomenally transparent: We seem to be directly aware of properties of objects rather than properties of experience itself. And it may seem that carping about mere intentional objects will not neutralize that observation. So the question I wish now to raise is this: Is there phenomenal paint? The concept of directedness purports to be of a phenomenal property in the sense of a property of experience rather than a property of an apparent object of experience. At the same time, the directedness of perception is not separable in imagination from the more specific phenomenal aspects of perception. It seems then that if there is phenomenal directedness there must be phenomenal paint. Is there not a phenomenal difference between visual and tactual perceptions of shapes, a difference in the felt qualities or qualia of vision and touch, which is to say, a difference between visual and tactual paint? Consider the obvious phenomenological differences between seeing and touching a quarter. The representationist’s reply, as I understand it, is this. What we are inclined to think of as specifically visual and tactual differences in how we perceive a quarter are in fact differences in its perceived qualities over and above its shape and size, differences between its color and luminosity, and its texture and solidity. So we are not forced by the phenomenal differences between sight and touch to admit differences in qualitative aspects of experience, that is, differences in how experiences represent rather than what they apparently represent. Bill Lycan, whose position is strongly representationist, has developed an interesting strategy for defusing apparent cases in which the representational content of two perceptions is shared while their phenomenal manner of presentation differs. The idea is to take the alleged difference in qualitative manners of presentation to be in fact differences in apparent properties of intentional objects. Lycan’s account requires finding multiple levels of intentional objects in problematic experiences. I will describe an experience and consider how Lycan’s strategy avoids qualia in accounting for it. Keep one eye open and use your fingers to stretch it in different directions. You see some apples in a bowl, say, in blurry distortion. Surely the perception does not represent the apples as themselves blurry. Rather, the proponent of qualia will say, the perception blurrily represents the apples. The blur is a qualitative aspect of the visual experience. Lycan says not so: There could be a scene that is objectively “blurry” in that way, and that (nonexistent) scene is a sort of secondary intentional object of the blurry visual experience. There are two levels of
244
Brian Loar
intentional objects here: the ordinary apples in the bowl, and the (nonexistent) objectively blurry-apples-in-a-bowl. I am willing to grant that there could be an objective scene that looks to the normal eye just like that.23 But it seems to me that Lycan’s is a forced and ultimately wrong account of visual blur. The blur is an aspect of how the perception represents its objects, certain normal apples; it is not on its normal role a perceived property of some abnormal apples. The question is whether that can be argued more or less conclusively on its own terms, or whether a larger argument for visual qualia is needed in order to give the qualitative account of visual blur and the like its proper force. The latter is in fact what I am inclined to think. But let me first present an analogy from ordinary depiction to nudge intuitions in the right direction. For there is an intuitive distinction between what is depicted in a picture and how it is depicted, where by the ‘how’ I do not mean the surface of the picture but something intentional. 7
Aesthetic Interlude: The How and the What of Pictures
Consider representational paintings that are not photographically realist, for example, one of Picasso’s portraits of Marie-Therese Walter. It represents its subject distortedly, if quite gracefully. Marie-Therese is captured with rounded swooping lines and bright colors, and fragmentedly—her head, say, has one half in profile and the other half full-face. Doubtless there could be a real three-dimensional scene that looks just like this picture. But we do not see the Picasso portrait as representing a Martian: It seems unmotivated to say that it represents Marie-Therese by way of representing a Martian. The picture does not, at least as I am inclined to see it, represent any object as having that distorted shape. Rather it gets you to see its object in a distorted way (and part of the visual pleasure lies in deciphering the picture, following here and there and back again how it represents its object). The distortedness is not a matter of intentional content but of intentional style, not a matter of what is represented but of how it is represented. There are Picasso pictures that do seem to represent a distorted object—a figure on a beach, made out of bony pieces, like a surrealist sculpture of a seated bather. Two points about these pictures are it seems to me instructive. First, if the picture is of a bony sculpture, then that sculpture is itself represented in a realistic way, and that is itself a manner of representation, an intentional style—one that we usually do not attend to, for it is, until noticed, diaphanous. The diaphanousness of a realistic portrait should not blind us to pictorial realism’s being an intentional style. Second, the realistically depicted surrealist sculpture itself represents something distortedly; and the sculpture does not represent a further surreal or distorted object, and so on. It represents a woman, surreally. You cannot get rid of the manner, the how. I am not speaking of the physical paint, but of something
Phenomenal Intentionality as the Basis of Mental Content
245
perceptual, an intentional way in which—as we visually engage it—the picture presents its objects. (The Picasso picture is even more interesting than I have made it. We can move back and forth between the picture’s realistically presenting a bony surrealist sculpture, and its surrealistically depicting a woman on a beach.) The same holds for visual representations. It is difficult to see how you can get rid of the how, or the manner, of perceptual representation. That manner is as accessible as the how of pictures; and it is intentional. This seems to me to be a coherent view of blurriness and the like. The question is now whether we can turn that coherence into something stronger. 8
Inverted Spectra and Inverted Worlds
Harman’s and Lycan’s representationism is externalist. Those aspects of visual experiences of which we are phenomenologically aware are their ordinary referential properties; they involve externally determined relations to external objects and properties. There is a familiar, quite elementary reason for rejecting this view, which I find persuasive. We can coherently, and easily, conceive of subjectively different color-experiences that are of the same objective properties of objects. (The idea of inverted spectra is one way of conceiving this.) We can also conceive of a single color experience that is, in different circumstances, of different objective properties. This is persuasively shown by Block’s Inverted Earth, a color version of Twin Earth.24 The arguments of representationists against these possibilities do not, I think, depend on phenomenological intuition so much as on externalist theory. The phenomenology of imagination—and that is after all our current field of play— seems squarely on the side of the qualia-exponent. And all the subjectivist apparently needs here is that the phenomenology is coherent, which it appears to be. We must add this: These color qualia are not pure qualia, for they are phenomenally intentional—they phenomenally represent (what we conceive of as) object-surfaces as having certain properties. How to interpret this remains to be explained, as it will be below; but that it makes sense seems clear to me from the basic thought experiments. There are color qualia, and they are intrinsically intentional. 9
Isolated Brains
Externalists, as we have seen, often have no trouble regarding visual demonstrative concepts that fail of reference (in a hallucination, say: ‘that hand reaching out from the wall’) as having genuine intentional content.25 But according to these externalists, that intentional content is essentially anchored in the properties that the perception represents the (merely
246
Brian Loar
intentional) object as having. Putting colors aside, these properties will include physicalobject-types, spatial relations, and so on. (The visual hallucination represents the hand as an object with protruding appendages thrusting in my direction.) According to the externalist, my perceptual state represents a merely intentional object as spatially located only if that perceptual state stands to certain externally constituted properties in externally determined reference-relations. Evidently standard Inverted Spectra and Inverted Earth thought experiments do not count against this point. For they hold constant the basic physical properties that visual perceptions represent. So not surprisingly, we require a more radical conceptual possibility than those if we are to establish intentional internalism. It must show that we can hold constant phenomenologically accessible intentional visual qualia while varying all the properties that they represent things as having. Brains in vats to the rescue. One of the interesting facts about the current debate about representationism is how tame the thought experiments are. If the game is phenomenology, then we really ought to exploit all possibilities that are phenomenologically conceivable and prima facie coherent. So, once again: I could have a mental twin whose brain is a molecule for molecule duplicate of me; and I can conceive that twin as having the same visual experiences that I have, even though its brain is isolated from all the normal causal relations to the world that give my visual experiences their actual references. The point is that when I imagine how the brain’s visual experiences represent their (merely intentional) objects, I apparently imagine those experiences as in some sense intentional, despite the brain’s difference from me in all its references. Is this coherent? Discussions of representationism and qualia avoid this thought experiment, it seems to me, because defenders of qualia think they don’t need it. This is because they are concerned with qualia and not intentionality; they want merely to show that color qualia make sense. Even if visual qualia are phenomenally intentional this will not in itself support a purely internalist conception of visual experience. What is wrong with the idea that my twin-in-a-vat can have visual experiences intentionally equivalent to mine? There seems no phenomenological incoherence in the idea. There has been thought to be a conceptual incoherence however. For if the brain’s visual experiences are intentionally the same as mine, then according to the referentialist about intentionality they must share references with mine, which according to externalism about reference is impossible. But of course the argument is fragile, for it ignores the coherence of nonreferential or phenomenal conceptions of intentionality. When I imagine the brain’s visual states and at the same time conceive of them as having no references in common with mine, what am I conceiving? Here we return to the how versus the what. What I hold constant in imagining the brain’s visual experiences is how it conceives things. That is, I can coherently imagine a complete sharing of my experience’s phenomenological details
Phenomenal Intentionality as the Basis of Mental Content
247
conjoined with a complete unsharing of its references, at least with regard to my outwarddirected states. So we need an analogue of the phenomenal directedness of singular perceptual demonstratives for the other representational factors in perception, that is, the factors that represent external properties and relations. Suppose we can extend the idea of directedness to those aspects of visual experience that purport to represent spatial properties, and so on. Then my twin’s visual experiences and mine share that directedness; but mine refer to spatial properties (metaphysically rather than phenomenally spatial so to say) while his do not. If all this can be made out, we are aware of directed qualia, qualia that internally purport to refer not only to objects but to basic properties. My twin and I conceive things thoroughly in the same intentional manner. This is what we hold constant across twin brains, namely, highly specific forms of property-directedness. There are no shared properties of intentional objects. To return to the promised relevance of conceptual roles, we hold constant not only intentionalized phenomenal experience but also conceptual roles. Internal intentionality is to be located primitively in perceptually based concepts. It will be derivatively located in nonperceptual concepts via their conceptual connections with perceptual concepts. The subjective intentional properties of nonperceptual concepts are always of matter of, as it were, looking sideways via their connections with perceptual concepts. The earlier complaint about the intuitive deficiencies of our conceptions of narrow conceptual roles—as purely syntactic, as not capturing how thoughts subjectively represent the world—is I think answered on this picture. 10
Recognitional Concepts
The idea of directedness can be extended to demonstrative concepts that purport to pick out, perceptually, kinds and properties rather than individuals, what we can call recognitional concepts. They are an important if somewhat elusive variety of kind-concept or property-concept. They appear to me to have an intuitively evident sort of internal intentionality, which may be thought of, by analogy with ‘object-independent’, as ‘kindindependent’ or ‘property-independent’ intentionality. As we have seen, committed externalists about intentional content concede that individual demonstrative concepts have an object-independent conceptual or psychological integrity. But they will draw the line at property-concepts, kind-concepts, relation-concepts, insisting that their psychological or intentional individuation must incorporate the kinds, properties, or relations for which they stand.26 I am quite committed to externalism about the reference of these property-concepts: That a given recognitional concept refers to a given kind is of course a matter of some external, for example, causal, relation between the thinker and the kind.
248
Brian Loar
What I wish to deny is that either the external relation or the kind itself is part of the intentional individuation of the recognitional concept. Recognitional concepts are personal, and they are perspectival. Their reference is determined by non–socially mediated actual and dispositional relations between the thinker and the kind. And these concepts are individuated, in part, by the perspective from which they are conceived—for example, a perceptual perspective. So a visual demonstrative kindconcept concept may pick out a kind by virtue of past perceptions of its instances and a disposition to pick out further instances from its defining visual perspective. It is important to be clear about the following point. Visual recognitional concepts are not descriptions of the form ‘the kind to which this thing, that thing and so on belong’, that is, they are not descriptions that embed singular visual demonstrative concepts. Suppose I have a solid recognitional conception of a species of elm, without knowing its name. I need have in mind no particular elms, nor any group of them: My conception is of them in general, but from a certain perspective, from which I can take one of them in at a glance, say, while being able to see its bark and its branches. So we are not proposing a descriptive account of the kind-independent intentionality of recognitional concepts. How then shall we intuitively conceive the directedness of a recognitional concept—its purporting to pick out a certain kind of tree? Think of exercising the concept in imagination without applying it perceptually. For example, one wonders whether there are trees of that kind in Philadelphia; here one points in imaginative memory to a kind. What is useful about these cases, where the concept is not applied directly in perception, is that it makes it easy to isolate the purported reference to a kind (‘that kind of tree’) from corresponding purported references to individuals (‘that tree’.) The question is whether we have here an object-independent intentional property. It won’t surprise you that it seems clear to me that we have. Conceiving of a given visual kind-demonstrative’s failing of reference is not hard: One can be wrong in thinking that, from a certain perceptual perspective, one has picked out, or is able to pick out, a kind. But one’s recognitional concept may nevertheless have been as coherent as any, and perceptually focused as if on a kind. So the analogy with the lemon-hallucination seems fair. We also would like to establish a further analogy with the lemon case, that is, that recognitional type-demonstratives hold constant across reference change. And so I have to inflict on you a familiar waving of intuitions, but with a new emphasis. Imagine, then, some worlds like ours, as superficially similar as you like, but populated with different underlying kinds. Could that same recognitional concept—the visually embedded concept that kind—occur in all those worlds even though it picks out different kinds in each? This strikes me as straightforwardly evident. And, as in the lemon case, what intuitively we hold constant across these worlds—certain conceptual and experiential factors—are not easily equated with some combination of functional or syntactic and
Phenomenal Intentionality as the Basis of Mental Content
249
purely sensational properties. A kind of intentional directedness is again present, associated with the kind-demonstrative. It is analogous to the directedness of singular demonstratives—and doubtless derives from it—despite the difference between singular demonstrative pointing and pointing to a kind via a recognitional capacity. So it seems quite easy to conceive of a recognitional concept kind-independently and yet intentionally, as purporting to refer, pattern recognitionally, to a kind. Let us be clear about what the point has been so far. It is not that we finally have shown that recognitional concepts have purely internally determined intentional properties. That cannot be so, for we haven’t yet dealt with those further general concepts that are presupposed by, say, an ordinary visual recognitional concept (e.g., of elms), especially the spatial concepts that are entwined with visual experiences, as well as the general concept of three-dimensional objects persisting through motion and change. What rather it seems to me we have shown is that recognitional concepts have kind-independent intentionality in this sense: Even though a recognitional kind-demonstrative fails to pick out a kind, it nevertheless has (and now we speak phenomenologically) an overall intentional content that is organized around the concept’s visual kind-directedness, that is, its purporting as a whole to pick out, visually, a kind, a property, and so on. A recognitional concept purports to refer in two ways: (a) It purports to refer by way of an imaginative capacity. This can only be conceived intentionally, for one has little grip on purely sensory visual imagination. This imaginative capacity somehow involves as it were schematic singular visual demonstratives. The generality of the directedness of a recognitional concept, and the schematic form of imagined occurrences of individual visual demonstratives, are somehow closely connected. (b) It purports to refer by way of a disposition to respond to singular demonstrative visual experiences, where, as before, these are conceived intentionally and object-independently. 11
General Concepts of Physical Objects and Spatial Relations
Now we turn to spatial and other concepts, which I argued belong among wide (class A) concepts, whose reference is externally determined; these concepts include certain general concepts of approximate spatial relations, shapes, and the like, and a certain conception of a three-dimensional object as it persists over time. If we can extend to these concepts something analogous to the treatment of recognitional concepts, we will have rather a strong reason to think that concepts whose references are externally determined can in general be individuated by internal intentional properties. Class C concepts will moreover play a crucial structural role in explaining the internal intentionality of the huge remaining class of wide terms and predicates.
250
Brian Loar
Certain basic physical and spatial concepts do not have and do not need socially deferential roles. Of course ‘isosceles’ may well, for many of us, have as socially deferential a role as ‘arthritis’. But having the concept of an isosceles triangle would not be possible unless we independently had recognitional—visual and tactual—conceptions of more basic spatial relations. That at least is how it strikes me. Of course our recognitional concepts are not very precise; we hardly acquire the concept of a (perfect) right angle from perception alone. And yet concepts of more or less angular and curvy boundaries, of spatial betweenness, of relative distance—that is, the raw material of further precision in spatial concepts—are plausibly regarded as recognitional concepts. To say that they are recognitional concepts is not to deny them structural interrelations. It’s not a topic we can pursue here, but there seems to me no fundamental difficulty in the idea of structural interrelations among recognitional concepts. Quality spaces, after all, are structured. We also appear to have a recognitional concept of physical object in general. This does not mean that we have an image of a physical object in general; but there is nevertheless a perceptual ability to group together three-dimensional objects of all shapes and sizes. To say ‘perceptual’ leaves it open that some recognitional concepts are transmodal, that is, apply on the basis of both visual and tactual information. But I take transmodal concepts themselves to be perceptual concepts and not (as it were) pure categorial concepts, that is, not amodal. Do these general recognitional concepts have the kind-independent phenomenal directedness we claimed for less general recognitional concepts? Here I must again appeal to the fully intentional mental life of my twin-in-a-vat. All externalists abandon me at this point, however indulgently they have followed so far. I hope though that now we have not merely an intuition, but something approaching a principled account of it. Given externalism about the reference of these concepts, none of that twin brain’s concepts pick out physical-objecthood or spatial properties. The internalist claim is that my twin’s concepts are exactly similar intentionally to my recognitional concepts of the various spatial properties and relations. For they conceive the properties and relations to which they purport to refer in precisely the same way as my concepts do, via the same highly specific visual and tactual experiences and guided by the directedness of ‘that property’, ‘that relation’, and so on. The twin brain has a fully phenomenally intentional visual field. Given that the special directedness of recognitional concepts, including spatial and basic-object concepts, derives from the singular directedness of perceptual experience, it makes perfectly good sense to regard the intentionality of the brain’s general (i.e., nonsingular) concepts to be identical with mine. We need not decide whether the twin brain’s spatial concepts refer to some nonstandard properties and relations—for example, properties of the visual system itself—or fail of reference entirely. It is not clear that this is an interesting question. But
Phenomenal Intentionality as the Basis of Mental Content
251
if we can make sense of intentional properties that persist through shifting kindreferences and the failure of kind-reference, then I cannot see why that should not also apply to spatial recognitional concepts. You may object. “The sense we made of intentional directedness in connection with less general recognitional concepts depends on qualifying concepts that themselves are somehow intentional. That was intuitively crucial in supporting the intuition that the recognitional conception of elms had its own kind-independent intentionality.” Well, yes. But that does not mean that we then depended on the intentional properties of those basical qualifying concepts being externally determined. It appears to me that it is quite coherent to ascribe object-independent directedness to recognitional concepts all at once, including basic spatial concepts and so on. 12
The Paint That Points
Before turning to nonperceptual concepts and the question of their intentionality, let us look back to the representationist-qualiphile dispute. I have agreed with the representationist that visual experience is intrinsically intentional but denied that this requires an externalist treatment. I have also argued that the notion of ‘mere intentional object’, which the representationist requires if he is to be true to the phenomenology, is dubiously compatible with externalism about intentionality. Moreover, I argued that what appear to be coherent phenomenological intuitions support the qualiphile’s thesis that we have intuitive conceptions of the qualitative aspects of experience, although we have no way of separating the qualia from the intentionality. But this is all right, given that the phenomenological intentionality of perception is to be explained via ‘directedness’, which is itself a phenomenal notion. Now if by ‘paint’ one means something that we can conceive independently of its intentionality—like the paint on a canvas—then, at least in vision, there is no (pervasive) paint. But if ‘paint’ means qualitative aspects of experience that are separable from referential properties, then there is such a thing as phenomenal paint. And it points. 13
Personal Systematic Concepts
Presumably there are concepts that are neither recognitional concepts nor socially deferential concepts nor logical concepts. Calling them ‘theoretical’ makes familiar sense from philosophy of language, but it is perhaps somewhat overblown, and what I mean is not all that grand. So let me call them personal systematic concepts. To begin with, here’s what
252
Brian Loar
I mean by ‘personal’. Suppose that Fiona thinks that one way of becoming a mother is adopting a child and caring for it. When we tell her that ‘mother’ means a biological relation, she replies, determinedly, “When I say ‘mother’ it means what I mean and not what someone else means.” What construal shall we give of Fiona’s undeferential concept? We might try a description, or a cluster of descriptions. But that would, at best, be a matter of local convenience, and not a strategy for cashing out her personal theoretical concepts en masse. The reason is circularity. It is doubtful that we could explain those concepts using ordinary descriptions or description-clusters that appeal only to recognitional and logical concepts. We have to invoke other concepts that are in the same boat, concepts such as ‘female’, ‘child’, ‘raising’, and so on. Getting its content from having a role in a network of conceptual connections with similar concepts is what makes Fiona’s concept systematic, or if you prefer ‘theoretical’. Now consider her personal systematic concepts as a whole. They are bound to be multifariously linked with recognitional concepts, including the general concepts of physical object and spatial relations. Recognitional concepts that pick out children, that pick out the subjective psychological state of attention, that pick out attentive behavior, that pick out feelings, that pick out kinds of physical activity will also play essential roles in giving content to Fiona’s systematic concepts. We come to the question: How are we to conceive of the internally determined intentionality of personal systematic concepts? What I want to suggest is that their intentional properties are dispositional. We do not take in the intentional properties of a systematic concept all at once. We do so rather by finding our way about among a systematic concept’s lateral interconceptual connections. You may ask how the conceptual role of a concept can amount to an intentional property. We are used to thinking of conceptual role as “syntactic” role (as is often said). But what we uncover is hardly just the concept’s syntactic or functional or inferential connections. For one constantly engages, at every turn, perceptual recognitional conceptions that have their own independent directedness. The phenomenological world-directedness of a personal theoretical concept, I want to propose here, derives from its intimate conceptual connections with perceptual intentionality. So the idea that every concept can be revealed in an introspective glance, or even in an introspective stare, is not essential to the defense of internal, phenomenological, intentionality. This is not simply to assert that the conceptual roles of concepts are crucial to their individuation; that does seem to me beyond doubt. The point is not so much about individuation as about intentionality. The intuitive world-directedness of a concept—that phenomenological property—need not consist in its having its own perceptual focus, as do perceptual demonstratives. Its intentionality may come rather from the accessibility of conceptual repositionings and sidelong glances.
Phenomenal Intentionality as the Basis of Mental Content
14
253
Socially Deferential Concepts
Socially deferential concepts include most of the proper names in one’s repertory, and the extensive group of kind-terms to which Burge has called attention. Socially deferential concepts are of course not perceptually based in the manner of recognitional concepts, for recognitional abilities do not fix their reference. And so perceptual focus does not give us an intentional property of such concepts as a whole. Socially deferential concepts have about them something more discursive and linguistic: They involve conceptions of other speakers and of the shared language. What I propose is that socially deferential concepts belong among the personal systematic concepts. This is perhaps perplexing: How can socially deferential concepts be personal concepts? There are two ways in which a concept can be said to be personal. The first concerns how its reference is determined, that is, whether it is socially deferential or not. The second—as used in the phrase ‘personal systematic concept’—concerns how the concept is individuated, which is to say by the systematic role that the concept has in one’s own thoughts. And when I say that the socially deferential concepts belong among the personal concepts, I mean simply that those concepts—including their internal intentional properties—are determined in that way. At the same time, their reference is determined socially. The link between the two is this: That a concept of mine is socially deferential depends entirely on its systematic role in my thoughts. If it has the socially deferential role in my thoughts, then its reference is determined socially. If not, then its reference is determined otherwise, perhaps in the manner of a recognitional concept, or in the manner of personal systematic concepts that are not social. As a last note on social deference, we might observe that my twin-in-a-vat of course also has socially deferential concepts, but only in the sense that he has concepts that are equivalent in their internal intentionality to my socially deferential concepts. Might they have reference? If they have reference, it is not via the expected routes; my twin-in-a-vat has no concepts that refer via the usage of other people. Perhaps its concepts refer to some states of its own? I doubt that our concept of reference applies here; better to say that, like most of the rest of my twin-in-a-vat’s concepts, his “socially deferential” concepts fail of reference. 15
Directedness and Reference
How are internal intentionality and reference connected? Intentional directedness is an object-independent property, and it does not involve relations to objects. Reference comprises various causal and other relations to objects, and it is absurd to think of those reference relations as somehow instantiated without objects.
254
Brian Loar
My answer may not be fully digestible without more explaining and consequent ruminating, but here goes. Directedness is an object-independent property. But it is intimately involved with what is often called the diaphanousness of perception. Directedness is diaphanousness without an actual object. Earlier I pointed to the inadequacies of “mere intentional objects” in furthering the representationist project. But that leaves intact the usefulness of intentional objects in phenomenological description; and, in a phenomenological vein, we might say that directedness is diaphanousness toward intentional objects, whether “mere” or real. Now imagine having one of the lemon-experiences without knowing whether it is veridical. You are strongly tempted to say ‘that object’. Your perceptual processing presses mightily on your belief-inclinations, so strongly that you seem both to commit yourself, by using a demonstrative, and to take it back at the same time: ‘that object may or may not exist’. The phenomenology gives you the feel of a sort of ontologically neutral object that could have the property of existing or not-existing; and directedness is phenomenologically very like a relation to that neutral object, which could turn out to be real. Suppose you then discover that it is real. At this point the question arises of the object’s actual nonphenomenological relation to your perception. It turns out that a certain optical-causal relation holds in all such cases. The ghostly internal relation gets embodied in something nonmental and out there. The point of this fanciful description is to explain the relationship between directedness and reference. But the explanation of course is phenomenological-psychological; it is from a combined first-person/third-person perspective that directedness is intimately connected with reference.27 16
Concluding Remarks
The lemon demonstratives had this property in common: They purport-to-pick-out-anobject. This was said in a phenomenological vein. We are, it seems to me, as entitled to speak of phenomenological intentionality as we are of the felt qualities of a sensation. And the Cartesian intuition that is rejected by externalists about content is after all primarily a phenomenological intuition. We might reject that intuition by rejecting phenomenological or subjective conceptions in the philosophy of mind. But the only way to reject phenomenological intentionality selectively is to show that there is after all no such apparent phenomenon, or that the idea is incoherent. It is hard to see that externalist arguments are of the right sort to show that it is incoherent. If there is no reason to deny phenomenal directedness and no reason to regard this phenomenal feature as object-dependent, then there is no warrant for the externalist idea that internalism about mental content somehow leaves mental content blind, or that then “it is all darkness within.” In fact it is odd of the externalist to see his theory as providing inte-
Phenomenal Intentionality as the Basis of Mental Content
255
rior illumination. The metaphor seems to flow in the opposite direction: If the only intentional content is externally determined then it is all darkness within. Still the thought naturally arises, how could something in the brain account for intentional directedness? But just this question arises about phenomenal features in general, and here I am content to put it aside. Notes 1. Classic texts are Kripke (1972), Putnam (1975), Burge (1979), (1982). 2. As Kent Bach puts it: “every case is a Paderewski case.” Loar (1986a,b; 1987). 3. See Evans (1982); McDowell (1986). 4. Your ascription of Guido’s belief does not even implicitly convey an exact conception of his highly specific visual conception; we rarely have an exact conception of another’s mode of presentation. What is conveyed rather is a type of mode of presentation to which Guido’s precise visual concept belongs. It is the specific modes of presentation that individuate beliefs; but we typically merely gesture in their direction. The theory that types of modes of presentation are implied contextually in the gaps between the worlds in that-clauses (so to speak) was first introduced by Stephen Schiffer (1977). He has subsequently argued against his own theory (Schiffer 1992), but I am not persuaded. 5. See, for example, Burge (1979), (1982), (1986). 6. I engaged some of Tyler Burge’s arguments in an earlier paper (Loar 1986). There I made four proposals: (i) Burge’s arguments depend on the supposition that the psychological content of the predicative aspects of thoughts is identical with what obliquely interpreted that-clauses capture. (ii) That-clauses do not capture psychological content—any that-clause can apply by virtue of different psychological contents. (iii) Thoughts with the same psychological contents will in different contexts require different that-clauses to express them. (iv) Psychological content can be understood in terms of “realization conditions.” It has for a long time seemed to me that the fourth proposal is incorrect, because it relies on semantic resources (basically possible worlds) equivalent to those mentioned in the text as referential. As will become clear below, any such proposal is vulnerable to further externalist argument about reference—putting aside issues about thatclauses. Another paper of mine (Loar 1987) proposed a notion of subjective intentionality that is nonreferential and hence different from that of realization conditions. My current account of “subjective” intentionality is different again from the 1987 account, which was not “phenomenal” in the same way. 7. That is, the externalists who pose the most serious threats to the internalist conception of intentionality. One can notionally conceive of an externalist position that holds that internalism is coherently conceivable but that, as a matter of fact, intentionality consists in externally determined properties. 8. See Spelke (1995). 9. Field (1994). 10. It is not uncommon for philosophers simply to conflate concepts and properties. 11. Kripke (1972), Putnam (1975), Burge (1979). 12. It is not to the point to note the infelicity of substituting a metalinguistic description for occurrences of ‘water’ in that-clauses. For as noted above, the current topic is not that-clause ascriptions, but the modes of presentation that can make true such ascriptions without being explicitly captured by them. (Doubtless the metalinguistic ascription is semantically inequivalent to the unmetalinguistic original.) 13. Cf. McGinn’s “weak externalism” in McGinn (1989). 14. Field (1994). 15. This is quite close to the theory I proposed in Loar (1981). There it was expressed without appeal to deflationism in the home case. Conceptual roles for primitive concepts determined truth-conditional contributions. The spirit of this part of that theory was certainly conventionalist.
256
Brian Loar
16. This is not to deny that spatial concepts may depend on transmodal perceptual concepts. ‘Transmodal’ does not imply ‘nonperceptual’. The transmodal spatial conceptual capacities of blind people will get their content, presumably, from tactile, auditory, and proprioceptive perceptual states. To the extent that we sighted people can understand the mental contents of blind people, to that extend presumably we conceive of tactile, proprioceptive, etc. modes of presentation. 17. Some proponents of this view regard the purely qualitative aspects of sensation as experienced only obliquely or even as inferred. See Hill (1991); Shoemaker (1994). 18. To use Ned Block’s name for the position. 19. Harman (1990), Lycan (1996). 20. The visual experience supports a complex intentional structure. If you shift attention to another aspect of the visual experience, that reveals a distinct intentional directedness. Evidently there is a compact and quite complex set of such intentional directednesses within most visual experiences. 21. Harman (1990). 22. Harman (1990). 23. I also grant that there could be an external arrangement of lights that look, undistorted, just like phosphenes, pace Block (1996). 24. Block (1990). 25. Burge (1986) and Récanati (1993), as against Evans (1982), and McDowell (1986). 26. See Burge (1986) and Récanati (1993). 27. See Loar (1995).
References Block, Ned. 1990. Inverted Earth. Philosophical Perspectives 4: 51–79. ———. 1996. Mental Paint and Mental Latex. Philosophical Issues 7: 1–17. Burge, Tyler. 1979. Individualism and the Mental. Midwest Studies in Philosophy 4: 73–121. ———. 1982. Other Bodies. In Thought and Object, A. Woodfield (ed.). Oxford: Oxford University Press. ———. 1986. Cartesian Error and the Objectivity of Perception. In Subject, Thought, and Context, McDowell and Pettit (eds.). New York: Oxford University Press. Devitt, Michael. 1990. Meanings Just Ain’t in the Head. In Meaning and Method, George Boolos (ed.). Cambridge: Cambridge University Press. Evans, Gareth. 1982. The Varieties of Reference. New York: Oxford University Press. Field, Hartry. 1994. Deflationist Views of Meaning and Content. Mind 103(411): 249–285. Harman, Gilbert. 1990. The Intrinsic Quality of Experience. In Philosophical Perspectives 4, Action Theory and Philosophy of Mind, James Tomberlin (ed.), pp. 31–52. Atascadero: Ridgeview. ———. 1996. Explaining Objective Color in Terms of Subjective Reactions. In Philosophical Issues 7: Perception, E. Villanueva (ed.), pp. 1–17. Atascadero: Ridgeview. Hill, Christopher. 1991. Sensations: A Defense Type of Materialism. Cambridge: Cambridge University Press. Kripke, Saul. 1972. Naming and Necessity. In Semantics of Natural Language. Davidson and Harman (eds.), pp. 253–255. Dordrecht: Reidel. Loar, Brian. 1981. Mind ad Meaning. Cambridge: Cambridge University Press. ———. 1986a. Social Content and Psychological Content. In Contents of Thought, Grimm and Merril (eds.). Tucson: University of Arizona Press. ———. 1986b. A New Kind of Content. In Contents of Thought, Grimm and Merril (eds.). Tucson: University of Arizona Press.
Phenomenal Intentionality as the Basis of Mental Content
257
———. 1987. Subjective Intentionality. Philosophical Topics 15: 89–124. ———. 1995. Reference from the First-Person Perspective. In Contents, Villanueva, E. (ed.) Atascadero: Ridgeview. Lycan, William. 1996. Layered Perceptual Representation. In Philosophical Issues 7: Perception, E. Villanueva (ed.), pp. 81–100. Atascadero: Ridgeview. McDowell, John. 1986. Singular Thought and the Extent of Inner Space. In Subject, Thought, and Context, McDowell and Pettit (eds.). Oxford: Oxford University Press. McGinn, Colin. 1989. Mental Content. Oxford: Blackwell. Putnam, Hilary. 1975. The Meaning of ‘Meaning’. Mind, Language, and Reality: Philosophical Papers, volume II. Cambridge: Cambridge University Press. Récanati, François. 1993. Direct Reference. Oxford: Blackwell. Rey, Georges. 1997. Contemporary Philosophy of Mind. Oxford: Blackwell. Schiffer, Stephen. 1977. Naming and Knowing. Midwest Studies in Philosophy 2: 28–41. ———. 1992. Belief Ascription. Journal of Philosophy 89: 499–521. Shoemaker, Sydney. 1994. Phenomenal Character. Noûs. 28: 21–38. Spelke, Elizabeth. 1995. Object Perception. In Readings in Philosophy and Cognitive Science, Goldman (ed.). Cambridge, MA: MIT Press/A Bradford Book. Tye, Michael. 1995. Ten Problems of Consciousness. Cambridge, MA: MIT Press.
12
Internalist Explorations
Noam Chomsky As I write, the sky is darkening and the radio warns that a storm is heading toward Boston, expected to bring heavy rain and strong winds, flooding of rivers and coastal areas, damage to trees and homes, and loss of power. The preceding statement, call it S (and pretend it to be spoken), is manifested in an external medium and understood in various ways by speaker and hearers. Informally, we say it has sound and meaning. S is also related to inner states of speaker and hearers, which enter into the ways they interpret it. Communication depends on similarity among these states. In such ways, language engages the world. These topics have been studied for millennia from many points of view. They are also matters of interest in ordinary life, and there are varying cultural and linguistic practices concerning them, sometimes called “common sense” or “folk science.” Plainly, the study of the topics themselves is not the study of such practices. The earth sciences are not bound by ideas and attitudes expressed in S, and the same holds for Hume’s “science of human nature,” which seeks to discover “the secret springs and principles, by which the human mind is actuated in its operations” (1975, p. 14). While the issues are clear enough for the earth sciences, they are more convoluted when we turn to the science of human nature, which counts among its concerns the investigation of common sense (what we might call ethnoscience). Nevertheless, it proceeds on its own course. Inquiry may begin with ordinary notions of language, sound, meaning, wind, river, and so on, but without expecting them to be a reliable guide beyond a superficial level. I am interpreting Humean “science of human nature” as individualist and internalist. It comes nowhere near exhausting the study of how humans function in the social and physical worlds. The broader inquiries presuppose, if only tacitly, ideas about the inner states that enter into thought and action, and commonly use what they can from the internalist study of systems of the mind/brain. Interchange flows in other directions as well, as in the study of other organisms. In the case of human language, the least remote analogues are perhaps in insects.1 Investigation of such properties as “displaced reference” in bee communication will attend to the (internal) nature of bees, their social arrangements, and their physical environment, mutually supportive inquiries. Apparent conflicts should be resolved by clarity about the enterprise being pursued. Take, say, discussion of wide and narrow content, specification of mental representations, or individuation of thought and belief. If the inquiry falls within ethnoscience, we ask how people think and talk about such matters—recognizing, however, that the question cannot be raised directly for ‘content’ and ‘mental representation’, used here in a technical sense; that ‘thought’ and ‘belief’ are words of English without close counterparts even in similar
260
Noam Chomsky
languages, whatever significance that may have;2 and that commonsense accounts of what people do are not to be construed as a form of theoretical explanation. Here we find ourselves in a morass that is largely unexplored. In the science of human nature, different questions arise. We look into the theoretical framework in which such notions as content and thought are formulated and assess its descriptive adequacy and explanatory force. It comes as no surprise that commonsense notions are not of much use, and that pickings remain thin. Accordingly, one should be cautious about putting much weight on how “cognitive science appeals to the meaning of mental representations” to express generalizations about cognitive processes and action, and “to help explain these generalizations.” Similarly, the shift from “linguosemantics” to “psychosemantics” on grounds that “psychological natural kinds” are likely to better “fulfill the purposes of psychological explanation” is significant only as far as psychological explanation reaches. Quite far in some domains (e.g., visual perception), but rarely in dealing with behavior.3 The term ‘cognitive science’ is sometimes used for the empirical study of cognitive capacities (vision, language, reasoning, etc., components of the science of human nature that may not form a unitary discipline); and sometimes for reflection on the nature of mind. In the latter sense, it may be plausible to hold that “Kant’s central methodological innovation, the method of transcendental argument, has become a major, perhaps the major, method of cognitive science”; but not the former. In both cases, Fodor’s “First Law of the Non-Existence of Cognitive Science” is pertinent, though for different reasons (Brook 1994, p. 12; Fodor 1983, p. 107). Psychological generalizations also come in several varieties. Consider, for example, the discoveries about “what infants know”: enough to distinguish the mother’s language from a different one a few days after birth; to individuate physical objects in terms of common fate and other complex properties not many months later; and much else.4 The science of human nature tries to account for such accomplishments in terms of inner states, sorting out innate and environmental factors, constructing explanatory theory at any appropriate level. Here we have substantive research programs concerning a particular biological organism. Call this category of generalizations PG1. Consider the psychological generalization PG2: If Peter wants X, thinks that obtaining X requires doing Y, and is easily capable of Y, then he will typically do Y. PG2 differs from PG1 in many ways. It purports to explain behavior; the generalizations of PG1 do not. The empirical content of PG1 is easy to detect; not so PG2, which holds of any organism we choose to describe in such terms. Unlike PG1, PG2 is evaluated by reflection, not empirical inquiry, and opens no research programs—except, perhaps, into ordinary use of the terms and concepts of rationality. PG1 falls within the science of human nature, but that
Internalist Explorations
261
is less clear for PG2. The idea that “cognitive science” tries to express and explain PG2 is correspondingly obscure, as are efforts to ground such “intentional laws” and to explore their implementation in computational or other mechanisms. Study of PG1 falls together with other branches of science. “Let chemical affinity be received as a first principle, which we cannot explain any more than Newton could explain gravitation,” the eighteenth-century British chemist Joseph Black recommended, “and let us defer accounting for the laws of affinity, till we have established such a body of doctrine as he has established concerning the laws of gravitation.” Unification with fundamental physics was delayed until the twentieth century, while chemistry proceeded to establish a rich body of doctrine, its “triumphs . . . built on no reductionist foundation but rather achieved in isolation from the newly emerging science of physics” (Thackray 1970, p. 279; Black cited by Schofield 1970, p. 226). A similar course is reasonable with regard to PG1.5 But PG2 suggests few ways to proceed to a body of doctrine, hence to eventual unification. 1
Mental and Physical Reality
When chemistry had achieved a sufficient “body of doctrine,” one might have chosen to call its constructs physical (though some eminent scientists did not); even more so after physics had changed enough to permit unification, departing even more radically from commonsense notions of the physical so as to “free itself” from “intuitive pictures” and “give up visualizability totally,” in Heisenberg’s phrase (Holton 1996). The lessons carry over to mental aspects of the world, including mental representations and processes that might be postulated by the science of human nature. Cartesian dualism had raised substantive questions: A mechanical concept of physical was proposed and arguments were offered to show that it was incomplete. The questions— though not the problems that had given rise to them—dissolved with the collapse of mechanism, and we “accustomed ourselves to the abstract notion of forces, or rather to a notion hovering in a mystic obscurity between abstraction and concrete comprehension,” as Friedrich Lange, in his classic scholarly study, summarized this “turning-point” in the history of materialism, which deprives the doctrine of much significance (Lange 1925, p. 308). A century before, Hume had taken the dimmer view that by showing “the imperfections of the mechanical philosophy,” Newton had “restored [Nature’s] ultimate secrets to that obscurity in which they ever did and ever will remain.” Efforts to grapple with the component of the obscurity called mental led some to the conclusion that “it is the organization of the nervous system itself” that “freely exercises in a healthy state all the
262
Noam Chomsky
properties” of mind (La Mettrie). But the problems that had troubled the Cartesians were never addressed, and no substantial “body of doctrine” was developed.6 Apart from its theological framework, there has been, since Newton, no reasonable alternative to Locke’s suggestion that God might have chosen to “superadd to matter a faculty of thinking” just as he “annexed effects to motion, which we can in no way conceive motion able to produce.” As Joseph Priestley later elaborated, drawing “the obvious conclusion to the thinking-matter debate” (John Yolton), we take those properties “termed mental” to be the result of “such an organical structure as that of the brain,” superadded to others, none of which need be comprehensible in the sense sought by earlier science. While continental materialism took a different tack, at its heart “lay the assertion, based on one reading of Newtonian physics, that motion is inherent in matter, that all of nature is alive, that soul and body are one, all material, all entirely of this world” (Margaret Jacob).7 With the notion of the physical abandoned, never to be replaced, we can go no further than to ask whether mental aspects of the world, or others, “can be accommodated within the framework of physical explanation, as presently conceived,” being “fairly sure that there will be a physical explanation for the phenomena in question, if they can be explained at all, for an uninteresting terminological reason, namely that the concept of ‘physical explanation’ will no doubt be extended to incorporate whatever is discovered in this domain, exactly as it was to accommodate . . . numerous other entities and processes that would have offended the common sense of earlier generations” (Chomsky 1968, p. 98). The study of language tries to develop bodies of doctrine with an eye to eventual unification. Its constructs and principles can properly be “termed mental,” and assumed to be “the result of organical structure”—how, it remains to discover. On these aspects of the way language engages the world, there is little more to say.8 2
The Faculty of Language
There is reason to believe that humans have a specialized “organ” dedicated to the use and interpretation of language, call it the faculty of language (FL). We can take FL to be common to the species, assuming states that vary in limited ways with experience. Interacting with other systems (cognitive, sensorimotor), these states contribute to determining the sound and meaning of expressions. Study of these topics may not capture commonsense notions of sound and meaning, sameness of meaning, repetition, and so on; and there is no clear question as to whether they count as theories of sound and meaning, as in the case of motion, rivers, life, and so forth. For concreteness, consider the expressions of (1):
Internalist Explorations
(1)
263
(a) John was (too) clever to catch (b) John was (too) clever to be caught (c) John was (too) easy to catch (d) John was (too) easy to be caught
If Peter’s FL has attained the appropriate state, he knows that with ‘too’ included, (a) and (b) are true if John was so clever that one could not catch him (John), and that with ‘too’ deleted, (a) is “deviant,” requiring some nonstandard mode of interpretation (while (b) is differently interpreted). He knows further that (c) is true if it was (too) easy to catch John (who wasn’t “easy”); and that with or without ‘too,’ the obvious analogies fail for (d), also deviant. The study of FL seeks to encompass such observations under broader generalizations of the category PG1 and to discover the principles and structures that underlie them. Though not explaining Peter’s behavior, these elements of inner states should contribute to an account of how he thinks and acts, insofar as there is one. There is a reasonably successful theory that addresses such facts on the assumption that FL is a computational system with largely invariant principles. Tentatively adopting it, we attribute to Peter the corresponding mental states, representations, and processes (to which he has no conscious access).9 Suppose Peter’s FL is in state L. We may then say that Peter has (speaks, understands, . . .) the language L. Here the term “language” is used in a technical sense: call L an Ilanguage—“I” to suggest internal and individual, and also intensional, in that L is a specific procedure that generates infinitely many expressions of L. One such expression of Peter’s I-language, call it RAp, enters into determining how Peter might interpret the radio announcement reported in the statement S above. RAp resembles expressions generated by the minds of the announcer and other listeners, if they understand the announcement more or less as Peter does. The part of the science of human nature that concerns itself with FL, the states it assumes, and the expressions these I-languages generate, we could call I-linguistics. The notion of I-language seems to be about as close as I-linguistics comes to the various commonsense notions of language. Though unproblematic for ordinary life, these are intricate and obscure. One description of ordinary English usage, as good as any I know, is that it takes a language to be “an (intentional) object of (mutual) belief, appropriately studied hermeneutically within a sociology of language” (Pateman 1987, p. 73)—though the notion is no more likely to be useful for sociology of language beyond the surface than the phrases of S are for the earth sciences: say, the term “coastal area,” which has something like the status of “language,” except that it is much less amorphous, shifting, and multidimensionally interest-relative. The ordinary terms are often used as shorthand, as in discussing general properties of Chinese versus Italian (for neither of which is there much
264
Noam Chomsky
in the way of mutual belief). We also say that Peter does or does not speak the same language as I do, or live in the same place. But the world does not consist of such areas or languages in any sense that interests the earth sciences or I-linguistics. Even to speak of Peter as having the I-language L is a severe simplification; the state of any person’s faculty of language is some jumble of systems that is no more likely to yield theoretical understanding than most other complex phenomena of the natural world. Peter is said to be multilingual when the differences among his languages happen to interest us for one or another reason; from another point of view, everyone is multiply multilingual. In English usage, having a language is called “knowing a language,” a fact that has led to attempts to impose various conceptions of the nature of knowledge, and to determine to what entity Peter stands in a cognitive relation when he has L. For reasons discussed elsewhere, I think the questions are misconceived, though others are worth pursuing. Thus when Peter has L, he knows many things: for example, that ‘chase’ rhymes with ‘lace’ and entails ‘follow’. To spell all this out is a meaningful and important pursuit; and there are others about the nature of knowledge of X generally, the cognitive content of knowing how, the relations of knowledge to ability, and so on.10 The expressions of L are constructed from lexical items, each a collection of properties; the simpler words of S come close. We speak informally of the sound and meaning of a word, the way it is pronounced and what it means. The nearest I-linguistic paraphrase refers to the properties of a lexical item LI that are involved in sound and meaning: its phonological and semantic features (call them its I-sound and I-meaning, respectively). LI consists of these, along with formal features (not necessarily distinct) involved in the computational processes that form larger structures. And it may have more complex internal structure. There is no separate substratum, the word, in which the properties inhere, and any feature change yields a different LI. Putting aside many interesting issues, let us assume that the language includes a lexicon that is the set of LIs, and that the lexicon is accessed by the computational procedures that form expressions.11 The meaning of words has elicited a good deal of attention and controversy; that any such thing as I-meaning (“semantic representation,” “narrow content”) even exists is now commonly denied. Comparable questions about I-sound have rarely been raised. The empirical disciplines seem to me to study them in much the same way: in particular, to assume that both involve invariant universal features of which LIs are constituted (and hence are not radically holistic). I will tentatively assume that postulation of I-sound and I-meaning is legitimate, returning to reasons for denying it. FL attains state L with little if any effect of instruction, training, or decision, passing through characteristic stages and partially stabilizing at fixed periods. To borrow Hume’s phrase, the operations of the mind proceed “by a natural transition, which precedes reflec-
Internalist Explorations
265
tion, and which cannot be prevented by it.” In these respects too FL seems similar to other bodily organs. The lexicon continues to change in certain ways and is subject to a degree of conscious choice (as are other parts of language, marginally). Thus the lexicon of my language includes the word ‘dour’, which rhymes with the final word of S, ‘power’. Peter’s language may have a different word with the same meaning but rhyming with ‘poor’. I might abandon my usage and adopt Peter’s, or adopt a somewhat different meaning while keeping the I-sound fixed; by decision, or without and beyond awareness. Such events fall within what Tyler Burge (1986b) calls that “vast ragged network of interdependence, established by patterns of deference which lead back to people who would elicit the assent of others,” and which, along with various power relations, social arrangements, personality factors, and much else, “set a norm for conventional linguistic understanding,” as informally construed. Whether “they also provide linguistic meaning,” as Burge suggests, seems to me a matter of terminology, not fact. And it is unclear to me how one might learn about such a heterogeneous complex except by chipping away at parts that lend themselves to closer inquiry. In any event, I-linguistics goes no further than to say that in the case in question, I have added a new item to my lexicon, perhaps abandoning usage of an older one; and more generally, seeks only to isolate certain factors, crucial it seems, that enter into the awesome complexity of human affairs. It is often held that “people’s spontaneous judgments, or their intuitions, as philosophers call them,” constitute the subject matter for linguistics and for the theory of reference, which aim to systematize “grammatical intuitions” and “reference intuitions.”12 One can define projects at will, but it is hard to see the interest in systematization of some category of judgments, or other selected data. Take the study of reference, in its two aspects: the study of how people use language to talk about things and of their ideas about such matters. For these endeavors, judgments might provide evidence, perhaps reliable or useful, perhaps not. A serious investigation of either topic might explore cross-cultural similarities, poverty of stimulus considerations, psycholinguistic experiments, brain scans, or anything else that can be devised. Neither endeavor is the study of judgments, though we could think of them as studies of intuitions in a different sense: what they really are, a topic for which intuitive judgments serve at best as a source of information.13 Intuitive judgments are data, nothing more; they might become evidence within the framework of some explanatory theory. The judgments reported in connection with (1) have been used as evidence to support the conclusion that the complement of the adjective is clausal with three empty categories: the null subject, an empty operator O, and the trace of O, notions explained within the theory and justified independently if the account of (1) is to have any force. About these matters, speakers have no intuitive judgments, any more than they do about tensors and undecidability.
266
Noam Chomsky
Forced intuitive judgments with ordinary expectations withdrawn have to be considered with particular caution. Suppose we ask Peter whether a Martian speaks his language if it shares his judgments concerning (1) and other expressions but uses different principles or has a different biochemistry; or whether a Peter-duplicate created this instant can talk about rivers or water. Judgments become unclear, fading toward insignificance as thought experiments strip away background beliefs presupposed in the ordinary use of language, moving to the realms of Twin Earth, Swampmen, and other strange worlds.14 Suppose we adopt a “strange worlds” scenario to investigate what falls under Peter’s concepts: Does his concept water include Twin Earth XYZ, for example? Would he say— or be right to say—that on Twin Earth water is XYZ, unlike here? That Twin Earth doesn’t have water, only XYZ? Or either, as conditions of the thought experiment are changed? Or perhaps nothing coherent? Answers might provide evidence for some account of Peter’s linguistic states and practices, and ways of thought, and might bear on the initial question about concepts if that technical notion figures in the theoretical account. In isolation, the judgments would tell us little even if they were stable as conditions of the thought experiment vary, which does not seem to be the case. The study of folk semantics should not lightly assume that practices and conventions of some cultural tradition are a good guide to commonsense understanding, that of the investigator or anyone else.15 At the very least, it should try to discover the analogues to FL and I-language in this domain, seeking to identify the innate component. Suppose Peter says that Joe Sixpack voted for a living wage because he’s worried about his child’s health. Are we entitled to conclude that Peter believes the world to be constituted of such entities as Joe Sixpack, living wages, and health, and relations like votingfor and worrying-about that hold among them? Would the parallel inference be legitimate when Peter says that Tom visited Boston? If Peter says that the bank moved across the street after it was destroyed by fire, does he believe that among the things in the world there are some that can be destroyed but still be around, so that they can move? Similar questions can be asked about the terms of S. Ethnoscience is concerned with folkscientific conceptions of such matters. The science of human nature tries to find out what is actually happening, to unravel “the anatomy of the mind,” in Hume’s phrase, and the ways its structures and processes are implicated in thought and action. The inquiries are different, though they might use similar data (perhaps intuitive judgments). Similarly, inquiry into the meaning of meaning or of sound might be concerned to discover (I) the semantic features (I-meaning) of the lexical items ‘meaning’ and ‘sound’ in some variety of English; (II) the ideas people have about the general domain of meaning and sound; or (III) the best theory of language and its use. (I) is a question about some (rather idiosyncratic) English words; (II) falls within ethnoscience; and (III) within the science of human nature. (I) and (II) pose perfectly serious questions. Thus pursuing (I),
Internalist Explorations
267
we find that names have no meaning: The question “what does ‘Stalin’ mean?” makes sense only if one is asking about etymology. We find further that the phrase “what does the expression E mean?” shares properties with “what does John weigh?” and “how does John feel?” rather than “what did John eat (say, mean)?” suggesting that what E means might have some kind of adverbial quality. The study of (I) and (II) has little obvious import for (III). Much the same holds of the study of thought, belief, concepts, and so on. 3
Interpretation of Interface Levels
Let’s turn to questions that fall within (III): questions about FL and the states it assumes, and how they are integrated with other components of the mind/brain in language use. One fairly standard assumption, adapting traditional ideas, is that an expression E of L is a pair ·PHON, SEMÒ, where PHON(E) is the information relevant to the sound of E and SEM(E) to its meaning. PHON and SEM are constructed by computational operations on lexical items. Suppose E is a word in isolation. PHON(E) is generally distinct from its I-sound by virtue of phonological operations, but SEM(E) could be identical with the Imeaning of E, depending on the facts about lexical decomposition and the like. PHON(E) and SEM(E) are elements at the “phonetic level” and “semantic level,” respectively; they are phonetic and semantic “representations.” The terms have their technical sense; there is nothing “represented” in the sense of representative theories of ideas, for example.16 These levels are the “interface” between FL and other systems, providing the information used by the sensorimotor apparatus and other systems of language use. There has been a great deal of illuminating work about such representations and how they are constructed by operations of the I-language.17 This work could be considered syntax in the technical sense; it deals with the properties and arrangements of the symbolic objects. On the sound side, the work is sometimes called phonetics, but with the understanding that the study of phonetic features, syllabic and metrical structures, and so on, only contributes to the more general investigation of how information made available by the I-language is used by the sensorimotor systems, and how the whole complex relates to external events. These are the topics of acoustic and articulatory phonetics, going well beyond I-language. The same practice would be appropriate, I think, with regard to the work often called “natural language semantics” and “lexical semantics.” It can be regarded as part of syntax, but oriented to a different interface and different aspects of language use. Insofar as the relation of rhyme that holds between ‘chase’ and ‘lace’ is based on properties of I-sound, and the relation of entailment that holds between ‘chase’ and ‘follow’ on properties of I-meaning, both fall under syntax, in a traditional sense. Virtually all work in syntax in the narrower sense has been intimately related to questions of semantic (and of course phonetic) interpretation, and motivated by such questions.
268
Noam Chomsky
The fact has often been misunderstood because many researchers have chosen to call this work “syntax,” reserving the term “semantics” for relations of expressions to something extralinguistic.18 The earliest work in modern I-linguistics (generative grammar) was concerned with the meanings of such expressions as (1), reviving concerns of traditional grammar. We can usefully distinguish aspects of I-language that are more relevant to sound or to meaning; but phonetics and semantics, in the sense of how language engages the world, lie beyond. Serious questions about the general picture arise at every point, from the assumed architecture of mind to details of implementation. One category of questions has to do with the location of the interface. On the phonetic side, it has to be determined whether sensorimotor systems are in part language-specific, hence within FL, so that the interface level should be “beyond” what is usually taken to be phonetic representation; there is considerable disagreement about the matter. On the semantic side, the questions have to do with the relations between FL and other cognitive systems. At either level, one can offer only reasonable guesses, taken as first approximation. Questions of language-world relation at the phonetic interface have been studied intensively with sophisticated technology, but the problems are hard, and understanding remains limited. Questions about the use of semantic representations are much more obscure. Far less is known about the language-external systems; much of the evidence about them is so closely linked to language that it is notoriously difficult to determine when it bears on language, when on other systems (insofar as they are distinct). And direct investigation of the kind possible for sensorimotor systems is in its infancy. Nonetheless, there is a huge amount of data about how expressions are used and understood in particular circumstances, enough so that natural language semantics is one of the liveliest areas of study of language, though questions of language use remain elusive. 4
Lexical Items
I have taken an expression to be a pair ·PHON, SEMÒ constructed from lexical items LI, each a complex of properties, including I-sound and I-meaning. PHON and SEM are interpreted by language-external systems. At these interface levels, there may be no subunit corresponding to LI. For the phonetic interface, the point is uncontroversial. A good deal of work in syntax/semantics assumes that LIs may be decomposed and reconstructed in the course of computation of SEM. For example, such items as who or nobody might yield operator-restrictor-variable constructions at the level SEM, something like: [[QUx, x a person] [John saw x]]. And there may be other ways in which the semantic properties of LIs are modified or distributed. But for simple words, we can generally assume that SEM = I-meaning (perhaps a reflection of our ignorance).
Internalist Explorations
269
With regard to the semantic component of LIs, alternatives to this picture are common. The questions also tend to be approached somewhat differently in more empirically oriented studies and in conceptual discussions of the nature of meaning and reference. The latter typically regard words and other expressions as phonetic (or orthographic) units, or as dissociated from either sound or meaning; accordingly a word can change its meaning, perhaps even both its sound and meaning, and still be the same word. It is not obvious that these conventions make sense. At least, they have to be explained and justified. The simplest thesis is that an expression E has no existence apart from its properties at the interface levels, PHON(E) and SEM(E) (if these exist). It is a useful heuristic, I think, to pursue analogies between the sound and meaning sides as far as they plausibly go. Specifically, we may ask whether some light can be shed on issues of semantics by looking at phonetic analogues, which often seem less contentious. Consider a “Mentalese” alternative to the picture outlined so far. Instead of taking LI to include I-sound and I-meaning, let us assume that one or the other is missing, or perhaps both. Accordingly, either SEM, PHON, or both are missing at the interface levels. To learn a language is to acquire rules that map LI into some other system of mind, Mentalese, which is interpreted to yield (aspects of) sound and meaning. If I-sound is missing, then LI is mapped into P-Mentalese. If I-meaning is missing, then LI is mapped into SMentalese. Or both. Language itself has no phonology/phonetics, or no semantics, or neither. These are properties of Mentalese. On the phonetic side, there are no such proposals, to my knowledge. On the semantic side, they are common. What is their substantive content, on either side? For concreteness, consider again the words of (2), or the words ‘persuade’, ‘force’, ‘remind’ for X in (3): (2)
chase, lace, follow
(3)
John X-ed Mary to take her medicine.
Suppose the corresponding LIs lack I-sound and that Peter has learned how to map them into regions of P-Mentalese that have phonetic interpretation. Peter knows a lot about the regions and their interpretations. Thus ‘chase’ rhymes with ‘lace’; ‘persuade’ and ‘force’ begin with lip constriction, though in different ways, and ‘remind’ does not; and so on. Standard approaches assign these properties to FL, taking them to be represented in PHON. The P-Mentalese alternative adds an extra layer of complexity and raises new problems, for example: What component of LI indicates the region of P-Mentalese to which it is mapped, if not the I-sound (as conventionally assumed)? At what point in the computation of an expression does the mapping to P-Mentalese take place? How are universal and
270
Noam Chomsky
particular properties of sound expressed in the interpretation of P-Mentalese? For good reasons, such questions have not been raised, and we may drop the matter. Consider the semantic analogue. We now assume that LIs have only I-sound and uninterpreted formal properties, and that Peter has learned how to map them into regions of S-Mentalese, which have semantic interpretation.19 Peter knows a lot about these regions/interpretations too. Thus if Tom chased Bill then Tom followed Bill with a certain intention, not conversely; if X = ‘persuade’ in (3), then John’s efforts were a partial success (Mary came to intend to take her medicine, but may not have done so); if X = ‘force’, John succeeded, but differently (Mary took her medicine, whatever her intentions); if X = ‘remind’, John may have failed (Mary may not have been paying attention), but if he succeeded, then Mary came to remember to take her medicine. The earlier picture assigns the relevant properties to FL, taking them to appear in SEM by virtue of operations on LIs and the constructions in which they appear. The S-Mentalese alternative adds an extra layer of complexity and raises new problems analogous to those of the phonetic counterpart. If we take LIs to have neither I-sound nor I-meaning, then both kinds of problem arise. One can be misled by simple examples, say, ‘snow is white’, or descriptive phrases of S: ‘the sky is dark’, and so on. But problems multiply with even the slightest extension of the paradigm. Consider ‘the rain looks heavy’, ‘the wind feels strong’, . . . ; and in general (4): (4) X (is, looks, tastes, sounds, feels, smells, . . .) Y Even such simple sentences impose translation problems, even for very similar languages. How are they to be translated into universal Mentalese?20 Answers to such questions might yield empirical consequences within more articulated theories of language and Mentalese, perhaps justifying the additional complexity. Standing alone, the proposals can hardly be evaluated. Suppose that we develop denotational theories of interpretation, either directly for linguistic expressions, or for Mentalese translations. With regard to sound, a standard assumption is that in producing or perceiving E, the sensorimotor systems access PHON(E). Instead, let us now suppose that LI has no I-sound but P-denotes some object that is external to the person; call it the Phonetic Value PV of LI (alternatively, of its PMentalese image), and suppose some computation on PVs yields the linguistic component of the sound of E, PV(E). PV could be something about the noises associated with utterances (or possible utterances) of E as circumstances vary (perhaps also as speakers vary, insofar as they are sufficiently alike); a construction from motions of molecules, perhaps. The proposal could be elaborated by taking PV to be determined by social and physical factors of various kinds. One might proceed to an account of communication, translation,
Internalist Explorations
271
acquisition, and other processes in these terms. Thus Peter is able to communicate with Tom because the same PV is denoted by their expressions in the language they share (but only partially know). The proposal leaves all problems where they were, adding a host of new ones. We understand nothing more than before about the relation of E to its external manifestations. The account of communication and other processes is worthless. There is no reason to suppose that such PVs figure in the process by which one person’s mind constructs some version of what another is saying. For such reasons, there are no proposals along these lines. Consider the semantic analogue.21 We now suppose that LI has no I-meaning but that it (or its S-Mentalese image, perhaps an “idea” or “concept”) S-denotes a Semantic Value SV(LI) that is external to the person, some construction from what is being talked about when E is uttered (speakers and circumstances varying), perhaps partially determined by social and physical properties. One might again offer an account of communication, translation, acquisition, and other processes in these terms. Thus Peter is able to communicate with John because their expressions S-denote the same SVs in the shared language that they partially know. We now take the SV for ‘Joe Sixpack’, ‘living wage’, ‘chase’, ‘persuade’, ‘look’, the words of S, and so on (or for the S-Mentalese images), to be Joe Sixpack, living wages, chasing, persuading, looking, the sky, Boston, rivers, damage, loss, power, . . . , while adding something about ‘who’, ‘nobody’, and so on. To account for the semantic properties of E = ‘Chinese is the language of Beijing and Hong Kong’, we take the SVs to be Chinese, language, Beijing, and so on. We would ask whether the external object SV(‘the fate of the Earth’) = SV(‘the Earth’s fate’) for the common language (or for someone who can be said ‘to know it’). We could go on to explore intuitive judgments, whatever that might mean within this quasi-technical array. So far, at least, the original project isn’t advanced, merely restated, with many new problems. We have learned nothing more about how expressions are used and interpreted. Adopting one or the other proposal, we still have to account for the properties of expressions: those of (1)–(4), for example. The phonetic and semantic cases are not the same, of course; only similar, but in what may be informative ways. Suppose we follow a different course, saying that properties of rhyme, inference patterns, and so on, do not relate to language (or its Mentalese images), but have to do with our beliefs about Values: the external objects, whatever they are. On the phonetic side, we say that Peter’s belief that PV(chase) rhymes with PV(lace) has a different status from his other beliefs about PVs (say, about their frequency). Similarly for other properties. Such a proposal has never been entertained, and we can again drop the matter. The counterpart on the semantic side would be to hold that the properties of (1)–(4) are accounted for in terms of Peter’s beliefs about the world; perhaps strength of belief, in
272
Noam Chomsky
Quinean terms. Such proposals are familiar, even close to orthodoxy. To evaluate them, we have to find out more about how beliefs are fixed in these highly intricate and strikingly uniform ways within and among languages, among other questions. Until they are addressed, the proposals are virtually without substance. For the present, it seems reasonable to conclude that the situation is much as on the phonetic side: The semantic properties of the words and constructions are determined by the ways they are constituted, with a rich innate contribution. The problem is to discover the properties of I-sound and I-meaning (whether for LIs, or their S-Mentalese counterparts), the ways they can be combined, the computations that yield interface representations and how these are interpreted by language-external systems. In both domains, there are many open problems, but substantive progress as well. Consider a different approach: The sound and meaning of an expression reduce in part to relations of the sort discussed in connection with (2) and (3). For LI, we have some (finite) pattern of relations to other expressions, phonetic relations Rp and semantic relations Rs, perhaps supplemented with P- and S-denotational properties. Similarly for more complex expressions. For ‘chase’, Rp consists of the properties rhymes with ‘lace’, begins the same way as ‘child’, has the same number of syllables as ‘pin’, and so on; and Rs consists of the relations to ‘follow’, ‘intend’, and so on, and other conceptual and inferential roles. On the phonetic side, the move again seems pointless. The standard featurecomposition approach suffices to express Rp along with other phenomena: the relation of components of ‘chase’ to articulatory gestures and noises, their distributional properties (e.g., consonant-vowel interactions), and so on. Furthermore, Rp(chase) shares properties with Rp(W) for other words W. Numerous facts of this sort are expressible under the standard view that LI is constituted of its properties, which enter into determining its phonetic relations to other expressions and much else. For such reasons, the proposal has never been entertained.22 On the semantic side, again, there are such proposals, and similar questions arise. Thus, Rs(persuade) shares properties with Rs(raise): “causative” properties, which have been extensively studied in many languages, with nontrivial results. A sensible version of LI should express such facts. It should also capture distributional properties that are not (sensibly) stated in terms of inferential and conceptual roles; for example, the fact that ‘deny’, ‘doubt’, ‘refuse’, and so on, occur with polarity items (‘any’, ‘ever’, etc.) in ways that ‘assert’, ‘believe’, ‘accept’ do not, and that in these respects the former are similar to ‘not’, ‘few’ (vs. ‘many’). Standard approaches seek properties of I-meaning and SEM in terms of which a wide array of facts can be expressed and explained, including inferences and their shared and dissimilar properties.
Internalist Explorations
273
So conceived, semantic and phonetic interpretation are somewhat analogous. The expression E consists of the interface representations PHON(E) and SEM(E), computed from LIs. PHON(E) provides information that is used by sensorimotor systems for articulation and perception; SEM(E), information that is used by conceptual-intentional systems to engage the world in different ways as the language user thinks and talks in terms of the perspectives made available by the resources of the mind. The referential use of language can attend in various ways to the component elements of I-meaning and SEM. Individuation commonly turns on such factors as design, intended and characteric use, institutional role, and so on. If something looks to me just like a book but I learn that it was designed to be a paperweight and is characteristically used that way, I could come to agree that it is a paperweight, not a book. Suppose the library has two indistinguishable copies of Middlemarch, Peter takes out one, and Tom the other. If we attend to the material component of the LI, they took out different books; if we focus on its abstract component, they took out the same book. We can attend to both simultaneously, using words with an abstract/concrete character, as in the expressions ‘the book that he is planning will weigh at least five pounds if he ever writes it’, or ‘his book is in every store in the country’. Similarly, we can paint the door white and walk through it. Or consider the word ‘bank’ (savings, river). We can say that (i) the bank burned down and then it moved across the street, that (ii) the bank, which had raised the interest rate, was destroyed by fire, and that (iii) the bank lowered the interest rate to keep from being blown up. Referential dependence is preserved across the abstract/concrete divide. Thus (i) means that the building burned down and then the institution moved; similarly (ii), (iii). But we cannot say that (iv) the bank burned down and then it eroded, or (v) the bank, which had raised the interest rate, was eroding fast, or (vi) the bank raised the interest rate without eroding; (iv) does not mean that the savings bank burned down and then the river bank eroded. The facts are often clear, but not trivial. Thus referentially dependent elements, even the most narrowly constrained, observe some distinctions but ignore others (pronouns, relatives, the “empty category” that is the subject of ‘being blown up’ and ‘eroding’). In the case of ‘bank’, the natural conclusion is that there are two LIs that happen to share the same I-sound (homonymy), and that one of them, ‘savings bank’, is polysemous, like ‘book’: it provides a way of looking at the world that combines abstract and concrete properties, allowing referential dependence across these perspectives.23 Such properties can be investigated in many ways: language acquisition, generality among languages, similar items within the language, invented forms, zeugma, and so on. If systematic similarities and differences persist, conclusions about lexical structure are supported. There is no a priori reason to expect that language will have such properties; Martian could be different.
274
Noam Chomsky
The question, “to what does the word X refer?” has no clear sense, whether posed for Peter, or (more mysteriously) for some “common language.” In general, a word, even of the simplest kind, does not pick out an entity of the world, or of our “belief space”—which is not to deny, of course, that there are banks, or that we are talking about something (even some thing) if we discuss the fate of the earth (or the earth’s fate) and conclude that it is grim; only that we should not draw unwarranted conclusions from common usage. The observations extend to the simplest referential and referentially dependent elements (pronouns, same, re-(‘build’), etc.); or to proper names, which have rich semantic-conceptual properties derived in large part from our nature, with some overlay of experience. Something is named as a person, a river, a city, with the complexity of understanding that goes along with these categories. Language has no logically proper names, stripped of such properties; one must beware of what P. F. Strawson called “the myth of the logically proper name” in natural language, and related myths concerning indexicals and pronouns. We can think of naming as a kind of “worldmaking,” in something like Nelson Goodman’s sense, but the worlds we make are rich and intricate and substantially shared thanks to a complex shared nature. Even the conscious efforts of the sciences and the arts are guided by such properties—fortunately, or they could accomplish nothing (Strawson 1952, p. 216).24 An approach to semantic interpretation in such terms has a traditional flavor. Seventeenth-century rationalist psychology held that innate “cognoscitive powers” enable people “to understand or judge of what is received by the sense,” which only gives the mind “an occasion to exercise its own activity” to construct “intelligible ideas and conceptions of things from within itself” as “rules,” “patterns,” exemplars,” and “anticipations” that provide relations of cause and effect, whole and part, symmetry and proportion, characteristic use (for all “things artificial” or “compounded natural things”), unity of objects and other Gestalt properties, and in general “one comprehensive idea of the whole.”25 “It is manifest,” Hobbes held, that “Names are signs not of things but of our cogitations . . . our conceptions”; the technical notion “sign of X,” holding of words, is better construed in such a way. These “conceptions” can be intricate, as we see from our manner of individuation in terms of constitution, form, origin, and other properties. A man “will always be the same, whose actions and thoughts proceed all from the same beginning of motion, namely, that which was in his generation; and that will be the same river which flows from one and the same fountain, whether the same water, or other water, or something else than water, flow from thence [as in the classical case of the ship of Theseus, Hobbes adds]; and that the same city, whose acts proceed continually from the same institution.” The inquiry into personal identity from Locke to Hume was concerned with organic unity, a broader notion. A tree or an animal “differs from a mass of matter,” Locke noted, by virtue of the “organization of parts in one coherent body, partaking of one common life” with “continued organization” that comes from within, unlike artifacts. The identity
Internalist Explorations
275
of an oak resides in “a sympathy of parts” contributing to “one common end” of “support, nourishment and propagation” of the form, Shaftesbury added. Hume largely agreed, though taking “the identity, which we ascribe to the mind of men,” and “the like kind . . . that we ascribe to vegetables and animal bodies,” to be “only a fictitious one” established by the imagination, not Shaftesbury’s “peculiar nature belonging to this form.” John Yolton makes a strong case that the core of the theory of ideas from Descartes to Reid took ideas to be “not things, but ways of knowing,” “not signs of the corpuscular structure, but signs in terms of which we know of or are acquainted with experience,” so that “The world as known is the world of ideas, of significatory content” (Hobbes 1889, pp. 16ff).26 Hume’s conclusion gains more force as we look more closely at the intricacy of the concepts. “[Person] is a forensic term,” Locke observed, “appropriating actions and their merit; and so belongs only to intelligent agents, capable of a law, and happiness, and misery,” as well as accountability for actions, and much else. Individuation of rivers and cities involves factors well beyond origin. The flow of a river can be reversed, or it may be diverted to a different course or even divided into streams that may later converge, or changed in all sorts of other ways, and yet remain the same river, under appropriate circumstances. The press reports intelligibly that scientists “have discovered the source of the Amazon” in an unexpected place, the sole source, though usually “rivers start off as myriad little ones.”27 Locke notes that the oak remains the same when a branch is lopped. Suppose the oak is transplanted elsewhere and replaced in its original location by the branch, which grows into a replica of the tree while the transplanted oak withers and dies— but is still the original tree, according to the fictitious identity established by innate cognoscitive powers. This barely touches the surface. Proceeding further, we find that these powers impose a rich framework of interpretation and understanding, which we would expect to be only marginally influenced by experience, as in the case of other complex organic structures. From such ideas about internally generated modes of cognition to which experience conforms, it is a short step to an analysis in terms of semantic features, or what Julius Moravcsik calls the “(generative) factors” of lexical structure.28 Recasting the enterprise in these terms, we try to unravel the anatomy of the mind, including FL and the systems at the interface, and to discover how experience and social interaction are shaped in terms of these internal resources. 5
Some Questions of Legitimacy
It is commonly held that this version of the science of human nature is needlessly complex, or misguided in principle. On one view, the evidence adduced for principles of FL “is
276
Noam Chomsky
much more simply accounted for by the . . . hypothesis” that FL indeed is “innate in human brains” but we need only say that there is “a hardware level of explanation in terms of the structure of the device” and “a functional level of explanation, describing which sorts of languages can be acquired . . .” (John Searle). Or, we should dispense with FL altogether in favor of the “competing hypothesis” that the innate structures of the brain “have as their original and still primary function the organization of perceptual experience, the administration of linguistic categories being an acquired an additional function for which evolution has only incidentally suited them,” thus overcoming the problem of accounting for the evolution of language among “other advantages” (Paul Churchland).29 That there is a “hardware level” is not in contention, if by that we mean that atoms, cells, and so on, are presumably involved in “the structure of the device” FL that is “innate in human brains.” For the moment, we can only follow Joseph Black’s good advice and construct a “body of doctrine” about FL; with progress toward unification there could be more to say—perhaps, as in the case of chemistry, that current assumptions about the “hardware” are misconceived. The “body of doctrine” is concerned with “which sorts of languages can be acquired” and also their properties, their interactions with other systems, the manner of their acquisition and use, unification problems, and anything else that lends itself to useful investigation. Working this out, we seem to be led back to the “deep unconscious rules” that Searle regards as dispensable. Searle is right that there is “no further predictive or explanatory power . . . by saying that there is in addition [to the hardware and functional level] a level of deep unconscious rules” of FL. But what has been proposed is quite different: specific structures and principles of FL, which yield at least a partial account of properties of language. Similarly chemistry is uninteresting if it says only that there are deep structural properties of matter, anything but as a body of doctrine about these is developed. At best, the debate is rather reminiscent of past controversy over whether chemical properties, molecular structure, and so on should be attributed to matter or regarded simply as calculating devices; all pointless, as largely agreed in retrospect, and falling under Burge’s apt observation that questions of ontology and the like are “epistemically posterior to questions about the success of explanatory and descriptive practices” (Burge 1986a).30 Churchland’s proposal could become a “competing hypothesis” if spelled out sufficiently to say something about the most elementary properties of language (discrete infinity, structure dependence, etc.) and on to the properties of (1) and others like them.31 It would also be necessary to deal with the fact that we do not find, as apparently predicted, uniformity of cognitive development and attained structures across domains, similarities of language use among species with similar modes of organizing perceptual experience, no dissociation of function under disabilities, homogeneity of brain structure, and so on.
Internalist Explorations
277
A more considered challenge is presented by Hilary Putnam in his critique of “MIT mentalism,” in part the view outlined so far (which he attributes to Fodor and to me).32 His goal is to “destroy the theory of innate semantic representations,” call it TISR, which asserts: (5)
[1] There are “semantic representations” in the mind/brain; [2] these are innate and universal; [3] all our concepts are decomposable into such semantic representations.
TISR holds further that the mind is a “Cryptographer”: “the mind thinks its thoughts in Mentalese, codes them in the local natural language, and then transmits them” to a hearer who “has a Cryptographer in his head too, of course, who thereupon proceeds to decode the ‘message’ ” in the lingua mentis. TISR goes well beyond I-linguistics. That representations generated by I-language map into a lingua mentis is a separate hypothesis. [3] also goes beyond the study of language, which has to do with FL, not other cognitive systems, which could be (and I suppose are) different in character. [2] requires clarification. Only the elements of which representations are constructed are taken to be innate (hence universal, available generally though perhaps unrealized). Thus the components and mode of composition of phonetic representation are presumably innate, but the representations are not; they differ for English and Japanese, even among siblings. The same is true of whatever is involved in fixing meaning—“semantic representations,” or something else. Languages differ in this regard, one of the many problems that bedevil translators. There is no controversy about this, nor, presumably, about the thesis that the elements of whatever is involved in fixing meaning are innate. It is hard to imagine an alternative. There are empirical grounds for believing that variety is more limited for semantic than for phonetic aspects of language. Phonetic data are available to the child in abundance, and the gap between target attained and data available seems narrower than for semantic subsystems. If so, variety is more easily tolerated. The study of meaning has to face the fact that extremely limited exposure in highly ambiguous circumstances suffices for children to come to understand the meanings of words and other expressions with remarkable delicacy, far beyond anything that the most comprehensive dictionaries and grammars begin to convey, with refinements and intricacy that are barely beginning to be understood. For such reasons, empirical inquiry has sought to discover semantic properties that are innate and universal. These problems have to be faced whether one adopts an I-linguistic (or more broadly, TISR) framework or any other. Putnam’s position seems to be that mechanisms of general intelligence suffice. Hence these must have the innate structure required to carry the mind
278
Noam Chomsky
from the data available to the cognitive systems attained. For language, the problem is now displaced from FL to general intelligence. We now face the problems that confront the “competing hypothesis” that everything reduces somehow to perceptual organization. The prospects look as unappealing as before, but there is nothing to discuss until something specific is proposed. For language, the thesis that Putnam aims to destroy is now reduced to (6): (6) [I] There are “semantic representations” in the mind/brain. [II] They are constructed of elements that are innate. [II] is innocuous if [I] holds. But [I] has nothing particular to do with “MIT mentalism.” Empirical semantics generally assumes something similar. Suppose nonetheless that [I] is false. Thus, neither FL nor any other system of the mind/brain involves “semantic representations.” But some internal state is involved in how we understand sentences, say S or (1). The alternative to (6), then, holds that such states do not involve “semantic representations.” Apparently, the intended alternative keeps the assumptions about states of the mind/brain relating to sound, and perhaps also those concerning the structural properties of FL that enter into establishing the meaning of expressions, but not “semantic representations.” The specific intricate knowledge that the child has acquired and uses is represented in the mind/brain somehow, but not in the manner developed in studies of natural language semantics, now cutting a very broad swath. That is not unlikely; current phonetic theory may also turn out to be wide of the mark. But again, comment is impossible. Putting this aside, let’s look at Putnam’s critique of [I]. It has several strands. One is that “meaning is holistic.” In the Quinean formula, sentences meet the test of experience “as a corporate body,” and revision can strike anywhere. For the sciences, the formula seems fair enough; Carnap apparently agreed, though preferring a different formulation (see Uebel and Hookway 1995). But the questions here have to do with human language, a biological object, not with the sciences that humans construct, using different faculties of mind, so it appears. Putnam holds, however, that “the language of ordinary life” has the same holistic properties as the sciences. The reason is that everyday discourse relies on unstated assumptions, so that “if language describes experience, it does so as a network, not sentence by sentence.” But language does not “describe experience,” though it may be used to describe or misdescribe it, or in countless other ways. The fact that hidden assumptions enter into use of language tells us nothing relevant here. Another strand of Putnam’s critique turns on scientific practice. Right or wrong, these arguments do not bear on human language, or other aspects of human thought, except on assumptions about uniformity of mind that surely require justification, so far lacking. Other parts of the argument rely on conclusions about lingua mentis and “public language,” and
Internalist Explorations
279
intuitions about synonymy, translation, and other matters, none of which would be relevant here even if tenable (throughout, I am skeptical; see Chomsky 1995). The rest of the argument has to do with “Chomsky’s Innateneness Hypothesis.” I have never understood what that is supposed to be. It is often refuted, but never formulated or defended, to my knowledge. Presumably, cognitive capacities, like all others, are rooted in biological endowment, and FL (if it exists) is some kind of expression of the genes. Beyond that, I know of no Innateness Hypothesis, though there are specific hypotheses about just what is innate. Putnam seems to identify the “Innateness Hypothesis” with (a) the thesis that the lingua mentis is innate and (b) the thesis that “the mental vocabulary” is innate. I-linguistics is not committed to (a) or (b)—at least, insofar as I understand these theses; admittedly, not very far. Whatever their content, furthermore, they are presumably distinct: The lingua mentis is not the mental vocabulary, just as English is not its vocabulary. Putnam then turns to the arguments that are widely alleged to undermine not only “MIT mentalism” but also an approach to the study of meaning and reference that reaches from Aristotle to Mill, Russell, Frege, and Carnap, the tradition that holds (A) and (B): (7) (A) When we understand a word or any other “sign,” we associate that word with a “concept”; (B) the concept determines the reference of the word (or sign). Putnam takes (7) to be refuted by the fact that reference is determined in part by “the division of linguistic labor” and the “contribution of the environment.” I-linguistics has no commitment to (7); nor could it, without some explanation of the technical notions. At most, I-linguistics is committed to (8): (8) (A¢) When X understands the word W, X makes use of its properties. (B¢) The properties might include I-sound and I-meaning, and if so, the latter play a part in determining what X refers to in using W. Beyond that, the chips fall where they may. The critique of (7) does not seem to bear on at least the I-linguistic component of “MIT mentalism,” but let us look at it anyway. To illustrate the division of linguistic labor, Putnam considers the word ‘robin’ in British and American English. Suppose PeterE in England and PeterUS in the United States are in relevant respects the same, and are unaware that (9): (9) The word ‘robin’ does not refer to the same species of bird in England and in the United States.
280
Noam Chomsky
PeterE and PeterUS have the same word ‘robin’ in their I-languages, but it has different extensions because “reference is a social phenomenon” involving reliance on experts. We must therefore abandon the traditional thesis (7). Taking (9) to be a statement of fact about language-world relations, we want to determine whether it is true. We first have to understand its terms: specifically, the phrases ‘the word “robin” ’ and ‘refer’, a relation alleged to hold between ‘the word “robin” ’ and a biological species. Let’s grant (much too quickly) that we understand well enough what is meant in speaking of ‘the word “robin” ’, an entity in a “public language” (as intended). What about ‘refer’? People use words to refer to things in various ways, but English has no term ‘refer’ or ‘reference’ used in the sense of (9);33 nor do similar languages, one reason why Frege had to make up technical terms and why there is much variation as to how to translate them, some preferring Latin words that make clear the technical status. Some work has to be done, then, to make it possible to evaluate (9) as an empirical claim. The context (resort to thought experiments, etc.) suggests that (9) is to be understood within the study of folk theories. If so, the conclusions have no obvious bearing on I-linguistics; or perhaps even on the tradition, if understood as offering a kind of rational reconstruction. Let us ask nevertheless whether (9) is well grounded within the study of folk theory. To avoid the (as yet unexplained) technical terms, let us select some ordinary English counterpart, perhaps (10): (10) PeterUS uses the word ‘robin’ to refer to one species of bird, and PeterE to refer to a different species. Is (10) true? The birds PeterUS has called robins are different in all sorts of ways from the ones PeterE has called robins, but that’s also true of PeterUS and his friend Charles, who have been neighbors all their lives. We have to know much more to evaluate (10). Suppose we ask what PeterUS would say if he went to England and saw the red-breasted things there. By assumption, he would call them robins, so this gets us nowhere. Suppose Jones would say that PeterUS is making a mistake when he calls the birds in England robins (I wouldn’t). We are then learning something about Jones that is of no relevance here. Jones may be assuming something like thesis (9). Perhaps Jones holds that PeterUS’s concept robin doesn’t cover the species in England; and that Earthly Oscar’s concept water doesn’t cover Twin Earth XYZ. But now we are back to the original query: How do we find out whether Jones’s claims are true? Suppose that PeterUS’s cousin Bill lives in a part of the United States where the birds called robins belong to a different subspecies. If PeterUS visits Bill and calls the thing on his lawn a robin, is he making a mistake? Can he understand Bill’s talk about robins? Suppose that PeterUS’s wife Mary grew up in his neighborhood but spent part of her child-
Internalist Explorations
281
hood in England. What is Mary referring to when she talks about robins? As cases vary, judgments do too, in all sorts of ways, and are often highly uncertain. The case does not seem problematic for “MIT mentalism.” By assumption, the above characters, alike in relevant respects, would have the same judgments about what is a robin. Further conclusions about whether they are right or wrong, or how “the word ‘robin’ ” is used to refer in “public languages,” or about their beliefs, raise other questions that may or may not be worth investigating once given a clear enough formulation. There seems little more to say. To illustrate “the contribution of the environment,” Putnam adduces Twin Earth and other arguments, all based on assumptions about what “a typical person would say” under various circumstances. Again, the arguments have no direct bearing on a theory of language T that adopts thesis (8). The most they could show is that T (or TISR) does not yield a full explanation of linguistic behavior or capture ordinary usage, but that is obvious in advance. The arguments (for ‘water’) are based on the assumption that water is H2O. To assess the status of this statement we have to know to what language it belongs. Not English, which has no word ‘H2O’. Not chemistry, which has no word ‘water’ (though chemists use the word informally). We could propose that chemistry and English belong to some “superlanguage,” but it remains to explain what this means (see Bromberger and Halle 1996). Putting such qualms aside, is it true that a typical speaker relies on constituency in deciding whether to call something water? Suppose two glasses G and G¢ are on the table, G filled from the tap and G¢ from a well. Suppose a tea bag is dipped in G¢. The contents of G and G¢ could be chemically identical: Maybe tap water comes from a reservoir that uses a “tea filter” to remove contaminants. Knowing that the contents are identical, I would say that the stuff in G is water, not tea; and that the stuff in G¢ is tea, not water. I suspect this is typical. Constitution is a factor in deciding whether something is water, but not the only one.34 The situation recalls the case of ‘book’ and others like it. Here too we can arrange circumstances so that we will attend to constitution, not other factors, in deciding what we are talking about. Under such circumstances, we might call the contents of both G and G¢ water. Empirical study might show that constitution is more of a core factor for ‘water’ than for ‘book’; presumably so, but that would still have no bearing on (8). In ordinary cases, there are no answers except in terms of complex and varying circumstances and interests that yield what Akeel Bilgrami (1992) calls “locality of content.” If, for example, Mary believes that there is water on Mars, and something is discovered there that she regards as water though it has the internal constitution of heavy water or XYZ, there is no general answer as to whether her belief is right or wrong.
282
Noam Chomsky
Reference to expert use adds new quandaries. A recent technical article opens by saying that “Glass, in the popular and basically correct conception, is a liquid that has lost its ability to flow,” and goes on to conclude that “most of the universe’s water exists in the glassy state (in comets, . . .),” as “naturally occurring glassy water” (Angell 1995). Suppose the tea-water scenario just described took place on Twin Earth, where they happen to make their glasses from tails of Earthly comets. Suppose earthly Oscar arrives on Twin Earth and asks for water, pointing to G. Is he right if he is referring to the glass and wrong if he is referring to its contents? My judgments are reasonably clear, and I suspect typical. Looking at the issues from a different standpoint, take Albert and Bill to be relevantly alike, and A, B to be indistinguishable apples, A an object of Albert’s experience, B of Bill’s. Each thinks about, looks at, and takes a bite of their respective apples, leading to identical state changes throughout. Shall we say that the thoughts, visual images, tastes, weight changes, and so on are the same for Albert and Bill but “directed” to different things? Or different for Albert and Bill, the external objects A, B being “part of” the thoughts, and so on? Hearing indistinguishable renditions of the statement S, do Albert and Bill have the same auditory and understanding experiences directed to different objects, or different ones incorporating the objects? Ordinary English may tolerate the “externalist” usage for thought and understanding rather than weight changes, though what we would learn from this is unclear. The science of human nature is too primitive for the question to arise. An internalist picture seems appropriate, though incomplete in the uninteresting sense that a study of Albert and Bill in their environments takes the latter into account. Ordinary examples are often more complex. Take a version of Kripke’s puzzle. Suppose that Peter says: “I used to think that Constantinople and Istanbul were different cities, but now I know they are the same,” adding: “but Istanbul will have to be moved somewhere else, so that Constantinople won’t have an Islamic character.”35 Has he adopted new lexical items? New beliefs? Something different? If, referring to Istanbul, he says “it will have to be moved and re-built elsewhere” (while remaining the same city), how are we to interpret the italicized items—which behave differently in curious ways as examples vary?36 We can only proceed sensibly as indicated earlier, it seems. Consider the issue of fallibility: Clearly we want to be able to say that Peter might be mistaken in calling something an X. Thus Peter might misdescribe the contents of G¢ as water, not knowing that it is tea, not water. Or he might mistake a paperweight for a book. Perhaps Peter is mistaken by his own lights: He would not call it X if he were aware of the facts. Or perhaps we are adopting a standpoint that relies on constitution to decide whether he is right or wrong, so what Peter takes to be water might be something else, maybe heavy water, or XYZ. Such moves are standard in the sciences, but that they are appropriate for natural language, and if so, in what respects, has to be shown. It would
Internalist Explorations
283
be necessary to outline the theoretical framework in which the questions are being posed, and if it uses such notions as concept, to define them in non-question-begging ways; not, say, by stipulating that concepts are specified by internal constitution. There is no clear question, hence there are no straightforward answers. Suppose that young Charlie has experiences that lead him to recognize that his usage differs from that of adults in his community.37 Suppose at stage 1 he referred to streamlined aquatic animals as fish and very large ones as whales. Finding that adults adopt a different usage for the nearest counterparts (pronouncing the words differently too), he moves to stage 2, adapting himself to adult usage, consciously or not. How do we describe what happened? Some might be inclined to say that what Charlie thought about whales and fish in stage 1, and the way he used the words and pronounced them, was wrong. By stage 2, he had corrected himself. He is improving his knowledge of English, the language of his community (ordinary usage provides no way of referring to his linguistic system at stage 1). Search for further understanding can follow the usual two courses. We can seek to learn more about how people talk and think about such matters, or about what is actually happening. An I-linguistic account is straightforward, though incomplete, in part because of its scope, in part because of lack of understanding within its scope. In stage 1, Charlie has Ilanguage L1 with lexical items ‘fish1’ and ‘whale1’. In stage 2, his I-language L2 has ‘fish2’ and ‘whale2’, differing somewhat in properties. The phonological features are different (by assumption); but the status of the semantic features is unclear. Do the new items have different features, incorporating the new criteria for referring to aquatic animals? Do they select different regions in a lingua mentis, conceptual space, belief system? Something else? What Charlie calls things will change in various ways, depending on accidental facts: for example, whether the large aquatic animals with which he had some acquaintance in stage 1 happened to be mammals or tuna fish. We could look for principles that enter into whatever happened and ask to what extent it could have followed another course had circumstances differed. So little is known about these topics that we can only speculate, but no obvious problems of principle arise. The enterprise would not be carried forward by invoking “the real meaning (denotation)” of words in a “common language” that is partially known and shared, “the collective mind,” “words” that remain constant as pronunciation and usage varies, and other such notions that remain mysterious. Suppose we approach the matter in terms of a notion of reference in a common language, perhaps a causal theory. We would then have to determine whether or not the denotations of ‘whale’ or ‘fish’ remained constant as Charlie changed what he calls things (including the objects of his earlier experience), and what happened to the content of his thoughts. If the technical notions are clarified, it might be possible to formulate
284
Noam Chomsky
significant empirical questions about how people think about these matters in one or another cultural and linguistic setting. For the science of human nature, it does not seem to me a very promising course. Consider finally a case discussed by Burge (1986b), illustrating an interesting genre. Suppose A shares with other English speakers the word ‘sofa’ and relevant experiences with things they call sofas. But he comes to believe that sofas “function not as furnishings to be sat on but as works of art or religious artifacts,” and are not “preeminently for” sitting. A and others agree about which things of their common experience are sofas, but disagree about the function of sofas; they may also disagree about whether sofas have really been used for sitting (A thinking that others are deluded about this). If A’s doubts prove well founded, Burge concludes, “the conventional meaning of ‘sofa’ would have to change,” but “it might remain appropriate . . . to attribute propositional attitudes involving the notion of sofa,” as just described. How might such events be described in the internalist framework, extending it now to the assumption that there is an I-conceptual and I-belief system alongside of I-language? Initially, A and others have the same LI ‘sofa’, the same I-concept sofa, and the same I-beliefs about sofas. Call this shared complex SOFA. Within it, sofas are identified as artifacts with certain physical properties and functions. For A, SOFA changes to SOFA¢ with a shift of beliefs about what sofas are for. Someone else, call him B, might change his beliefs about constitution, concluding that sofas are typically flat surfaces with iron spikes, though still used for sitting; for B, SOFA changes to SOFA≤. All agree about which of the things around them are sofas, but A differs from others on the function, and B on the constitution, of the category to which these things belong. So far, there is no difficulty in describing the events and the (I)-mental states of the participants. But we have said nothing about what happened to conventional meaning, thoughts, and beliefs as the story unfolds; or about where in SOFA the changes took place. The first question cannot be addressed until the notions are clarified. The second could be relevant here, but it is still unanswerable. By assumption, changes in the I-belief component of SOFA took place, but this leaves open the question whether A and B changed the LIs of their I-languages or some other aspect of the complex SOFA. Whatever the answer, a straightforward account seems available. Burge argues that it would be “unacceptably superficial” to say that A changed his language when his doubts arose, because “we have no difficulty understanding that he is raising questions about what sofas really are,” and know how to investigate the questions. But granting all of this, we still don’t know whether A changed his I-language, replacing one LI by another. If his I-language remained fixed, he would now be saying that what people thought about sofas was wrong; if it changed as indicated, he would now be saying that people were mistaken in calling these things sofas—they were really something else.
Internalist Explorations
285
Either way, we can understand his questions and know how to investigate them. There are empirical issues lurking here, and perhaps they can be extricated. But it is not clear that anything more is at stake. Similar questions arise about whales and fish. Suppose whales are considered to be fish in Peter’s community, but he decides that a different classification would make more sense, and revises his usage accordingly. Again, we have no difficulty understanding that he is raising questions about whales and fish (what they “really are,” perhaps, though it is not obvious that this is the most apt locution), and we know how to investigate those questions. Inquiry into such cases in their dazzling variety seems to yield answers that vary widely under slight changes of assumed circumstances, arousing some skepticism about how much can be learned by proceeding in this way. However that may be, such phenomena do not seem to me to bear on the soundness of internalist approaches to linguistic and other mental aspects of human life, as far as they can reach, or to suggest a preferable alternative. Notes 1. See Donald Griffin and Steven Austad, in Gajdusek (1994). 2. See Rhum (1993) for comment. 3. Quotes from Lormand (1996). 4. See Mehler and Dupoux (1994), Spelke (1990). 5. On some analogies, and a number of issues by-passed much too quickly here, see Chomsky (1995). 6. For discussion, see Chomsky (1966, 1968) and later publications, including (1995). La Mettrie, cited by Wellman (1992), p. 147. On Newton’s struggles with the basic problem, see Dobbs and Jacob (1995). 7. See Yolton (1983) chapters I and VI; Chomsky (1995); and Jacob (1991), p. 200. 8. John Searle and I have discussed these issues for some years. We apparently agree on the incoherence of monism, dualism, materialism, etc. (cf. Searle 1992, p. 25, and Chomsky 1968), and on the essential accuracy of eighteenth-century conceptions of mind/body of the kind just mentioned. But not on how to account for the properties of language. See below. 9. Note that I do not agree that the choice lies between interpreting “grasp and understanding as conscious states” or as “mere training-induced reaction patterns” (Gaifman 1996, endorsing a view that he attributes to Michael Dummett). Understanding (of (1), S, etc.) appears to involve states and processes that fall under neither category. 10. For some discussion, see Chomsky (1975b), (1986). 11. On how it is accessed, there are various ideas. See Halle and Marantz (1993), for critical discussion of some of these and a “late insertion” alternative. I will ignore all such matters here. 12. Stich (1996), pp. 38ff., reporting (not advocating) standard formulations, which he distinguishes from (I-) linguistics and the “proto-science” of reference. 13. Stich (1996) looks at the matter somewhat differently. 14. See Stich (1983), p. 62n.; Fodor (1994), app. B. Note that there is no contradiction in accepting Wittgenstein’s cautionary remarks on these matters along with quite strong conclusions about invariants of sound and meaning.
286
Noam Chomsky
15. Thomas Reid is the best known of those who argued in the manner of modern ordinary language philosophy that the conception of an idea as “the object that the mind contemplates” is based on a misinterpretation of surface grammar; his argument could be extended to thought, belief, and other cases. On ideas as objects of thought or modes of mind in seventeenth- and eighteenth-century thought, see Yolton (1984), who argues that Reid and other commentators have misread the tradition; see below. 16. In the earliest work of the kind considered here, it was assumed that an I-language generates “markers” at the several linguistic levels (phonetic, word, phrase structure, etc.), each “representing” PHON(E) as a predicate holding of it. Thus PHON(E) is a . . . , where . . . is its phonological (word, phrase structure, etc.) “representation” (for details, see Chomsky 1975a). PHON(E) (hence indirectly, markers at all levels) could be taken to “represent” utterances in a similar way. Since utterances are associated with states of speakers, the predication could be construed as holding of these, the course taken by Bromberger and Halle (1996), discussing phonological levels in terms of intentions of speakers (understood as supervening on brain states). Their purpose is to compare competing theories, a good reason for more careful foundational work, which has otherwise rarely been undertaken. 17. On the semantic side, see, inter alia, Larson and Segal (1995), Pustejovsky (1995), and sources cited. 18. For similar reasons, while the thesis of “autonomy of syntax” has been vigorously rejected, it has never been defended, to my knowledge; nor formulated in any intelligible way by its opponents. 19. See Fodor (1990), chap. 7, a review of Stephen Schiffer’s Remnants of Meaning (1987), for several versions of such views. 20. For similar reasons, a theory of T-sentences runs into problems when object and metalanguage differ, so that informativeness of nonhomophonic T-sentences does not provide good grounds for justifying the approach. Whatever its merits, which are real, it leaves untouched the question of how language engages the world, much of the heart of the traditional theory of meaning. See also Fodor (1990). 21. Not to be confused with it is postulation of semantic (or phonetic) Values as mental entities, with (LI, Value) relations that have formal properties of refer and denote in their technical sense. That has to be assessed alongside of postulation of other syntactic objects. It seems to me appropriate (though unconventional) to construe much work in natural language semantics in these terms. 22. One might, perhaps, understand some structuralist proposals along these lines, but that would be a dubious interpretation, I think. 23. On some traditional problems, often obscure and complex, see Lyons (1977), II, 13.4. 24. For some further discussion, see Chomsky (1975b, 1995). 25. Quotes from Ralph Cudworth, but the point of view is general; and influential, at least in the Kantian version. See Chomsky (1966). 26. Other quotes here and below from Mijuskovic (1974), pp. 97–113; Yolton (1984), pp. 213ff. 27. Tim Radford, Guardian, July 24, 1996. 28. Adapting Aristotelian notions and applying them broadly to lexical semantics, Moravcsik (1975, 1990) takes the factors to be “constituents, structure, function, and agency.” See Chomsky (1975b) for some comment, and Pustejovsky (1995), for elaboration of similar ideas. 29. Searle (1992), 244; I am overlooking irrelevant terminological differences. Churchland (1981). 30. See Chomsky (1986, pp. 250f.), (1995); also note 8, above. Searle argues further that postulation of unconscious rules is illegitimate, but on grounds that seem to me without merit; see Chomsky (1990). His reductio using the analogy of a “vision faculty” is not relevant because the principle he rightly rejects lacks any explanatory force. 31. There has been serious work with a vaguely similar flavor, both traditional and modern. See, e.g., Jackendoff (1994), chap. 14, and sources cited. 32. Putnam (1986). I will put aside questions of accuracy of attribution where not relevant. 33. The observation is familiar; see, e.g, Strawson (1950). 34. For some experimental work concluding that H2O content is only weakly correlated with judgments about what is water, or even prototypical water, see Malt (1994). Braisby et al. (1996) review various ideas and exper-
Internalist Explorations
287
imental work on such matters, and present findings of their own, which, they argue, “show that natural kind terms are not employed in an essentialist manner.” Understanding is limited, hence confidence in interpretation of data. 35. For real examples of this kind see Chomsky (1995). 36. See Chomsky (1995) for some illustrations; also above. 37. There are many interesting insights on such cases in papers by Tyler Burge, among them (1986b, 1989). It is not entirely clear to me if, and if so where, we differ substantively about them. See Mercier (1992) for one interpretation.
References Angell, C. A. 1995. Formation of Glasses from Liquids and Biopolymers. Science 267: 1924–1935. Bilgrami, Akeel. 1992. Belief and Meaning: The Unity and Locality of Mental Content. Oxford: Blackwell. Braisby, Nick, Bradley Franks, and James Hampton. 1996. Essentialism, Word Use, and Concepts. Cognition 59(3): 247–274. Bromberger, Sylvain. 1996. Natural Kinds and Questions. In Essays on Jaakko Hintikka’s Epistemology and Philosophy of Science, Matti Sintonen (ed.). Poznan Studies in the Philosophy of Science and the Humanities. Amsterdam: Rodopi. Bromberger, Sylvain and Morris Halle. 1996. The Content of Phonological Signs. Unpublished ms. Brook, Andrew. 1994. Kant and the Mind. Cambridge: Cambridge University Press. Burge, Tyler. 1986a. Individualism and Psychology. Philosophical Review 95: 3–45. ———. 1986b. Intellectual Norms and Foundations of Mind. Journal of Philosophy 83(12): 697–720. ———. 1989. Wherein Is Language Social? In Reflections on Chomsky, A. George (ed.). Oxford: Blackwell. Chomsky, Noam. 1966. Cartesian Linguistics: A Chapter in the History of Rationalist Thought. N. Chomsky and M. Halle (eds.). New York: Harper and Row. ———. 1968. Language and Mind. New York: Harcourt Brace Jovanovich. ———. 1975a. Logical Structure of Linguistic Theory. New York: Plenum (excerpted from unpublished 1956 ms.). ———. 1975b. Reflections on Language. New York: Pantheon. ———. 1986. Knowledge of Language. New York: Praeger. ———. 1990. Accessibility “in Principle.” Behavioral and Brain Sciences 13(4): 600–601. ———. 1995. Language and Nature. Mind 104(413): 1–61. Churchland, Paul. 1981. Eliminative Materialism and the Propositional Attitudes. Journal of Philosophy 78: 67–90. Reprinted in Folk Psychology and the Philosophy of Mind, Scott Christensen and Dale Turner (eds.), Erlbaum (1993). Dobbs, Betty Jo and Margaret Jacob. 1995. Newton and the Culture of Newtonianism. Atlantic Highlands, NJ: Humanities Press. Fodor, Jerry. 1983. Modularity of Mind. Cambridge, MA: MIT Press. ———. 1990. A Theory of Content. Cambridge, MA: MIT Press. ———. 1994. The Elm and the Expert. Cambridge, MA: MIT Press. Gaifman, Haim. 1996. Is the “Bottom-Up” Approach from the Theory of Meaning to Metaphysics Possible? Journal of Philosophy 93(8): 373–407. Gajdusek, Carleton, Guy McKhann, and Laiana Bolis, eds. 1994. Evolution and Neurology of Language, Discussions in Neuroscience. 10(1–2), April (special issue).
288
Noam Chomsky
Halle, Morris and Alec Marantz. 1993. Distributed Morphology and the Pieces of Inflection. In The View from Building 20, K. Hale and S. J. Keyser (eds.). Cambridge, MA: MIT Press. Hobbes, Thomas. 1889. The English Works of Thomas Hobbes, vol I. William Molesworth (ed.). Oxford: Oxford University Press. Holton, Gerald. 1996. On the Art of Scientific Imagination. Daedalus 125: 183–208. Hume, David. 1975. Enquiries Concerning Human Understanding and Concerning the Principles of Morals, third edition. Oxford: Clarendon Press. Jacob, Margaret. 1991. Living the Enlightenment. New York: Oxford University Press. Jackendoff, Ray. 1994. Patterns in the Mind. New York: Basic Books. Lange, Friedrich Albert. 1925. The History of Materialism. London: Kegan Paul. Larson, Richard and Gabriel Segal. 1995. Knowledge of Meaning. Cambridge, MA: MIT Press. Lormand, Eric. 1996. How to Be a Meaning Holist. Journal of Philosophy 93(2): 51–73. Lyons, John. 1977. Semantics. 2 vols. Cambridge: Cambridge University Press. Malt, Barbara. 1994. Water Is Not H2O. Cognitive Psychology 27: 41–70. Mehler, Jacques and Emmanuel Dupoux. 1994. What Infants Know. Oxford: Blackwell. Mercier, Adele. 1992. Linguistic Competence, Convention, and Authority: Individualism and anti-Individualism in Linguistics and Philosophy. Ph.D. dissertation, Philosophy, UCLA. Mijuskovic, Ben Lazare. 1974. The Achilles of Rationalist Arguments. Martinus Nijhoff. Moravcsik, Julius. 1975. Aitia as Generative Factor in Aristotle’s Philosophy. Dialogue 14: 622–636. ———. 1990. Thought and Language. London: Routledge. Pateman, Trevor. 1987. Language in Mind and Language in Society. Oxford: Oxford University Press. Pustejovsky, James. 1995. The Generative Lexicon. Cambridge, MA: MIT Press. Putnam, Hilary. 1986. Meaning Holism. In The Philosophy of W. V. Quine, L. E. Hahn (ed.). La Salle: Open Court. ———. 1988. Representation and Reality. Cambridge, MA: MIT Press. Rhum, Michael. 1993. Understanding “belief.” MAN 28(4). Schiffer, Stephen. 1987. Remnants of Meaning. Cambridge, MA: MIT Press. Schofield, Robert. 1970. Mechanism and Materialism. Princeton: Princeton University Press. Searle, John. 1992. The Rediscovery of the Mind. Cambridge, MA: MIT Press. Spelke, Elizabeth. 1990. Origins of Visual Knowledge. In An Invitation to Cognitive Science, II, D. N. Osherson, S. M. Kosslyn, and J. M. Hollerbach (eds.). Cambridge, MA: MIT Press. Strawson, Galen. 1994. Mental Reality. Cambridge: MIT Press. Strawson, P. F. 1950. On Referring. Mind 59: 320–344. ———. 1952. Introduction to Logical Theory. London: Methuen. Stich, Stephen. 1983. From Folk Psychology to Cognitive Science. Cambridge, MA: MIT Press. ———. 1996. Deconstructing the Mind. Oxford: Oxford University Press. Thackray, Arnold. 1970. Atoms and Powers. Cambridge, MA: Harvard University Press. Turing, Alan. 1950. Computing Machinery and Intelligence. Mind 59: 433–460. Uebel, Thomas, with comments by Christopher Hookway. 1995. The Vienna Circle Revisited. London: Centre for the Philosophy of the Natural and Social Sciences. DP 6/95. Wellman, Kathleen. 1992. La Mettrie. Durham: Duke University Press. Yolton, John. 1983. Thinking Matter. Minneapolis: University of Minnesota Press. ———. 1984. Perceptual Acquaintance. Minneapolis: University of Minnesota Press.
TYLER BURGE REPLIES
13
Descartes and Anti-Individualism: Reply to Normore
Calvin Normore’s essay illustrates his remarkable ability to use historical expertise to provide insight into current philosophical discussion. In addition to interesting historical points, his paper raises some important issues about anti-individualism. I am gratified to have called forth some posthumous remarks from Descartes and will try to contribute to the interchange at a level worthy of the Master’s attention. I made two large mistakes in “Cartesian Error and the Objectivity of Perception” (1986b). One concerns the interpretation of Descartes. The other lies in the handling of an argument against individualism. I try to rectify these mistakes in sections I–III. Sections IV–VI discuss some theses about the role of concepts in reference and knowledge. In sections VII–VIII, I confront differences between my view and Descartes’s on these same topics. I conclude in section IX with remarks on social factors in determining the nature of thought. This is a long paper whose many topics may appeal to different readerships. I hope to provide an example of how history of philosophy and contemporary philosophy can combine to yield rewards that neither can easily produce alone. I My first mistake was the presumption that Descartes is an individualist. I realized this mistake shortly after making it and signaled my awareness of it in a subsequent paper.1 What is individualistic in Descartes is the presentation of the demon thought experiment in Meditation I. What I said about individualistic intuitions that might be drawn from that thought experiment—my criticism of moving without argument from knowledge of one’s actual thoughts to knowledge of what one’s thoughts would be in counterfactual situations—still seems to me right and to the point. This invalid transition is easy to fall for. It has seduced others. Descartes himself blurs the distinction between self-knowledge and metaphysical knowledge, in such a way as to encourage the transition. My attribution of individualism to Descartes was nevertheless badly grounded. Descartes holds that the thought experiment of Meditation I is not ultimately coherent. As Normore emphasizes, God is invoked as a necessary principle for making our intentional mental contents (ideas) conform to basic natures in the objective world. This is to give God a role prima facie analogous to the role of causal history and evolutionary design in determining contents. The general caste of Descartes’s account is in many ways congenial to anti-individualism. Whether Descartes is an anti-individualist, however, is complex and not completely clear. Descartes does not say much about the individuation of mental states. For that matter, he says little about the individuation of minds or of physical objects. I want to explain why the issue is interpretatively complex.
292
Tyler Burge
Individualism is consistent with Descartes’s claim that God ensures that our clear and distinct ideas apply to objective natures. If one held that there are no individuation conditions on what it is to have a given idea of an objective nature that make reference to anything outside the mind, one could regard God’s role as simply to ensure that our ideas always correspond to objective natures. For example, one could hold that what it is to have an idea requires that no further explanatory conditions be met. God’s necessarily making a match between the world and ideas so conceived would not entail that there are anti-individualist conditions on the individuation of thoughts. It would entail only that there is a necessary match between our ideas and their objects. Necessity is one thing. Individuation is a further thing. The appeal to God’s veracity is congenial with antiindividualism, but does not entail it. Descartes also has a causal principle: There must be as much reality, “formally,” in the cause as there is in the effect. The principle requires that the objective reality (roughly, the representational content) of an idea be caused by something whose formal reality (roughly, intrinsic reality of the object or referent) is as much or greater than the reality purportedly represented by the idea. Descartes understands degrees of reality in terms of explanatory priority (Normore 1986). The principle would require that the representational nature of an idea of a rock be explained in terms of a cause that has an ontological status at least equal to that of a rock. Thus a rock would do. In its focus on explanation, this principle is closer to bearing on individuation conditions than the invocation of God’s veracity. But the principle is too weak to entail antiindividualism. The finite thinker will have an ontological status sufficient to explain the objective reality of all the thinker’s ideas except for the idea of God (for which God must be invoked, according to Descartes’s ontological argument). So the principle is again consistent with individualistic individuation conditions for objectively and empirically referring ideas.2 It appears to me, however, that the spirit or main drift of Descartes’s reasoning is to explain the representational nature of one’s ideas in terms of the objects that those ideas represent. The appeal to God’s veracity seems to function as a guarantee that this natural direction of explanation, especially for reference to objects in the physical world, can be relied on, as long as we as thinkers avoid invoking materially false or confused ideas.3 For example, consider this passage from Meditation III: And although the reality which I am considering in my ideas is merely objective reality, I must not on that account suppose that the same reality need not exist formally in the causes of my ideas, but that it is enough for it to be present in them objectively. For just as the objective mode of being belongs to ideas by their very nature, so the formal mode of being belongs to the causes of ideas— or at least the first and most important ones—by their very nature. And although one idea may perhaps originate from another, there cannot be an infinite regress here; eventually one must reach
Reply to Normore
293
a primary idea, the cause of which will be like an archetype which contains formally all the reality which is present only objectively in the idea.4
The phrase “same reality” and the causal account of the objective reality (or intentional content) of ideas seem distinctly anti-individualist in spirit. Our ideas of objective physical space, for example, seem to be explained in terms of our being causally related to spatial shapes. Since Descartes does not explicitly discuss individuation conditions of ideas, this reading requires some judgment. Still, I take the spirit of Descartes’s work to be broadly anti-individualist. In my formulation of individualism in “Cartesian Error and the Objectivity of Perception,” I indicated some awareness of these issues. I explicitly excluded appeals to God as means of preventing one from being an individualist.5 I do think that there are serious difficulties with Descartes’s invocation of theology to guarantee clear and distinct ideas in his account of veridicality and content-determination. The appeal is not only epistemically tenuous. I think that it, together with his extreme reductionist conception of physical reality, may have led him to underrate how large a role mundane, empirical, macro-objects and properties play in content determination. I will return to this point. On the other hand, no historically sensitive account of early modern rationalist views can afford to bracket the role of God. I believe that my indulging in such bracketing led to underrating anti-individualist elements not only in Descartes but also in Leibniz. I have come to think that there were fewer individualists prior to the twentieth century than I had formerly supposed. Despite the new-science criticism of Aristotelian metaphysics and of commonsense epistemology by nearly all the great early-modern figures, the enormous influence of the anti-individualism of Aristotle carries deeply into the early modern period. II Another issue needs to be explored in a discussion of whether Descartes was an antiindividualist. This is how individuation of mental states in terms of relations to objects outside the individual’s mind is to be squared with Descartes’s strong form of dualism. Anti-individualism limits the metaphysical independence of mental states and events from the nonmental. But Descartes’s view of mind as a substance whose principal essential attribute is thought is prima facie incompatible with ontological dependence of mind on the nonmental. Finite substances are independent for their existence and nature of anything else except God. Insofar as Descartes is not an individualist, Descartes must reconcile these views by holding that mental states and events that are about body are not essential to any given mind, or to being a mind.6 Mental events about body are for him modes of mind
294
Tyler Burge
(contingent properties). Such mental events are not attributes (necessary ones). (A principal attribute is an attribute that is constitutive of the entity’s nature or essence and explains other attributes.) Anti-individualism does not itself maintain that to have a mind, one must have thoughts about body. It claims that, as a matter of individuation, to have certain thoughts, including specific ideas of body, one must be in causal relations to a wider environment. It is open to Descartes to agree that having particular thoughts about body, even having innate ideas like that of extension, is metaphysically dependent on the thoughts’ nonmental objects, without conceding that mind itself is thus dependent. So although this sort of reconciliation involves, to be sure, other difficulties, it does make Cartesian dualism prima facie compatible with anti-individualism. Of course, conceding to anti-individualism that thoughts of body are dependent for their natures not only on mind but on body opens a view of Cartesian dualism that is different from its common construal. On this new construal, some particular thoughts, which Descartes would count as modes of a mind, are necessarily dependent on physical properties. So not just aspects of sensation and imagination, but even aspects of some acts of intellection are, on this position, explanatorily dependent on body. Aside from God, only mental substances (particular minds) and mental attributes derivative from the principal mental attribute (the attribute, thinking) seem to be completely independent of physical substances. It is important to see how far the exclusion of ideas of body from essential attributes of mind might extend in the Cartesian scheme, if it is to be anti-individualistic. For Descartes the basic idea of body is the idea of extension. Extension, considered in abstraction from actual matter, is supposed to be the fundamental idea in geometry. Geometry is not, according to Descartes, about corporeal extension. It does not depend epistemically on perception of the physical world. It is not committed to the existence of corporeal substance. Descartes seems, however, to think that the shapes studied in geometry are not ontologically independent of possible or actual shapes that can be instantiated in corporeal substance. Descartes’s view, like most early modern views about why geometry applies to actual physical space, depends on the claim that such shapes are essentially possible modes of matter, studied independently of whether they are actually instantiated. So the objects of geometry are not ontologically independent of corporeal substance, in Descartes’s view. An anti-individualist account of ideas and thoughts about geometric shapes would be committed to some dependency relation between the abstract shapes (the objects of geometrical reasoning) and having those ideas. If such objects of geometrical reasoning, the shapes, are not ontologically independent of (corporeal) extension, then thoughts and ideas about them are not ontologically independent either. If this line of reasoning is correct, Descartes should hold that having geometrical ideas is inessential to being a thinking mind.
Reply to Normore
295
What of arithmetical ideas? Unlike geometrical ideas, they seem not to be essentially associated with physical substances or properties. He classes them with duration and order as modes of thought, independent of what they are applied to in counting (Principles I, 55, 58). It follows that they are independent of geometrical ideas. I find this interesting. Despite his discovery of analytic geometry, and sophisticated use of algebra, Descartes joined the tradition, dominant since the Greek mathematicians, that held that geometry is more basic than arithmetic or algebra. Geometry was supposed to be the foundation of mathematics. (I shall return to this point in section VII.) But the exact sense in which it is foundational is important for our purposes. It is clear, I think, that Descartes regarded proofs in geometry as necessary epistemic bases for some beliefs about number (for example, belief in the real numbers). But it is unclear to me how far this epistemic dependence was supposed to extend. In any case, Descartes did not hold, as many mathematicians in his day did, that number is ontologically grounded in geometric proportions. As Frege noted, number, unlike geometrical lines and shape, is applicable to all things (through sortals)—so to thoughts as well as to bodies. Descartes seems to have anticipated this insight, although he would not have joined Frege in regarding number as derivable from logic. I think that Descartes was free to regard the idea of number as essential to mind. If ideas of both body and geometrical shape are inessential to mind, what would remain? One could imagine a view that no particular thought- or idea-types are necessary for mind, as long as the mind has some thoughts and ideas. On this view, all specific intentional (or in Descartes’s terminology, “objective”) content is contingent to mind and to minds. This position is at least suggested by Normore’s attribution to Descartes (p. 8) of the view that God could have made different natures, but “left our minds the same,” and “then formally the same mode of mind would have had a different objective content.” Two versions of the position can be distinguished here. One is that the content (for Descartes, objective reality) of particular thoughts could be different while those thoughts as events (their “formal reality”)—and by extension the mind—remained the same. I think that as a substantive view, this position is untenable. Our only individuative grip on the identity of particular thoughts involves their intentional or representational content—primarily the concepts with which the thoughts purportedly represent referents. The other version of the position is as follows. Although particular thoughts would be different if their contents were different, the mind is metaphysically independent for its identity not only from the identity of any particular thought events, but also from having any particular contents or idea-types. That is, there are no particular idea-types that a mind must have in order to be a mind. I will come back to the question of what the identity of mind would then consist in. But it would have to consist in something (perhaps a field of consciousness and some type of thinking or other) that is what it is independently of the
296
Tyler Burge
content of any particular mental activity or process. For all particular thoughts, and all particular types of intentional content, would be contingent modes of a mind. This second version holds that thinking is necessary to being a mind. It holds that it is not necessary that any particular kinds of thoughts be thought. And it holds that it is not necessary that a mind have any particular idea-types or concepts. This view simply denies that there are any thoughts, ideas, or concepts, that are constitutively necessary and universal to all thinking minds. This is an interesting view, not obviously false, I think. Aristotle and Kant represent a tradition of maintaining that there are fundamental, universal categories of all thought. Their lists, which include a number of traditional metaphysical notions, have not persuaded many. There remains for some an inclination to think that having a few simple logical notions associated with negation, conjunction, and implication, perhaps simple arithmetical ideas, perhaps some few further ideas, is necessary to being a thinking mind. I share such an inclination. Justifying it is, however, a difficult matter. One must confront the variety of nonstandard logics and the apparent possibility of taking different connectives as fundamental even within classical logic. As regards arithmetical notions, one must explain why thinking cannot proceed without sortals and counting, and get by with mass concepts and notions of more and less. I doubt that either of the positions just outlined is Descartes’s. I believe it likely that Descartes held instead that having available to reflection the innate ideas of thought and God is essential to being a mind. Perhaps Descartes believed that from the idea of thought or of God, other ideas, such as those of objective reality and cause, can be derived by reflection. I conjecture that knowledge associated with the cogito and the ontological argument are thought by Descartes to be necessarily available to any thinking mind. If he is an anti-individualist, however, I think he must hold that thoughts about the material world and about geometrical shapes are, though central to our actual mental histories, not essential to being a mind. The view that thoughts and knowledge of mind and of God are more basic—in the sense of necessarily more fundamental to the essence of any thinking mind—than thoughts and knowledge of body and mathematics is, of course, deeply un-Kantian. As an account of all thinking minds, I find such a view unacceptable. Thoughts about thought, let alone thoughts about God, seem absent from some minds—the minds of animals and young children—that nevertheless think about body and can engage in simple counting. On the other hand, I think that the closely related question of whether thoughts of mind are conceptually independent of thoughts of body—and the question of whether knowledge of mind is conceptually independent of knowledge of body—is more complicated than most neo-Kantian, post-Strawsonian discussion has suggested. We have been discussing Descartes’s dualist view of mind in the context of antiindividualism. I want to consider now his dualist view of particular minds in the same
Reply to Normore
297
context. Descartes claims that one cannot conceive of mental substance without its principal attribute, thinking. As we have seen, this principal attribute must be regarded as independent of any particular thought events. What are particular minds for Descartes? And what more can be said about the principal attribute of minds? If one regards the principal attribute, thinking, as a generic essence common to all minds, one must ask what individuates particular minds. I think that there is no evidence that Descartes thought of mental substances as immaterial “soul stuff.” Understanding immaterial substances on such a model is in effect to treat them as material and immaterial at the same time. Such a view misses what is special about mind, and part of what is interesting about Descartes’s dualism. Descartes’s mental substances are not, I think, best construed as distinguished by a special kind of constitution or stuff. There is reason to believe that Descartes saw the principal attribute of a mental substance as not (or not merely) generic and common to all mental substances, but as particular and concrete. Thus in a late letter to Arnauld, he writes: I tried to remove the ambiguity of the word ‘thought’ in articles 63 and 64 of the first part of the Principles. Just as extension, which constitutes the nature of body, differs greatly from the various shapes or modes of extension which it may assume; so thought, or a thinking nature, which I think contributes the essence of human mind, is far different from any particular act of thinking. It depends on the mind itself whether it produces this or that particular act of thinking, but not that it is a thinking thing; just as it depends on a flame, as an efficient cause, whether it turns to this side or that, but not that it is an extended substance. So by ‘thought’ I do not mean some universal which includes all modes of thinking, but a particular nature, which takes on those modes, just as extension is a nature that takes on all shapes.7
I will discuss two issues raised by this passage. First, the last sentence of the passage suggests that thinking is considered as a nature that is particularized in the individual mind. Yet in its particularity, it is seen as independent of any particular act of thinking. What could thinking be, so understood? It seems clear that thinking can be understood generically, as common to all particular acts of thinking. But what would thinking be, understood as both particular (or concrete) and independent of particular acts of thinking? I reject the idea that there is actual thinking independent of particular intentional or representational content. I see no reason to think Descartes was committed to such an idea. Perhaps, though, Descartes thought that a reflexive self-attribution, expressible as I think, implicitly attaches to every thought. All instances of such an attachment, somehow regarded as a continuous generative activity, might be seen to constitute the principal attribute of the mind, as a “particular nature.” I think that this is the most promising starting point for understanding Descartes’s notion. (See Principles I, 7–9.)
298
Tyler Burge
One might also take thinking, the “particular nature,” to be a reflexive consciousness. Such a view is at least loosely suggested in Principles I, 9. This view would allow the field of consciousness to take on particular images and ideas (which for much of his career Descartes thought about imagistically) as forms, on a loose analogy to matter taking on particular shapes. Perhaps the active element in Descartes’s chosen description of the relevant particular nature—thinking—could be seen as necessarily conscious. So the relevant type of consciousness might be understood in terms of the reflexive selfconsciousness, expressible as I think, mentioned above. The reflexive self-consciousness involved in the continuing I think is filled out by particular thoughts, which are themselves contingent modes of mind. I would regard such a view as hyper-intellectualized, if it were applied to all thinking minds. Animals and children think, but lack a concept of thinking. They do not think about thinking. But there might be variants worth exploring. One could confine the view to critical reasoners. One could avoid requiring—as Descartes does—that such thinkers be thinking at all times (and be thinking I think at all times!), if they are to continue to exist. Perhaps some less intellectual, less fully self-conscious, form of conscious awareness could be attributed to lesser thinking minds—to minds that are not capable of critical reasoning. By reflecting on the role of the different sorts of consciousness, perhaps one could learn something about different levels of mind. Here I think that there are interesting issues about the relation between intellectual reflexive self-consciousness and ordinary phenomenal consciousness. There are also interesting issues about the relation between reflexive consciousness in the activity of thinking (through implicit attachment of I think to thoughts) and consciousness of oneself as intellectual agent. This brings me to the second issue from the quoted passage that I wish to discuss. There is an obscurity in Descartes over the relation between the mental agent and its agency— between the mind and its thinking. There is no obscurity about Descartes’s view of the relation between mental substance and its particular contingent thoughts. Descartes is emphatic that the mind and its modes, its thinking events (whether active or passive), are to be distinguished.8 But the matter is more problematic regarding the relation between mental substance and its principal attribute, thinking—especially inasmuch as thinking is considered a “particular nature” in the individual thinker. Neither a field of consciousness, nor reflexive consciousness in the form I think, nor whatever else thinking itself might be—is the mental agent. The agent is a being that engages in such reflexive acts, that has the relevant self-consciousness. To identify the agent with any of these would be to make a category mistake. An ontological view of mental substance that takes the agent simply to be entirely exhausted by thinking seems to me incoherent. Thinking must be the activity of an agent of thinking. The primary essence, or principle attribute, must allow for or include an agent of thinking.
Reply to Normore
299
It is clear that Descartes is not committed to mental substances as bare particulars, or as having some further constitution, or stuff, beyond their thinking natures. Thinking is the principal essential attribute of the mind. No other essential property is, on his view, needed for the mind to be a substance. Neither the substance nor the attribute could exist or be understood without the other, and substance is known only through its principal attribute. But Descartes speaks of the mind as “producing” thoughts. He often seems to respect the distinction that I am emphasizing, by writing of the mind as a thinking thing. Moreover, as I have noted, he is very firm that particular thoughts, modes (contingent properties), are not the same as the thinking thing, the thinker or mind that does the thinking in particular cases. On the other hand, he sometimes speaks of thinking, as a principal attribute (essential property) and particular nature, as if it just is the substance. (See Principles I, 63.) Some scholars have interpreted these passages in a way that risks attributing the category mistake that I have warned about.9 In the passage quoted above, Descartes draws an analogy between thinking and a flame, which might be taken to suggest that thinking, considered as principal attribute, is a process that does not inhere in some further thing, a thinker. The other analogy, in the passage quoted, between thinking, as a principal attribute of the mind that can take on particular thoughts as modes, and material extension, which as a “particular nature” can take on different shapes, raises similar worries. This analogy seems to leave no room for a further distinction between thinking and the mind, as agent—the thing that thinks. For in the case of the corporeal substance, the principal attribute, material extension, takes on further shapes as contingent modes; but there is no need to distinguish between extension and anything further. In my view, although reflexive consciousness, as principal attribute of the mind, might take on particular thoughts as contingent modes, reflexive consciousness (I think) must be the consciousness of an agent that has such consciousness. Descartes’s official account of the distinction between a substance and its principal attribute in the Principles I, 60, 62, does not address the issue but may help us in illuminating it. In those passages, he discusses three relevant distinctions. First, he cites a real distinction, which holds between substances, which God could separate. Second, there is a distinction regarding mode, which holds between substances and their contingent properties or modes. Third, he mentions a conceptual distinction (or distinction in reason), which holds between a substance and an attribute of that substance without which the substance is unintelligible. Descartes says that the latter distinction is recognized in our inability to perceive clearly the idea of the substance if we exclude from it the attribute in question. This latter distinction holds between the thinking thing and thinking, its principal attribute construed presumably as a particular nature in the sense of the letter quoted above. Mental substance and its principle attribute are merely conceptually distinct, or distinct in reason, in this technical sense.
300
Tyler Burge
It is natural, and perhaps correct, to read ‘merely conceptually distinct’ (or ‘distinct in reason’) in a way that would require that thinking substance and its principal attribute are exactly the same at the ontological level, only thought about in different ways. This view seems, at least at first blush, to fall into the mistake of identifying agent and act, when it is applied to thinking substance. I think, however, that the view does not, or need not, fall into such an error. The view that the thinking agent can be understood essentially and purely in terms of its thinking seems to me interesting and characteristically Cartesian. Such a view might resist the move of insisting on conceptual grounds that the agent have a further constitution, one more basic than its thinking.10 But there remains the need to explicate some distinction between thinking agent and the activity of thinking. If there is no distinction between thinker and activity indeed some ontological distinction, then Descartes would be in my view committed to a category mistake. The problem of making room for a distinction between an act and its agent (hence, in Descartes’s scheme, between thinker and thinking) has no analogue in the relation between bodily substance and its principal attribute, material extension. I think that Descartes is in a position to solve this problem, even assuming that he identifies substance and principal attribute in some ontological sense. The difference between the principal attributes, material extension and thinking, itself seems to provide a basis for drawing the categorial distinction that I am after. If the principal attribute of a thinker, as a particular nature, is continuous reflexive self-consciousness, then the attribute itself, as expressed in I think, requires a distinction between thinker (what is indicated by I ) and attribute (what is indicated by think), between agent and act. The principal attribute would allow for an agent of thinking, because in having an essentially reflexive, “I think” structure, thinking itself contains essentially an agent-act ontological structure. If this conjecture is right, then Descartes’s account of thinking as a particular essential nature yields the desired asymmetry between the principal attributes of material and thinking substances. Before leaving this point, I want to say a bit more about Descartes’s conceptual distinction, or distinction in reason, mentioned above. This is a distinction that holds between any substance and its principal attribute. In a letter of about 1645, commenting on the distinction, Descartes states that conceptual distinctions must have a “foundation in reality.”11 Descartes contrasts conceptual distinctions, or distinctions in reason, with a distinction made purely by the mind—a rationis ratiocinantes. Descartes says that he does not recognize a rationis ratiocinantes. Traditionally, a purported example of this latter sort of distinction would be a distinction between definiens and definiendum in a real definition— between man and rational animal, for example. Descartes apparently would say that reason (as opposed to the language) draws no distinction in this case. Descartes denies that the
Reply to Normore
301
distinction between substance and principal attribute is a distinction made purely by reason, and claims that the distinction has a foundation in reality. Descartes seems to associate distinctions in reason with the distinction between essence and existence. In any given case, the essence of a corporeal body and the existence of the same body are for Descartes the same. Although the thought of Peter is different from the thought of humanity, “in Peter himself, being a man is nothing other than being Peter.” Similarly, in a corporeal substance, being an instance of material extension is nothing other than being the corporeal substance. The essence can, however, be thought about independently of the existence. And this difference in thought has a foundation in “objective reality” inasmuch as the difference in thought contents is not created by reason (it is not a rationis ratiocinantes), but is grounded in objective ways of thinking. Descartes’s nominalism about natures prevents him from regarding this difference in “objective reality” as being grounded in an ontology of universal natures. I am not sure that with his thin nominalistic resources he can give a satisfactory account of the objectivity (or “foundation in reality”) of the distinction between substance and attribute—an account of why the distinction in thought is objective. In any case, I think that antiindividualism will require him to say more than he does. But as far as I can see, he has the resources to avoid the particular incoherence of identifying agent and agency that I have been worrying about. There remains, of course, the notorious question—pressed by Kant in the Paralogisms section of The Critique of Pure Reason—of how, in the context of his substance dualism, Descartes individuates individual minds. Suppose that we do allow that there is an agent of thought, and (for the sake of argument) that this is the fundamental mind-substance, or at any rate a factor in mind-substance. What is it about this agent that individualizes it? Not the particular thoughts it thinks. They could be different while the agent remains the same. Not the general attribute of thought. That is common to different thinking agents. There is no evidence that Descartes appeals to “mental stuff.” As noted, such an appeal would lose the insight that mind is deeply different from body. It would miss the role of intentionality and point of view in making thinking agents what they are. There is the neoKantian route of demanding a body as a necessary condition for individuating a mind. The arguments for that view are tantalizing, but I do not think them decisive. At any rate, given his dualism, Descartes cannot appeal to the brain or to “external” physical objects to help individuate mental substances. Is there any way to think coherently about the individuation problem purely from the point of view of Descartes’s version of dualism? Perhaps one should take seriously Descartes’s apparent tack of simply regarding individuation of particular minds as primitive: We individuate a mind by conceiving it as an agent of particular mental acts. The same mind could have produced other thoughts instead. We count a mind the same by
302
Tyler Burge
reference to some type of continuity of a changing point of view. (I would insist that the agent has not only conscious active thinking, but powers, faculties, concepts, and other mental dispositions that are present even when mental activity is not.) I believe that Descartes may be on to something important in regarding thinkers as consisting not in some special sort of stuff, but in particular instances of the special type of agency, power, consciousness, and point of view involved in thinking. What is interesting and challenging here is to explain why immaterial constitution and bodily constitution are not basic—and why the aforesaid mentalistic features are the basic properties of minds or thinkers. A central challenge for a serious dualism should be to explain why the fundamental sortals can be activity and power sortals rather than constitution-sortals. Another Kantian challenge is to show that a continuous thinker can be made sense of using only mentalistic concepts. Whether or not these challenges can be met, I believe that reflection on these issues so far has not exhausted all possibility of progress. Descartes’s dualism seems to me more interesting than traditional or current caricatures of it allow. I have no interest in reviving a substance dualism, where ‘substance’ is taken in the oldfashioned sense—requiring complete ontological independence from anything else in the same ontologically basic category (e.g. the category substance or the category entity). I am not sure that anything is a substance in the old-fashioned sense. It is not obvious to me, however, that it is mistaken to suppose that mental agents and their mental powers, acts, and states are in no literal sense physical. For the present, I am impressed with anti-individualistic elements in Descartes’s account of mind. I think that anti-individualism is prima facie compatible with some form of dualism. These are profound historical and substantive issues that need more development. III My second mistake in “Cartesian Error and the Objectivity of Perception” lay in the presentation of the argument for perceptual anti-individualism. I set out three premises. I then used them to argue for a case in which a creature’s internal physical states and dispositions are compatible with being in different perceptual states. I still believe that this argument is sound. However, the case fixes on a rather special situation in which all a perceiver’s dispositions to action are equally well fitted to either of two perceptual contents. Moreover, my giving the argument suggested that such antisupervenience arguments are necessary to establish anti-individualism. They can support anti-individualism in a particularly vivid way. But they are not necessary to establishing anti-individualism. My argument also misleadingly suggests that failure of local supervenience of intentional states on the individual’s physical states is to be identified with anti-individualism.12
Reply to Normore
303
The third premise of the argument is initially stated: “some perceptual types that specify objective types of objects, properties, and relations as such do so partly because of relations that hold between the perceiver—or at least members of the perceiver’s species— and instances of those objective types” (Burge 1986b, pp. 127–128). This premise is not as sharply stated as it should be. “Because” could be interpreted causally or individuatively (which includes causation but goes further). As I go on to explain the premise, it is clear that I intend individuation. For example, I wrote, “The third premise states that some of a perceiver’s perceptual types take on their representational characters partly because their instances interact in certain ways with the objective entities that are represented. . . . It makes no sense to attribute systematic perceptual error to a being whose perceptual representations can be explained as the results of regular interaction with the physical environment and whose discriminative activity is reasonably well adapted to that environment” (ibid., pp. 130, 131). Understood this way, the third premise already entails antiindividualism!13 Anti-individualism about having perceptual representations is fundamentally supported by reflection on how intentional content is individuated and by the seeming unintelligibility of individualistically construed, ordinary perceptual representations. The perceptual state-types about objective, particularly physical entities, and hence what intentional content they have, is individuated by reference to those entities that causally interact with the perceptual system and explain successful perceptions and successful practical or motor responses to the discriminated objects. The objects give content to the representations both by forming them and by embodying what counts as the endpoint of a successful perception. Perceptual discrimination is normally associated with practical activity to which the discriminated objects are relevant. To attempt to account for perceptual intentional content while abjuring causal connection to the represented objects seems to me empty and ultimately unintelligible. What the three premises do, independently of their establishing, for a special case, failure of supervenience of perceptual states on local physical states, is to provide a framework for understanding the anti-individualistic nature of ordinary perceptual representation. The first premise indicates that perceptual reference is to objective entities. These are entities whose nature is independent of any particular person’s actions, dispositions, or mental phenomena. The second indicates that perceptual representations specify objective entities in definite ways. Percepts represent edges as edges, boundaries as boundaries, corners as corners, cones as cones, and so on. The third indicates that these representational or intentional specifications are individuated partly through interaction with the objective entities that are represented.
304
Tyler Burge
IV Normore discusses a version of anti-individualism that invokes “bare concepts.” He sees this position in Ockham and thinks it suggested by ways of pursuing ideas in Kripke. Normore claims that Descartes and I would agree in rejecting the “bare concepts” view. He seems to me broadly right about this. There are, however, four theses to which he attaches the “bare concepts” label. I would like to distinguish these, and discuss them separately. I discuss the first two of these theses in this section, and the remaining two in sections V and VI. There is, first, the thesis that “the mental quality or act that is the concept could have been produced by other things . . . and had it been so produced it would be a different concept” (p. 5). I believe that this view is incoherent. Something cannot be a concept and be capable of being another concept. Mental entities might express concepts, or be the vehicles of concepts, and might remain the same while being associated with another concept. I doubt that concepts must in general have vehicles. But I can leave that issue open. The thesis as stated, however, seems to me entirely unacceptable. Concepts are individuated by their intentional contents. In fact, they are “their” intentional contents. In my view, it makes no sense to take something that could remain the same while its content varied to be the concept. That is simply to change the subject. The second “bare concept” thesis is that mental representation or reference is never representation-as. This thesis is better understood as a no-concepts, no-representation view. According to this view, one thinks and perceives objects or properties, but the thinking does not specify them or represent them in any particular way. I regard this view as untenable as well. All thought and perception is perspectival. Each and every reference in thought and perception is perspectival. We represent and perceive objects and properties only from a point of view, by way of abilities that provide partial, incomplete, and sometimes erroneous perspectives on them. Concepts type such perspectives and abilities. One can have different perceptual representations from different angles of perception on the same property, even representing it as the same property. This is the essence of perceptual constancy—the ability to perceive something as the same object or property even though the perceptual mode of presentation, the perspective on the object or property, is different. One can also represent the same property in different sense modalities or conceptual modalities. Representation, in both thought and perception, is typed not only in terms of the referent, but in such a way as to reflect the relevant referential mental abilities that constitute the thinker’s perspective on the objects and properties that are represented. These abilities are subject to normative standards that attach to or are embodied in concepts and perceptions.
Reply to Normore
305
It would be absurd to think that finite beings can perceive or think about ordinary objects and properties neat—incorporate them into perception or thought without doing so by representing them in some way. No mental ability corresponds to such a view of thought or perception. We lack cognitive power to get on to (refer to) ordinary objects or properties in no way at all, or to incorporate them whole, apart from any representational means that constitutes a partial, usually fallible, perspective on them.14 Mental representations type cognitive or perceptual abilities. So mental representations must type the partial, perspectival abilities that we in fact have. The abilities so typed enter into psychological explanations. The mental representations enter into accounts of the veridicality of the states and the representational perspective and epistemology of the cognitive process. Although the basic grounds for my views on this second thesis derive from apriori reflection on human abilities and on the nature of human epistemology, there are empirical grounds as well. Psychological explanation takes the processing of representations that type mental abilities as fundamental. The processing of representations by perceptual subsystems cannot be separated in empirical theory from the end-product perceptual representations attributed to the whole perceiving animal or person, as well as to the perceptual subsystem. Many conceptual abilities and representations are parasitic on these perceptual ones. I think it untenable to hold that any mental reference or application is neat or can be fully typed apart from some representation. All reference in thought is typed in terms of some representation—some mode-of-presentation that in turn types some perspectival referential act or ability of the individual—where the ability itself is individuated in ways that are answerable to normative standards of various sorts. The point applies to general perceptual representations and to concepts, both of which mark representational abilities. The point also applies to demonstrative singular representations that are individuated in terms of—and mark—token, demonstrative singular applications in thought and perception. Representations fall into two main categories. The first category includes perceptual and conceptual representational types that mark general representational or psychological abilities, abilities that are in principle repeatable. The abilities and the representations that mark them are not simply abstractions from a single token application or event or any specific group of act or event tokens. Both perceptual representations of property types and of relation types and nearly all conceptual representations fall into this category. For example, the ability to apply the predicative concept cat does not depend on any specific encounters with cats. Any perceptual encounters with any cats, together with appropriate background information, might suffice to enable one to have the concept. Similarly, any encounters with someone who told one about cats might suffice. Let us call such representations ability-general, since they type general psychological abilities. General
306
Tyler Burge
psychological abilities are psychological abilities that are freely repeatable: They are not individuated by reference to any specific token act(s) or event(s). Most concepts and all perceptual representations in this first category are general in a further sense. They are capable, according to their form, of applying to various satisfiers of the representational type. Let us call such representations formally general. (Unless I indicate otherwise, I shall use ‘general’, unqualified, in this latter sense.) Thus a predicate concept is formally general. The predicate concept restaurant is open to applying to numerous restaurants (all restaurants) according to its form—even if there were in fact only one restaurant in all the world. By contrast a concept that is fully expressed by a context-free singular term would not be formally general. Some representations that are ability-general are not formally general. These are individual concepts. They apply according to their form to single objects, if to anything. Perhaps the concepts God and three (understood as singular rather than adjectival) are examples. But like most formally general concepts, and like all ability-general perceptual representations (which are all formally general), they type freely repeatable psychological abilities. Having the concept God or three is an ability that is not individuated by reference to any specific token acts or events, but rather a cluster of inferential, applicational, and predicational abilities, commonly associated with relations—causal or constituting relations—to the subject matter.15 The second main category of representation marks a singular application. Such singular applications are of two types. They may be acts of applications in thought— applications of concepts. Or they may be applications in perception—applications of ability-general (and formally general) perceptual representations. These context-dependent singular sorts of representation mark particular acts of context-dependent application to particulars. The particulars may be property- or relation-instances, as well as individuals. The representation can be an abstraction, in that it may remain, in memory, after the moment of the activity that it marks is past. It can even be maintained through interlocution. It is nevertheless individuated ultimately in terms of a specific token act or specific acts, not general abilities. Such token singular representations in thought are expressed by particular tokens of demonstratives like ‘that’, and by pronouns taking such tokens as antecedents. I believe that there are analogous context-dependent singular elements (individuated in terms of token occurrences, if not acts) in perceptual representation.16 I think it obviously incoherent to hold that any representational perspective by a finite being can do without formally general representations. All thought must make use of predication and predicational grouping. Predication is necessarily a representational type that is formally general; and it necessarily presupposes and must be associated with abilitygeneral predication. All perception must group or type perceptual objects under general perceptual rubrics, at one or another level of abstraction. It must involve perspectival rep-
Reply to Normore
307
resentation of aspects of particulars. Such representation constitutes an in-principle freely repeatable psychological ability. In addition to the thesis that all representation is perspectival and must include some general representations, I think a stronger thesis is true as well. Very roughly, the stronger thesis is that each representational context-dependent singular element—representations of the second, “token” singular kind—in autonomous thought and perception must be associated with a nonschematic representation that is both formally general and abilitygeneral. It must be associated with some general typing of the entities represented, as suches. More generally, the thesis is that representations in all positions—perceptual, referring, predicational, functional, and quantificational positions—in every autonomous representation must be or be associated with a formally general and ability-general representation. Thus, every singular representational element in every representation (simple or complex) has such association. ‘Autonomous’ is meant to rule out thought that depends essentially on communication. I shall discuss this provision briefly in section V. I take the relevant guiding formally general and ability-general representations to be nonschematic: They are not themselves purely demonstrative or indexical context free representations (like this or that) that are in need of a singular element to purport to have a definite referent or application. Such schematic representations do count as formally general and ability-general. But they cannot guide a singular application by themselves.17 The notion of association here is a technical one. A development of the notion, and an exact statement and defense of the thesis that I am broaching, will have to be postponed to another occasion. The main idea, however, is that any autonomously applied contextdependent singular representation must be tied—in perception or imagination, or through inference, predicative determination, memory, or anaphora—to a representation that is both formally general and ability-general, that is somewhere in the individual’s repertoire. According to the representational content of the individual’s overall representational perspective, the referent of the singular representation satisfies the formally general and ability-general representation. General representations include perceptual, imagistic, and conceptual representations of types.18 The conjecture that I have just broached does not maintain that the formally general and ability-general representation that must be associated with a context-dependent singular applicational representation must veridically apply to the referent of the singular representation. It does not maintain that the referent (if any) of the context-dependent singular representation must as a matter of fact satisfy the associated general representation. I will discuss conditions on reference in a moment. The point here is conceived as a condition on thought and perceptual representation purportedly about objects. It is not conceived as a condition that must help determine what the actual referent is. To purport to represent
308
Tyler Burge
singularly and autonomously, to purport to refer to a definite thing in a singular way, the singular representation must be associated with a general representation. The idea of the conjecture is that autonomous applications of singular representational abilities must be guided by some general representational ability, some ability to group or categorize objectinstances as instances of a type or as having some property, or as being in some relation. This conjecture is less restrictive than most claims in this area. Whether it can fare better than they remains to be seen.19 V The third thesis that Normore associates with the “bare concepts” view is that “the reference of a thinker’s thoughts . . . is wholly fixed by relations, usually causal relations, to a world outside the thinker” (p. 4). I shall assume singular reference to be primarily at issue here. But I believe that there is a parallel issue about the counterpart of singular reference for general representations. This thesis is to be distinguished from the previous one in that it concerns the reference relation, not merely the nature of mental representation. There is a straightforward reason why, on a certain construal, this thesis is untenable. The reference of a thought even to empirical objects outside the thinker depends not only on causal relations to such objects. It depends on the application of a referential apparatus, which depends on a network of intellectual or perceptual abilities in the individual. All reference in thought or perception must occur within some logical and conceptual structure or some perceptual structure, which types a network of representational abilities. This is fundamental to psychological explanation as well as to a reasonable epistemology. Reference in thought or perception necessarily depends on structural elements in the cognitive subject. To be an element in a thought that refers, the element must fit into the logical or grammatical form of the thought. It must bear relations to other elements in thought. These elements type representational abilities—abilities to predicate, abilities to group the referent as an instance of some kind or type, abilities to make inferences. No reference in thought could be established or fixed if it were not embedded in a network of cognitive or perceptual relations among intentional elements (representations) within the thinking individual. A perceptual representation of an object must be related to an ability to type the object referred to in perception. It must also be related to an ability of the perceptual system to fit the object into a wider perceptual scene or relate it to an array of other perceptual parameters. Thus reference to external objects even in perception must be fixed by a combination of causal relations with a network of representational abilities. Causal relations alone cannot fix any reference.20
Reply to Normore
309
The formulation of the third thesis might be understood in such a way as to avoid the preceding objections. It can be understood to hold that no formally general representational element need be true of the referent of a singular element in the world external to the individual thinker. For given the perceptual and conceptual framework within which the singular element is used, causal relations suffice to fix the referent. So construed, the thesis denies a role in fixing reference to any general mode of presentation, whether perceptual or conceptual, which guides the reference. Of course, the thesis is obviously false if singular elements are understood to include context-free definite descriptions, or their analogues in thought. I shall not understand ‘singular elements’ in this way. I understand the phrase to apply to demonstrative-like token applications in thought and perception. The thesis is, then: Given the perceptual and conceptual frameworks within which a formally singular context-bound (demonstrative-like applicational) element is employed in the perceptual or conceptual system, no formally general representation in either of these systems (other than trivial mathematical or logical representations) need be veridically applied to, or presupposed to be true of, the referent of the singular element. Given these frameworks and a singular application within them, causal relations alone fix the referent. I believe that this thesis, too, is unacceptable. Although it has other liabilities, I shall consider the thesis only as applied to cases of empirical reference to physical objects. The thesis can be made vivid by fixing on cases where the singular element succeeds in referring to an object, even though salient general representational elements that accompany it fail to be true of the object.21 One can refer to the man in the corner pretending to drink a martini even if one thinks of him as the woman along the wall drinking a coke. One can refer to Socrates even though one calls him ‘Hebrides’ and is mistaken about what he is known for. One can see a white toy behind one—where one’s sight is guided by a mirror or prism one is unaware of—even if one perceives it as a brown rat in front of one. Such cases show how reference in thought or perception can succeed even though the salient general representations that accompany the singular demonstrative-like application do not apply to the referent. I take for granted that for many singular representations, no set of general representations in the repertoire of the individual need or can uniquely fix the referent by being true of it.22 This point applies to many demonstrative, indexical, and proper names, and many incomplete definite descriptions in thought. It also applies to many perceptual representations, which pick out individuals or spatial locations even though an indiscernible look-alike might be, or even is, somewhere else in the world. The question I want to discuss is whether there are any singular representations in thought or perception that can succeed in referring by ordinary empirical means to physical entities, even though no formally general representation conceptually attributed by the
310
Tyler Burge
individual or perceptually attributed by the individual’s perceptual system applies veridically to the referent. This is a complex and delicate question. It is arguable that through interlocution an individual might refer, even though literally no general representation that the individual is disposed to apply to the object is true of it. Two adults may agree to call some particular atom or some warp in space-time ‘Sam’ and tell a child some falsehoods about Sam. The child accepts the stories, relies on them, and builds up a fantasy life about Sam. We can imagine that the child has no metaconcepts. So the child cannot think of Sam as the object the adults were talking about, or as the thing referred to by the name or as the referent of its thought. The child might garble the name and think of the object as Slam, wondering what kind of thing Slam is. So the child lacks the ability to count the object one of the Sams. If we suppose that the child has the superordinate concept of a spatial object, we may also suppose the named object to be a number. No general representation true of the referent is available to the child. Some will deny that the child is thinking about anything. I do not see a sound basis for such denial.23 In such cases, I doubt that any true attribution available to the recipient is required for reference to succeed. I think that we can imagine a child that has no relevant concepts beyond those built on perceptual representations. It is important that in these cases interlocution is in play. I believe that names and other demonstrative-governed elements in thought pick up reference that is grounded in others’ representations. Here the relation is quasi-anaphoric. I want to lay interlocution aside. So I shall discuss the problem by centering on singular reference in perception, or singular reference in thought that is supported purely by perceptual concepts.24 If one is to perceive or have a perceptual thought about a particular, must one’s perceptual system apply some veridical perceptual representational type to something perceived or something thought about demonstratively through perception? In considering this question, one must bear in mind that perceptual systems attribute perceptual representations at different levels of abstraction. Abstraction is understood here in terms of inclusion of more perceived subtypes. Thus a perceptual representation of something as an edge is more abstract than a perceptual representation of something as an edge-type of such and such a length and width, in such and such an orientation.25 I believe that the perceptual system can mistake nearly all nongeneric characteristics of an object, and the individual can still succeed in perceiving (and thinking about) the object. The visual system can be mistaken about the color, texture, shape, surface properties, spatial location, size, distance, motion, and sortal type—all in a given instance—and yet still enable the individual to see an object that is appropriately causally related to the perceptual representation. Perceptual reference can succeed even if the perceptual system is fooled into representing an object as being in front, when in fact the object causing the representation—and being perceived—is behind. One perceives the object even though
Reply to Normore
311
one misperceives its location and most of its features.26 For example, one can see an object through an unknown distorting prism and have so limited a view that one gets its shape and sortal type wrong. One can see something as an object with a definite surface, whereas it is in fact a coherently formed and trackable wisp of fog or a strikingly salient beam of light, perhaps a hologram. There are, however, limits on how mistaken the perceptual system can be and still succeed in having a perceptual referent. I think that to be visually perceived, an object must produce visual representations in such a way that the object’s properties form a coherent, trackable group that bear some systematic correspondence to the representations, even if the representations are mistaken. Suppose that light came from an odd angle and, because of its reflections off particles in the air, caused a representation of an object as straight ahead. Suppose that neither the light nor the particles form any coherent, trackable shape analogous to the apparently trackable shape of the apparent objects. The light is not a flash with a shape that is a deformation of the shape represented, nor is it a hologram. Then I think neither the light nor the dispersed particles are perceptual referents, with misperceived features. They are not objects of perception at all. There is only illusion. Perception fails not merely because the light is not where the perceptual system represents an object as being. As noted, we can perceive things while mislocating them. Perception does not fail because we cannot see light or dust and mistake them for more mundane objects. We can. Moreover, we can misperceive the shape and surface disposition of things we see. The problem is that, by hypothesis, the light and particles lack any coherence that is like the bodies and surfaces normally represented and tracked by the visual system. The visual system’s binding together various representations into a representation of a single entity does not correspond to any such system of properties in the environmental cause (or causes) of the complex representation. We would be seeing the light or particles if they formed a bounded shape of some sort, and some deformation of that shape were represented. The seen entity must have something like the boundedness of a trackable object. This does not mean that the “object” must be internally spatially connected. We see constellations, flocks of birds, and so on. In some of these cases, we are seeing several bounded objects at once, and successful reference depends on seeing a sufficient number of the component objects as bounded. In other cases, the boundedness of the whole is all that matters.27 I think that this sort of point suggests a requirement that some formally general perceptual representations must be veridical if singular perceptual reference is to occur. I will not try to specify such a requirement in detail. But I conjecture that if one is to see a physical object, the perceptual system has to get right certain spatial connectedness relations
312
Tyler Burge
among the parts of an object seen. It might also have to get right some of the relations between the object and other objects, or a background, in the presented scene. I also conjecture that similar limitations occur for other types of perceptual representations, besides object-representations. But developing these points is a matter for other occasions. I believe that to perceive a physical object in the environment, there is a more abstract point that the visual system must get right. It is a fundamental feature of the visual system, for example, that it represents its objects as in particular spatial locations outside itself. This feature is fundamental to the role of the perceptual system in generating motor activity that is geared to finding, fleeing, or otherwise coping with perceived objects. I have noted that we can perceive objects even though we are quite mistaken about where they are. But if the perceptual system represents an object as being in a specific location and the representation is caused by some internal event or malfunction, then the perceptual representation fails to have a perceptual referent. Nothing is perceived. It fails because the causes of the perceptual representation are not spatially located in the environment of the perceptual system. The system’s commitment to the object’s being spatially located outside the perceptual system must be veridical if normal visual reference is to succeed. Perceptual reference cannot succeed if its cause has no location in the environment—or at least on the surface of the perceptual system—at all. Not understanding this point lies at the root of the hoary mistake that we see sense data. It does not follow from this point that there must be a veridical perceptual representation that accompanies successful perceptual reference. In particular, it seems implausible to presume that the perceptual system has as abstract a representation as spatially located in the external environment. Although such general conceptual representations make a partition for a mature thinker, they do not provide any usable distinction for a perceptual system. So spatially located in the external environment is not a perceptual category, a representation available to the perceptual system. Subhuman primates and young children probably lack any such concept, even though they incorporate their perceptual representations into a belief system. So the point does not directly aid the third thesis. Still, there is information in visual representations from which a general concept of spatial location can be extracted. As noted, applications of the perceptual demonstrative representation is at angle such and such to the right of straight on might be mistaken, and one might still see the object through a distorting prism. The concept of spatial location in the environment can be conceptualized or inferred from particular representations by an individual with the requisite conceptual maturity. I conjecture that such a concept must be true of a referent of a singular element in visual perception (with provision for a special case to be noted shortly), even if it is not the case that an individual perceiver must have the concept. Such formally general representations are reasonably regarded as presupposed by the visual system.
Reply to Normore
313
We can imagine a sophisticated adult in highly disorienting circumstances thinking as follows: “I do not care whether that is spatially located in the environment in the usual way; it may be a reflection on the retina; or it may be an internal, even hallucinational, image; I want to know what that is”—where the use of ‘that’ in the thought is accompanied by the perceptual presentation. Here the individual has in effect canceled the commitment to an environmental spatial location presupposed by the perceptual system. The perceptual demonstrative ‘that’ is understood to apply, in a default manner, to a physical object if the perceptual system is successful in perceiving such an object, and to a retinal or internal image if it is not successful. It does seem possible to refer in this way. Is the reference in thought unaccompanied by any general representation that is true of the referent? I believe that to carry out such a reference, involving qualification of the normal commitment to an environmental location for the object of reference, the thinker must have concepts of appearance, perceptual representation, retinal image, and so on—in addition to physical object concepts. I think that an individual who knew nothing of reflections on retinas or perceptual representational images could not make reference to them. Thus the thinker has and applies a disjunctive concept—either physical object or retinal image or perceptual representation— that is true of the referent. I believe that such a thinker probably must also have the concept of cause or the concept of explanation. The sophisticated adult understands the perceptual demonstrative that to apply, in a default manner, to a physical object if the perceptual system is successful in perceiving such an object, and to a retinal or (presumably as a third back-up choice) internal image if it is not successful. The individual could refer to a perceptual representation even in the case of veridical perception, if he or she intended to. Children and apes need not have the ability to think of their perceptions as being caused by objects. Their perceptual systems do not represent causal relations between objects and perceptual representations.28 However, an individual who thinks the sophisticated thought just indicated probably must have such an ability. The individual who can take reflections on the retina and internal images as possible referents of his perceptually guided demonstrative must have some rudimentary concept of there being other possible causes or explanations of his perceptual representations than the ones that established the intentional content and reference of ordinary successful perceptions. I think that such an individual presumes a metaview of the referent as a cause or explanatory factor of the perceptual representation. So the individual is disposed to attribute a general relational representation— that of the cause or explanatory basis of the perceptual representation—that is in fact true of the referent.29 Issues regarding the limits on perceptual-conceptual referential error are a rich topic for further inquiry.
314
Tyler Burge
VI I want now to consider the fourth thesis that Normore associates with the “bare concepts” view. This is the thesis that all of our basic categorematic concepts are such that “there is nothing that we can see to be true about them apriori” (p. 3). I take it that this thesis is not meant to exclude purely logical apriori truths, such as that if Socrates exists then Socrates is self-identical or that orcas are orcas. I will ignore the thesis insofar as it is meant to apply to mathematical or philosophical concepts. I am interested in the thesis insofar as it is meant to apply to empirical concepts. One way of understanding the fourth thesis to see it as postulating empirical concepts as being introduced to apply to whatever best empirically explains this introduction, where it is a radically open empirical question how to explain the introduction.30 But one could regard the relevant empirical concepts as nonindexical and still maintain the thesis under discussion. I believe that this thesis is mistaken. In the first place, there are apriori knowable truths associated with judgments derived from basic limitative principles governing the reference of perceptual representations in perceptual systems. These are nearly entailed by what I have already said about the third thesis. Take, for example, the judgment that if that exists (where the demonstrative that relies basically on an ordinary visual representation), then that is spatially located. For any individual who has a concept spatial location, such a truth is apriori knowable. By hypothesis the individual must have the relevant concepts and the visual experience to make the judgment. But sense experience does not seem to figure essentially in the warrant for the judgment. The judgment is warranted by reflection on the referential norms governing the visual system. (There are analogous metarepresentational truths that are also apriori.) Are there further apriori truths of categorization, beyond those associated with principles governing perceptual reference? Are there apriori truths about the referent of an empirical concept that are not mere instances of mathematical or logical principles and that do not derive from principles governing the referential limits of a perceptual system? Earlier I maintained that all reference depends on there being inferential connections between the judgment in which the inference is made and judgments involving other concepts. I shall take this as granted by all sides. A negative answer to the question just posed maintains that none of these inferences is a good nonempirical inference, unless it is an inference from instances of principles of logic or mathematics or from principles governing perceptual systems. I doubt this thesis as well. But I find it interesting and challenging. Putnam’s claim that cats are animals is not apriori points in the direction of this thesis.31
Reply to Normore
315
I think it plausible that most empirical concepts are associated with superordinate concepts that provide conditions for application of the subordinate concepts and for singular reference accompanied by or guided by the subordinate concept.32 I think that one can be defeasibly apriori warranted in believing and even knowing certain general limitative principles governing the reference of concepts. For example, I think that we can know apriori that water is, if anything, physical and occupies space; that if something is yellow, it is colored; that a cat is, if anything, something with a physical body that has causal properties.33 It seems to me that many of our empirical concepts have apriori connections of this categorizational sort to other concepts. Despite my rejection of the fourth thesis, I believe that the kinds of connections that are apriori are extremely abstract. The point that we know only empirically that a kind like gold or water is a natural kind and has a unifying empirical principle is made very explicitly by Kant. Perhaps Descartes was on to the point as well. However, our conception of the taxonomic arrangement of genus and species is vastly more fluid and empirically sensitive than the conceptions prevalent in the early modern period. The apriori connections of the classificational sort that I have discussed here are mostly between relevant concepts and concepts for extremely generic features or relations in the world. I assume that Normore is broadly right that Descartes is unsympathetic with the four theses just criticized. I turn to differences that Normore sees between my views and Descartes’s. VII Descartes’s conception of the ways we might fall into error is a perpetual challenge to attempts to answer scepticism.34 He is sensitive to the fact that some of our representations are composites of other representations (griffins, satyrs). He challenges us to distinguish the representations that apply to genuine realities from implicitly composite ones that do not. Moreover, he is aware that some of our sensory systems are geared not to detect objects and properties as they really are, but rather to signal contrasts and changes that are potentially relevant to our survival or other practical needs.35 Again he challenges us to distinguish veridical perception from practically useful but epistemically unreliable perception, and objective detection from practically useful sensory signals that do not function to detect objective properties at all. These points go very deep. They enrich the sceptic’s arsenal in ways that are often not adequately appreciated today. Scepticism is not our primary topic, but some remarks on Descartes’s view of reference are in order. Descartes’s view of reference is, as far as I understand it, simpler than mine. I think it too simple. He tends to see unsuccessful reference with kind concepts as the result of our
316
Tyler Burge
making a fictitious combination out of basic ideas for simple natures or out of parts of simple natures. The idea of a satyr is a prime example.36 There are, however, other ways of making referential errors with kind concepts—ways that Descartes does not seem to recognize. The concept phlogiston is not a composite built out of representations for simple natures. It is the product of an explanatory theory that is constitutively dependent not on combination from simpler elements, but on an inference from observational beliefs. Descartes might, of course, extend the notion of combination to this case. He might insist that although we may not think that the concept of phlogiston is composite, it nevertheless is. But the notion of composition or combination would then seem to be so flexible as not to be very informative. I see no evidence that Descartes made use of what we now think of as scientific theoretical explanatory inference in his account of concept formation. On Normore’s exposition, Descartes holds that reference succeeds only when the explanatory cause of the mental event is the same as the explanatory cause of the content of the mental event (its objective reality). In such a case, the cause is identified with the referent. This view, too, appears to incorporate too simple a causal picture. A Martian scientist could refer to H2O even though he or she bore no causal relation to H2O and did not bear causal relations to all the factors postulated in the theory. Suppose that the scientist has causal relations to oxygen and hydrogen and, despite lacking any experimental causal relation to the particular sort of bonding connection between them, guesses or hypothesizes—near enough—the correct bonding relation. Then the object of the idea, H2O, is not the explanatory cause of either the representational content or the mental event. Moreover, there is no straightforward sense in which the cause of the mental event is the same as the explanatory cause of its representational content. I do not see that explanations of psychological events and explanations of representational content are likely to track one another in the case of complex theorizing.37 Nothing depends here on the referent’s being a compound. It could be an element or even a type of elementary particle. A scientist could correctly postulate and refer to such kinds, without having a causal relation to their instances. Descartes might not have counted particles, elements, or compounds as genuine elements of the world—as simple natures. The paradigm for him is geometrically shaped matter. But I do not see that he has the resources to form a plausible account of how we come into a referential relation to actual physical kinds that we now recognize as kinds and that we bear no causal or perceptual relations to. The tools of perceptual reference and combination seem inadequate to the task. Descartes may have thought that our only genuine referential relation to simple natures was quasi-perceptual.38 I think that Descartes’s oversimple account of reference is associated with his seeing all representation as a sort of perception, or a combination of perceptions. This picture
Reply to Normore
317
underlies the tendency to see reference to an object or kind as dependent on causal relations to that object, or else “combinations” of representations each of which bears causal relations to an object. It may seem surprising that Descartes was guided by such a picture, given his focus on the mathematicization of nature. The picture was encouraged by a venerable but now dated conception of mathematics. Although Descartes’s unification of geometry and algebra began the process that eventually freed mathematics for a more abstract view of its subject matter, Descartes joined a dominant tradition, which ran even into Newton’s early mathematical practice and motivated Kant’s philosophy of mathematics, of seeing geometry as epistemically basic in mathematics. Geometry was supposed to be an abstraction from our perceptual experience of objects in space. Thus Descartes seems to have seen the methods of mathematics as quasi-perceptual at their basis, with an attendant abstraction from empirical assumptions of actual existence.39 Descartes saw geometry as studying the shapes of the physical world, with no presumption that they were actually materially instantiated.40 Both mathematics and the use of mathematics in physics have become increasingly independent of their geometrical origins. The progressively more abstract conceptions of mathematics and of physical explanation have forced a more complex picture of the representation of physical reality. The recognition that representation and mathematicization of physical reality can be tied to perception in only very loose and complex ways, involving theoretical explanatory inference, has been forced on us by these developments in the physical and mathematical sciences.41 Descartes’s difficulty with theoretical reference to physical kinds to which one bears no causal relation was hidden by two elements of his philosophy. One is his extremely austere ontology, which admits only geometrical forms of matter as simple physical natures. Such an ontology disallows not only nearly all commonsense macro-objects but even most of the natural kinds of present-day science. The other element is his tendency to blur the distinction between mathematical and physical kinds. Given the austere ontology, he did not need to worry about the sorts of theoretical kinds that I have mentioned. (See note 36.) Given his view of physical kinds as being instances of geometrical kinds, he could believe that all the relevant basic kinds are available to perception informed by geometrical structures. More complicated kinds are constructible by geometrical reasoning from the simpler ones. But Descartes regards even the results of construction ultimately in quasiperceptual terms rather than terms of proof or formal construction.42 It must be said that an analogue of the problem that I have raised for Descartes regarding reference to theoretical physical kinds faces us today. Once a modern distinction between mathematics and physics is in place, one needs to account for reference to mathematical objects (or functions) that are not in any straightforward sense physical properties. We have no causal relations to the objects. I think that we cannot plausibly help
318
Tyler Burge
ourselves to the idea of a theoretical explanatory inference from perceptual references that are grounded causally. Cartesian mathematical perception of geometrical properties does not seem viable either, as a full account of reference to mathematical objects. Geometry is not a foundation for all mathematics. Moreover, pure geometry itself can no longer be seen to directly concern physical space. I think that accounting for mathematical knowledge and mathematical reference requires notions that Descartes did not employ. It requires notions of objective formal structures that inform thought and reason, and rational commitment to entities associated with these formal structures—a commitment that is implicit in the very practice of mathematical reasoning. This is a complex matter whose exploration is not in place here. I differ with Descartes in two other ways that are associated with the difference in our conceptions of reference. One difference lies in our conceptions of physical reality. The other lies in our conceptions of representation-as. I begin with the former difference. As I have noted, Descartes has an extremely austere conception of physical reality. For him, physical reality is made up of extension and parts of extension. This is an impoverished conception even of the world of physics. The subsequent history of physics, beginning with Newton’s recognition of forces as fundamental and continuing with the addition of dynamical and field-relations to mechanical ones in the nineteenth and early twentieth centuries, has made Descartes’s conception seem even more impoverished than it did to his contemporaries. Moreover, I do not accept Descartes’s apparent reductions of physical reality to physics. There are chemical and biological kinds that are fundamental natural kinds. Equally important, there are ordinary physical kinds that do not fit neatly into the sciences; and there are perceptible, artifactual, and social kinds that Descartes tends to treat as modes, or perhaps even constructs, acceptable in everyday practical life but unacceptable as basic in a serious account of reality. I take clouds, rainbows, brisket, rocks, the North Sea, arthritis, redness, shadows, cracks, rough-texturedness, sounds, cold, sofas, clothes, symphonies, the United States—as well as human bodies—to be kinds or individual entities that need not be reduced to parts of extension or of matter or to sequences of collections of particles. And they are not mere projections of our minds. Yet all of these are kinds or properties of physical entities, with the possible exceptions of symphonies and the United States. Thus if Normore is right that Descartes believes that we make such kinds in the sense that we “project principles of unity” for them, then Normore is isolating a basic difference between Descartes’s conception of physical reality and mine. (I am not convinced by this reading of Descartes, incidentally. But there is no question that Descartes thought that such objects have some kind of ontologically secondary status.) I do not agree that such objects or kinds are “ideal” or merely practical. I do not agree that they are in any sense con-
Reply to Normore
319
structed by us. Of course, most artifacts are dependent on our intentionally making them, causing them to come into existence more or less according to some plan. Once made, the artifacts are what they are, regardless of how we regard them. An amplifier is not a kind of thing only by courtesy of our “projecting” a principle of unity whose reality lies entirely in our projection. We fix on and represent kinds, features, and relations in the world. Often our representations reflect interests and needs special to us. One should not, however, conclude that since we represent a pattern only because it corresponds to some need or interest of ours that the pattern is a product or projection from our needs or representational abilities. The world is made up of individuals that instantiate a rich, hierarchical, cross-quilt of patterns made up of properties, relations, kinds. Science deals with those that submit to relatively deep explanatory systematization. But pattern is not less real by being local, or by being perceptible only by certain sensory modalities, or by being constitutively dependent on causal processes that do not fall under the systematic principles of some science. The realities that we represent are largely independent of our “projecting” principles of unity. The unities and similarities that we make use of are for the most part quite independent of us, even where they are of special interest to us, and might be of no interest to some other species. Granting Normore’s historical account, Descartes’s apparent combination of ontological reductionism about physical reality with conventionalism or idealism about ordinary kinds is fundamentally at odds with my view. Even if Descartes does not hold this sort of conventionalism (as I suspect he does not), his reductionistic picture of physical kinds is, I believe, unacceptable. I believe that Descartes’s view of physical reality can be seen, in retrospect, to be one of the more flamboyant products of intellectual hubris. It is no longer a rationally warranted view of the physical world. VIII The difference regarding our conceptions of representation-as is perhaps best developed by considering the thought experiment that Descartes posthumously puts to me in a letter unearthed by Professor Brown, which Normore quotes.43 I will assume for the sake of argument that the letter is genuine, although I harbor some doubts, textual and anthropological, as to its authenticity. Descartes thinks that we have two ideas of the Sun. One is of the Sun as a small disk about the size of a Canadian dollar. The other is of the Sun as a huge gaseous body larger than the Earth. Descartes insists that these are different ideas, not different judgments involving the same idea. Descartes imagines that there is a twin world in which Twin Sun is smaller than Twin Earth, and is made of different gases from the gases our Sun is made
320
Tyler Burge
of. Given the differences in the Suns, Descartes infers that I will count counterpart representations of the Suns as different representations with different referents. He believes that for me the difference must lie in the “information encoded” in representations on earth and Twin Earth. And he further infers that on my view, our Earthean representations encode the view that the Sun is a heavenly body larger than the Earth. He takes this view to be absurd. Astronomy revised our conception of the Sun on precisely this point. Normore produces a further gloss on the argument. He takes Descartes to infer from my holding that perceptual representations specify perceptual kinds as such that I hold that our representation of the Sun specifies the Sun “as such.” He takes Descartes to infer further that such specification means that our conception of the Sun has been different from the Twin Sun conception even before the astronomical discoveries, and different in a way that leaves it veridical of the Sun. This conflicts with our sense that astronomy corrects our conception of the Sun and that we did not have the true conception all along. Normore goes on to criticize this Cartesian argument for presuming that I would say the same thing about as “complex a representation as that of the Sun” as I would the about perceptual representations. And he proceeds to construct a new argument that challenges my position and illustrates Descartes’s different views about representation. Before discussing the new argument, I want to indicate what is wrong with Descartes’s own criticism of my position. It turns on a misconstrual. I agree with Descartes that there are at least two conceptual representations of the Sun. I think that there were at least two before the astronomical discoveries. One such conceptual representation is closely tied to the visual appearance of the Sun. It applies to the object visually represented in such and such a way in roughly such and such visually determined positions over the course of a day. The other one is a conceptual representation dependent on language, that could be shared by blind and sighted alike, or by anyone who did not use the concept in such a way as to tie it essentially to the visual presentation of the Sun. Presumably the reference of the second concept was originally fixed through use of the first, but the mode of presentation came to be separated from essential connection to perception. This second concept is probably the concept normally expressed by the word ‘Sun’ (or Latin, French, etc. counterparts) prior to the astronomical discoveries. Both of these conceptual representations specify ‘as such’. It is crucially important to be careful about exactly what specifying as such amounts to. If the letter to me is genuine, Descartes holds that the visually based conceptual representation is committed to taking the Sun to be about the size of a Canadian dollar. This is surely a mistake. The visual system specifies size and distance in terms of a variety of factors. It does not, of course, indicate size and distance purely in terms of the proportion of the visual field occupied by the visual image. Angular disparity between the images of the two eyes is a central and representative factor for determining distance and hence size.
Reply to Normore
321
In the case of the Sun, the usual cues that would indicate how far away the Sun is are not available to the perceptual system, apart from elaborate astronomical experiments. Ordinary changes of position, on Earth, with respect to the Sun would not discernibly change the angular disparity. So the visual system will not produce a representation of how far away or how large the Sun is. It will produce a representation of an object that specifies a certain shape, color, and (over time) series of positions in the sky. The visual concept that conceptualizes the visual perception should have similar commitments. Given the anomalies about distance and size, the perceptual concept will remain uncommitted about the other objective characteristics of the object, much as a perceptual concept of an object known to be seen at a distance and poor light would be. Thus I believe that reasonable people who do not know the astronomy might well leave open whether some of the represented visual characteristics of the Sun are real or merely apparent. The nonperceptual concept of the Sun also specifies it as such. The specification is merely as the (or a) Sun. Neither concept “encodes” mistaken opinions or astronomical knowledge about what the Sun is. Mistaken opinions as to the Sun’s size did not arise merely from analyzing a common concept. It is plausible to me that ‘the Sun’ in the latter nonvisual usage, before the astronomical discoveries, was a proper name rather than a specification of a specially salient Sun, among many other Suns. In that case, it is especially obvious that the conceptual representation does not encode mistaken descriptional information. I shall, however, discuss the case in a way that remains neutral as to whether the second conceptual representation is a proper name or an indefinite description that uses the general concept of a Sun. Of course, there is something to Normore’s point that people had a mistaken conception of the Sun, which got corrected by astronomical discoveries. I distinguish concept from conception.44 A conception incorporates (correct or incorrect) explicatory glosses on the concept. Some of the glosses might include empirical information or misinformation. It might well be that before the astronomical discoveries people who had both concepts went beyond what the visual system presented regarding the size of the Sun and took it to be less large than the Earth, perhaps even as small as a Canadian dollar. (As I have intimated, I think that this latter view would have been irrational; perhaps both views would have been irrational.) These views might even have been considered basic explicatory truths by some. So it might be right to hold that astronomy came to correct the common idea or conception of the Sun, in that sense. I see no reason to think that the mistake was incorporated into the concept of the Sun before the relevant discoveries. The mistakes did not ensure that their concepts failed to apply to the Sun; nor did mere mastery of their concepts ensure that they made mistakes about the Sun. As to the thought experiment, I believe that if Twin Sun looks from Twin Earth exactly as the Sun does from Earth, and follows an observationally similar course through the sky,
322
Tyler Burge
then the perceptual representations and perceptual concepts will be type-identical. The difference in reference would be in the demonstrative-like applications of the perception and perceptual concept to different individuals.45 The nonperceptual concepts, respectively used on Earth and Twin Earth, would similarly specify different individuals. Whether the Twin and Earthean nonperceptual concepts are different concepts depends more on the usage of the concepts than Descartes has specified. In the likely case that the concepts are expressed by proper names (‘the Sun’), then they would seem to be type-identical names—and type-identical name-concepts—with different etymologies that are contextually applied to different individuals (as ‘Smith’ might be applied to different Smiths).46 If the concepts are kind concepts, then their anti-individualist individuation would seem to depend not on composition of the objects that cause them— the Sun and Twin Sun. Individuation would seem to depend on the objects’ being heavenly bodies that bear some cyclical spatial relation to Earth and Twin Earth respectively. Since the cyclical relation is by hypothesis the same, it seems broadly plausible to me that the kind concepts are (prior to astronomical discovery) the same. They are applied contextually, in a demonstrative-like way, to different individual Suns. I see no reason to think that in either case the nonperceptual concept changed or needed to be corrected after the astronomical discoveries on Earth. It was indeed a surprise that the cyclical relation consisted in the Earth (Twin Earth) revolving around the Sun—and at a huge distance—rather than vice versa. I do not, however, see any reason to think that it was built into the perceptual concepts or the language-dependent concepts that the cyclical relation had the particular character that pre-discovery physicists (or common folk) thought that it had. Even if these false astronomical theories were used in conceptual explications, they were not built into the concepts being explicated. The concepts themselves carry no essential, heavy astronomical commitments. That is why they continued to refer to heavenly bodies, even though the astronomical theories, and conceptions associated with the concept, were rejected. Whether a new scientific concept that represents the Sun emerged after the Copernican revolution, or other astronomical discoveries in science, seems to me unclear, and indeed doubtful. Certainly, new scientific accounts of the Sun and new conceptions of the Sun emerged. The main trouble with the Master’s Twin Sun objections, at least as presented in his alleged letter to me, lies in a misconstrual of how much I build into the notion of specifying a referent in a given way—as some such. I see concepts, especially empirical concepts, as usually noncomplex in themselves. The concept water specifies water as water. The linguistically dependent concept of the Sun specifies it as the Sun. The perceptual concept of the Sun is no more complex than the image together with abilities to reidentify and track the Sun visually over a day. Visual specification is phenomenally but usually not theoretically complex.
Reply to Normore
323
Having the concepts requires having inferential relations to other concepts. In some cases, some inferential relations to specific concepts are essential to having a given concept. But in most cases, I think it a philosophical mistake to regard strong, detailed empirical commitments as built into ordinary macroconcepts. The ordinary specification of water as water is held in place by its causal relations to the environment, by very generic classificational relations, by perception, and by a network of perhaps collectively necessary but certainly individually dispensable empirical inferential connections and connections to perception. What holds a concept in place is not in general built into the concept as intrinsic components. Concepts are not bare. The concept water has a definite content that specifies water in a particular way. But the concept is not richly articulated individually and in itself. Conceptual richness derives from inferential connections to other concepts, from applicational relations, and from relations to perceptual representations. Some of the inferential connections are apriori. Others are empirical. As I have emphasized, in most cases the apriori connections are very generic ones. Breaking empirical inferential connections does not normally undermine the concept or its success in referring, except in the cases of massive empirical error. If the term is applicable more or less directly on the basis of observation, the error usually must undermine the perceptual basis for the referential application of the concept if it is to show the concept fails to indicate a property. If the term is more theoretical, the error usually must radically undermine the type and direction of the theory in which it is embedded (as in the case of the concept of phlogiston).47 Normore goes on to offer a variant of Descartes’s challenge. He presents a slippery slope “between the view that some of our basic percepts are veridical apriori and the view that we cannot be mistaken in any of our perceptual representations” (p. 11). He writes, “Somewhere along this slope Burge will dig in his heels, and it would be very helpful to know where and why.” He sees my view as differing from Descartes’s in taking perceptual kinds rather than true and immutable natures as “basic” (p. 10). And he attributes to me the view that “for perceptual kinds the perceptual representation must represent the kind as it is.” But this is not quite my view. I do take some perceptual concepts as basic in the sense that they are not made up of more basic representations and in the sense that they are among the concepts that can be utilized in producing genuine knowledge. I do not take them as more basic—in these senses, or in any other senses I can think of—than mathematical concepts or other concepts that Descartes might regard as specifying true and immutable natures. I have not set foot on the proffered slippery slope. So I need not dig in my heels anywhere on it. My anti-individualist view about particular perceptual representations like crack, edge, dark was not defended from a purely apriori view. Recall that I assumed that
324
Tyler Burge
perceptual objects and properties—the particular ones that we represent—are objective, mind-independent entities.48 I think that in the absence of special argument, this assumption must be construed as an empirical assumption. I assume that there are such things as edges, bars, boundaries, (approximate) cones, surface textures, and so on. I assume that we perceive the objects and kinds in the world that we think we do. Given these empirical assumptions, I maintain that the contents of perceptual representations and concepts essentially associated with such representations constitutively depend on causal relations between individuals and an environment. Thus I did not assume or argue that we know apriori that our basic percepts are veridical of the perceptual kinds they purportedly represent. Such a view would be relevant to certain discussions of scepticism. My account of individuation conditions has explicitly bracketed attempts to answer scepticism.49 I did and do claim that those perceptual representations that veridically apply to objective kinds get their content through complex interactions with instances of some perceivable kinds. I think that this claim has apriori status. But maintaining apriori that those perceptual representations that actually do veridically apply to objective kinds do so by virtue of a network of causal relations between the relevant kinds in the environment and the perceptual system is quite different from showing apriori that representations that purport perceptually to refer to mind-independent entities in fact do so. I believe that anti-individualism yields material relevant to answering the sceptic. I think, for example, that it is apriori that our purported representation of particular spatial relations would be impossible if we or our perceptual systems were not at some time veridically and reliably in perceptual touch with things in space.50 I think that it may be apriori that some or most of our purportedly perceptual representations veridically apply to kinds that were at some time actual. Perhaps there are more specific apriori points that would be relevant to contending with the sceptic without assuming from the beginning the objectivity of perceptual reference, as I have in most of my work. Anti-individualism certainly suggests a challenge to the sceptic to show how representations that purport perceptually to represent objective entities could do so unless they bore some relation, through theory or composition, to some successful application of some perceptual representations to some objective kinds, somewhere in the history (perhaps the evolutionary history) of the formation of such perceptual representations. I believe that ultimately the sceptic cannot coherently meet this challenge in such a way as to leave open the possibility that all of our purportedly perceptual representations fail to refer to mindindependent entities. I have not probed all the complexities here. As I have noted elsewhere, some perceptual presentations are systematically unreliable, though useful for survival.51 How are we to show apriori that some are epistemically reliable? Can we know apriori which ones are?
Reply to Normore
325
Can perceptual mistakes derive from bad theory? Can such a possibility be accommodated in an answer to scepticism? Some representations may seem to perceptually represent intrinsic objective properties, but in fact represent only changes in our sensory states or our bodily surfaces. There is the standard example of plunging one hand into lukewarm water after it has been in hot water, and the other hand into the same lukewarm water after it has been in cold water. It is easy to think mistakenly that the products of the relevant sensors indicate the objective temperature of the water, whereas they in fact give indications of sudden temperature changes on the skin’s surface or even sudden changes in sensory qualities. The sceptic might claim that as far as we know, we confuse sensory presentations with purportedly perceptual representations and these in turn with perceptual representations. This would constitute a claim that we are ignorant of the intentional content of our apparently perceptual representations. These and more issues would have to be addressed in a discussion of scepticism. I believe that such sceptical moves would face serious difficulties. Pursuing this dialectic would, however, be complex. The apriori principles that I have alluded to, and that are relevant to confronting a sceptic, do not seem to me to threaten a slippery slope toward claiming that we cannot be mistaken in any of our perceptual representations. I have allowed that error in the application of particular perceptual representation tokens, purportedly of objects or kinds, is always possible. Thus it is always possible to perceive an individual as being smaller or closer than it is, or to mistake a crack for a discoloration, or even to have a perceptual representation that lacks an individual referent. I have so far not claimed apriori of any particular, nongeneric, purportedly perceptual representation that it applies to a genuine mind-independent kind. I allow that error about the existence of purportedly perceptual kinds can occur through false theory. We can misconstrue sensory presentations of contextual changes local to the individual with perceptual representations of objective entities. I also grant that one can misconceive the nature of perceived objects. Whether the perceptual mistakes, or even the mistakes of conception, that I have allowed for could be systematic and generic in such a way as to raise rational doubt about the physical world seems to me questionable at best. The nature of content determination suggests that these various sorts of errors must be confined. I believe that it is apriori and relevant to scepticism that some (purportedly) perceptual representations (purportedly) of objective entities depend on successful applications. Representation of types is parasitic in complex but inevitable ways on successful application to instances of some types. For all that, these are suggestions that require far more development and defense than I have attempted to provide.
326
Tyler Burge
IX Before concluding, I want to comment on the role of the social in individuating representational content. I have already indicated that if Descartes sees the social as projecting principles of unity so as to put brisket or arthritis or the North Sea into the world, then my position is quite different. I think that these are kinds and individuals in the world that we in no sense put there. Communities may take an interest in some pattern, property, or kind that is not a natural kind, in the sense that it is not subject to some systematic comprehensive law of nature. Communities may introduce words or develop concepts for these kinds. And individuals in the communities may develop patterns of reliance on other individuals to connect them to these kinds through interlocution, or to set standards for correct application of the words or concepts. In my view such kinds are still not ideal, socially independent, or mind-dependent. Communities are not metaphysically necessary for the development of kind concepts (natural or un-natural). They are not even necessary for establishing conditions for the having of those nonperceptual, non–natural kind concepts that allow for Twin Earth thought experiments. One can be doubtful or agnostic about the application conditions of almost any empirical concept that one has, even if one does not rely on a linguistic community to fix the referent. Such doubt or agnosticism can fuel anti-individualistic thought experiments.52 The key factor in bringing out the anti-individualistic character of having empirical concepts is the gap between what we must know about a referent of a nonindexical representation in order to represent it as a kind, type, property, or relation, and the way the referent actually is. It is objectivity and lack of omniscience, not natural kinds or community, that lie at the root of anti-individualism. Communities figure in the individuation of our actual concepts. This fact had been neglected, and I tried to highlight it.53 I do not agree with those who hold that one cannot have concepts except in a community. Nor do I think that being in a community fixes one’s concepts regardless of one’s intentions. The role of communities in individuating concepts is not metaphysically necessary. Though profound and central in actual human culture, it is not dictatorial. Descartes gives the appearance of being more centered on the individual’s thought, and less interested in the ways individuals depend on their communities, than I am. But it is hard to determine whether this is oversight or principled disagreement. My view of the role of community in determining what concepts an individual has does not fall under any simple generalization that I know of. I certainly do not, for example, think that communal ways of individuating intentional contents trump other ways no matter what the individual intends. Individuals can quite self-consciously opt out of communal usage. Or they can be legitimately idiosyncratic, indifferent, or oblivious. As long as they abide by their
Reply to Normore
327
idiosyncratic usage and do not unconsciously rely on or draw on—or otherwise become responsible to—communal usage, their words and concepts need not be dependent communal usage, or the ways that usage is tied to the physical world. So communal practice does not trump individual intention.54 Individual intention is not, however, sufficient to opt out. The individual’s intention must accord with the individual’s practice. An individual can stubbornly resist correction from communal usage and be mistaken to do so. He must tailor his usage to his resistance. Or rather the resistance must be tailored to his usage and the commitments of that usage. An individual can intend to use ‘arthritis’ in a way that disregards communal practice and communal causal ties to a referent, but be disposed in various unrecognized ways to be corrected by that practice, or be otherwise subject to the practice and its referential connections to the world. Although concept individuation is a matter of choice, narrowly understood, only in special cases of stipulated (and adhered to) usage, it is in all cases heavily dependent on the individual’s own attitudes and on how and whether the individual relies on his fellows. It is a deep fact about human culture—dare I say “human nature”?—that we do rely on others in ways that make others’ intentional activity sometimes play a role in individuating our own. Descartes underemphasizes this fact, even though his view is probably in the broadest sense anti-individualist. Like most early modern philosophers, his concern was with the individual’s relation to God, to himself, and to the physical world. History and community seemed to him sources of superstition and inertia. A full understanding of individual minds cannot ignore, to the degree the early modern period did, ways that individuals depend on others for knowledge, reference, and intentional content. Notes 1. See note 4 of my (1988). 2. I have one caveat about Normore’s presentation of Descartes on these matters. Normore reads Descartes as an externalist “in the sense that the content of our ideas depends on their causes—but on the cause of the objective reality of the idea, not the cause of its formal reality” (p. 7). It is not sufficient to be an anti-individualist (or externalist) that one hold that the content of our ideas depends on their causes. One needs the further points that the causes are external to the individual and that the dependence is individuative, not merely causal. The caveat does not alter the fact that Normore seems right to hold that Descartes is an anti-individualist. 3. I read Meditation III and VI as illustrating this drift. I think that this issue in interpreting Descartes is worth further investigation. I have benefited from several conversations with Normore in my remarks about Descartes. Normore recommends further reflection on God’s creation of real possible natures. God’s idea and the natures are created in the same act. Nevertheless, Normore sees Descartes, here as elsewhere, as tending to take the nature that the idea refers to as explanatorily prior to the idea. 4. Descartes (1988), p. 29; Oeuvres de Descartes, (1964–76) VII, pp. 41–42. I am indebted to Calvin Normore for discussion of this passage and other passages in Descartes’s interchange with Arnauld in the fourth
328
Tyler Burge
set of Objections and Replies. The essay as a whole has benefited from several conversations with him and with Deborah Brown and Lilli Alanen. 5. I wrote, “Individualism is the view that an individual person or animal’s mental state or event kinds . . . can in principle be individuated in complete independence of the natures of empirical objects, properties, or relations (excepting those in the individual’s own body . . .)—and similarly do not depend essentially on the natures of the minds or activities of other (non-divine) individuals” (my retrospective emphasis). See my (1986b), pp. 118–119. 6. God’s mental events are special and may not be subject to anti-individualist considerations. The objects of God’s mind depend on his creating them through his thinking them. See, however, note 5. Moreover, I will assume that Descartes’s holding that a human mind’s idea of God depends on the existence of God is not incompatible with his substance dualism, since no finite substance is completely independent of God. Similarly, I assume that the fact that Descartes holds that God could bring about anything, including the falsity of eternal truths, does not show that Descartes was an individualist. Even though he thinks that relations between having ideas about body and being in relation to the physical environment are, in this sense, contingent, I take the issue to remain alive. If anti-individualist principles were, like mathematical truths, necessary except for the qualification about God’s power, then I would regard Descartes as an anti-individualist. I was led to make this point through a criticism by Carl G. Anderson. 7. Descartes to Arnauld, July 29, 1648, in Descartes (1981). I am indebted to Deborah Brown for calling my attention to this passage. 8. See Descartes’s reply to Hobbes’s worry that Descartes had reduced the thinking agent to thinking—the second objection in the Third Set of Replies (Descartes 1988), pp. 122–124. 9. There is a body of scholarly opinion that takes Descartes to hold that mental substance consists in its principal attribute. I have some sympathy with this view, but I think that it should be developed in the light of the difficulties that I am raising. As will become clear, I believe that Descartes has a way out of these difficulties. As far as I know, however, the problem that I have been raising has not been addressed. For a brief discussion of the construal of Descartes that takes mental substance to consist in its principal attribute, see Marleen Rozemond, Descartes’s Dualism (1998), pp. 8–12. 10. This is a move that Sellars pressed. I am not convinced by Sellars’s view. I shall discuss the matter elsewhere. See Wilfrid Sellars, “. . . this I or he or it (the thing) that thinks . . .” (1970). 11. See Descartes, Letter to *, 1645 or 1646, Descartes (1981), p. 187. (The letter is to an unknown recipient, jocularly called “Starman” in the trade.) I owe this reference and useful discussion of this issue to Lilli Alanen. For discussion of the distinction, see her “On Descartes’ Argument for Dualism and the Distinction between Different Kinds of Beings” (1986). 12. See my “The Indexical Strategy: Reply to Owens,” this volume. 13. Thus I do not agree with Normore’s remark that “the individualist as such need not reject Burge’s three premisses” (p. 2). When I wrote the paper I seem to have acknowledged that the third premise already entails antiindividualism, at least when conjoined with the second. See my (1986b), p. 135. I claimed, however, that the first premise—and implicitly the whole argument—was needed to show that nonindividualist methods of individuation do not “in principle” have counterpart methods that are individualistic. I think that I wanted to show that the possibility of error made it impossible, contra Leibniz, to reduce the relations that are necessary for the individuation of mental states to nonrelational properties of the individual. But I seem not to have firmly recognized that if intentional mental states and events are individuated relationally, then their individuation conditions are necessarily relational. I do (and did) not understand individuation merely as a contingent practice. It is a condition on the nature of the mental states and events—a necessary fact about them. I made substantially the same error, involving a similar blurring of the difference between failure of local supervenience and antiindividualism, in my (1986c), pp. 3–45. 14. Traditionally, God was said to have such a power to think of things without any discursive or general representation associated with the thinking. The power was called “intellectual intuition.” I regard this view as of doubtful coherence. For present purposes I maintain the more circumspect view that such reference in thought is impossible for finite beings, whose perspective on any object is limited. Russell made the mistake, at one stage in his career, of in effect thinking of acquaintance as the fundamental representational power, where acquain-
Reply to Normore
329
tance has all the key nonperspectival aspects of intellectual intuition except for bringing the objects of thought into being. For many of Russell’s purposes the objects are treated as perspectives. This is obvious in his treatment of sense data as the objects of acquaintance. Qualitative elements of consciousness are one thing. Singular representation of them in thought is another. Treating them as data for perceptual belief is a third. Singular representation of qualitative elements utilizes demonstrative-like applications associated with general concepts that make use of the qualitative features as repeatable types. Russell runs these things together. Russell took universals both as properties of objects and as perspectives of the mind on objects. I believe that this is another fundamental conflation or confusion. Russell never provided any discussion or defense of his fantasy about human epistemology and the mental abilities that go into making reference possible. All of the foregoing concerns the natures of belief and perception and of human psychology and epistemology. It seems to me a separate question whether an illuminating theory of aspects of language can abstract from the perspectival character of thought and perception in some instances. Even in this area, I am inclined to think that the perspectival character of linguistic representation is never fully obliterated in linguistic natural kinds. 15. As noted, I am inclined to believe that concepts like God and three are individual concepts, the conceptual counterparts of individual constants. They are ability-general, but formally singular. Unlike ordinary proper names, they are not associated with demonstrative-like determiners. See my (1973), pp. 425–439. I believe that the rejection of the second thesis applies to individual concepts. I think that these concepts also cannot be thought autonomously unless they are associated with and guided by formally general, predicational concepts. Thus in thinking the concept three, we must associate the concept with formally general concepts that guide the referential application. We presuppose a capability to think that three is a number, or three is the (natural) number immediately following two. Singular uses of three are conceptually associated with adjectival or quantifier uses, which have an obvious second-order formal generality: There are three oranges on the table. Some such associations of three with formally general concepts help guide its application, in the sense that they are taken as prominent in determining the referent. Similarly for other individual concepts. Traditionally, the concept God was understood as entailing certain general attributes—deity, power, knowledge, agency, and so on. One could not have the concept without associating it with at least some guiding, formally general concepts of associated attributes. Autonomous applications of concepts like Tlaloc for more primitive deities also seem necessarily to be associated with such concepts as agent or god of rain, as well as with certain images. In sum, I believe that individual concepts require some formally general conceptual associations to enable them to be context-free. 16. I discuss this singular sort of context-dependent representation, insofar as it occurs in thought, in my (1977), pp. 338–362; and “Russell’s Problem and Intentional Identity” in Agent, Language, and the Structure of the World (1983). I discuss such singular representations insofar as they occur in perception, in “Perceptual Entitlement,” forthcoming. Such perceptual singular elements must be recognized if one is to account for the fact that perception involves representation of particulars that need not be uniquely specified by the formally general perceptual representations of aspects (or properties and relations) of the particulars. Perception is of the particular objects and property or relation instances that the perceiver interacts with. It represents those particulars, not some look-alikes that the perceiver is not interacting with. Analogous singular elements in thought are needed to account for the fact that we can think about objects that we do not fully specify through ability-general or formally general conceptual representations. We apply the general conceptual representations to particular objects. The singular, context-dependent application must be seen as associated with, or marked by, a singular representation in order to account for the fact that beliefs about such particulars are true or false. Of course, such demonstrative-like token-based representations in thought are commonly guided by counterpart singular contextdependent elements in perception. The guidance may rely on memory or interlocution. I think, however, that there are singular, demonstrative-like representations in thought (such as applications of the first-person concept) that are not ultimately guided by perception, even by way of memory or interlocution. Thus not all de re thought is ultimately perception-based. There are context-dependent formally general representations, like such, next, later. I think that an analogue of the thesis about to be stated applies to them. They require association with non-context-dependent formally and ability general representations. I also conjecture that deictic application of these context-dependent formally general representations necessarily involves at least implicit use of a context-dependent singular representations. Thus deictic use of such an animal is associated with singular representation of a particular animal, or of a picture of an animal. 17. The thesis actually needs some further qualification that I will not go into here. I believe that context-bound, formally singular egocentric indexing elements that provide origins for frameworks, such as spatial frameworks
330
Tyler Burge
or motivational frameworks, are exceptions to the thesis as stated. Similar qualifications will be needed for my rejection of the next thesis. I shall develop these matters further on another occasion. 18. I think that each context-dependent singular representation in perception must be accompanied by a general perceptual representation. A general element accompanies a singular element if the general and singular elements are part of a single complex representation, and the meaning of the complex representation is such that the general element is supposed to be veridical of the referent (if any) of the singular element. Contextdependent singular representations in thought can be associated with a general representation by being accompanied by a general conceptual representation. (I do not count the most general representations like entity as fulfilling the requirement of association. Young children probably lack such concepts. Moreover, such concepts do not serve the purpose of putting a putative constraint on the putative referent of the singular representation.) Context-dependent representations in thought can also be associated with general conceptual representations by being anaphorically connected (through memory or inference) to singular elements in other thoughts that are themselves accompanied by general conceptual elements. I am inclined to think that association in thought can be even more permissive. I am inclined to think that a context-dependent singular representation in thought can be associated with a general perceptual representation—if the singular thought is guided by an unconscious perceptual representation of a particular, where the general perceptual representation has not been conceptualized and where it accompanies a singular perceptual representation that picks out an object. Imagine that one believes that something is bothering one. One can particularize the thought to the extent that one thinks “there it goes again, bothering me.” But one cannot say what kind of thing it is that is bothering one. We might suppose, however, that by psychological tests we could find that the perceptual system is picking something up and referring to it, even though the belief system, or system of propositional thought, has not categorized it. 19. Thus I think that principles governing when singular application is possible that have been articulated by Quine, Strawson, Evans, and others have obvious counterexamples. The theoretical underpinning of these views seems to me insufficient to deal with the counterexamples, or with other theoretical considerations. Much current discussion of these issues is influenced by Kant’s dictum that intuitions without concepts are blind. Some of this discussion has lost touch with Kant’s view of concepts as essentially ability-general representations. Moreover, some of the discussion of Kant’s famous dictum is committed to what I regard as empirically refuted views— for example, that insects that lack concepts cannot have perceptions that single out individual objects or cannot have perceptual categories for physical objects or properties. I discuss these issues in “Perceptual Entitlement,” forthcoming in Philosophy and Phenomenological Research. 20. The point of these three paragraphs is, of course, a variant of Kant’s points that concepts are elements of propositional attitudes (paradigmatically for him, judgments), and that perceptions have a formal structure. They are the origin of Frege’s point that meaning occurs only in the context of (or only through presupposing) a proposition, and that propositions are essentially bound up in inferential structures. It is also the origin of Wittgenstein’s point that naming can occur only in the context of a larger representational context. 21. In the study of language these cases were introduced with special vividness by Keith Donnellan in his (1966). 22. See Keith Donnellan (1966, 1970); Kripke (1980); Kaplan (1989); and my (1977). I think that this is a necessary truth about empirical thought, as I argue in “Belief De Re” (1977). Certain types of singular reference involve ineliminable singular elements whose referents depend partly on causal relations external to the thinking individual. 23. For a defense of such a denial, see Evans (1982), pp. 105–120. Evans requires that to refer to an object, one must know which object it is by discriminating it from other objects by perception, description, or recognition. More generally, he requires that one know what sort of thing would make one’s thought true. And in working out what this requirement means, he places restrictions on reference through ordinary cognitive capacities that I regard as poorly motivated and quite unacceptable. I find his arguments for this view as applied to memory and interlocution (pp. 127ff.) unpersuasive, his account of thoughts about natural kinds (e.g., p. 117) mistaken, and his strictures on perceptual belief quite excessive. I think that his view is backed by a sophisticated but rearguard defense of the old over-reliance on agent knowledge and control in determining a referent, an over-reliance driven by the traditional philosophical hyperintellectualization of accounts of thought. A detailed critical discussion would be out of place here. I will ignore universal categorizations like “object” or “entity.” The child need not have these. More important, I regard the thesis as making a stronger claim than such a defense would defend.
Reply to Normore
331
24. I am taking the distinction between perceptual representation and conceptual representation for granted here, as I did in “Cartesian Error and the Objectivity of Perception.” I shall discuss the distinction in future work. 25. For empirical discussion of such levels of abstraction see Sutherland (1960), (1971), and Marr (1982), esp. chapters 2–5. 26. I think that the same applies to seeing objects through magnification when the individual fails to realize that the seen objects are not in the vicinity. This is contrary to Evans’s view, which I regard as in other respects too restrictive an account of spatial reference. See Evans (1982), pp. 151–170. For criticism of Evans that I largely agree with see Marleen Rozemond, “Evans on De Re Thought” (1994). The details of these issues deserve further development on another occasion. 27. Event perception probably complicates the picture further. I think that seeing a flash of light, or a scattered group of explosions, again requires getting on to the boundedness of either the whole flash or a sufficient number of the individual component events. But I believe that event perception requires further reflection. 28. The view that causation between object and perception is a perceptual category was defended as a thesis in psychology some years ago. See Michotte (1963). There are numerous empirical objections to the theory, and it is no longer taken seriously in psychology. John Searle gave a more apriori argument for the view in Intentionality (1983). I criticize his argument in my (1991). 29. I believe that appeals to such metarepresentations in discussions of reference are usually cheap and mistaken. In the present case, the presence of such a metarepresentation is, I think, forced by the sophisticated and somewhat artificial (but still possible) move by the individual of canceling the standard presumption of spatial location generated by the perceptual system. Of course, even in this case, the metarepresentation is not sufficient to specify the referent. Which relevant cause or explanatory factor the referent is is a matter that the individual is usually not going to be able to specify, unless he is a very able philosopher. The general metarepresentation provides only a loose restriction on the reference, a necessary condition, not a sufficient one. The demonstrative element in the thought remains irreducible. 30. This is a form of indexicalism. For discussion of other forms of indexicalism see my “Phenomenality and Reference: Reply to Loar.” See also “The Indexical Strategy: Reply to Owens.” I think that all positions that take all our empirical concepts to be essentially indexical are out of touch with the specificity of our actual conceptual and perceptual representations. I think it beyond serious doubt that our ordinary empirical concepts are not introduced in the way indicated in the text. They are commonly constitutively tied to perception or to other concepts. Would concepts so introduced provide supporting instances for the fourth thesis? The apriori connection between the concept and the concept of an explanatory factor would remain. The introduction, associated as it is with perception, must presume that what the concept applies to has causal properties. If the concept is allowed to depend on the perceptual representations of the introduction, there is the further connection to the concept of spatial location, and any other concepts that categorize limitations on the relevant perceptual reference. I think a thorough discussion of the case would require investigation of conditions that limit what counts as an introduction of a concept—the perceptual and presupposed unifying conditions that enable an explanation to get started. 31. Putnam (1962). Whether Putnam ever intended to hold the thesis in full generality is unclear to me. 32. I will not discuss in detail here some interesting issues involving proper names. I think that we know that if Socrates existed, Socrates was not a transfinite number and was a temporal being. This is not to deny that some names could be introduced through interlocution into the repertoire of a recipient (as with the two adults and the child, discussed above) in which there are no conceptual representations available to the individual that restrict what kind of referent is involved. I think that in the case of the most famous Socrates, however, too much lore has accumulated around the name to admit assimilation of this case to the case of the child’s use of a name that is grounded in stipulation by the adults. If the name ‘Socrates’ as we use it began as such a joke about an irrational number, the right conclusion would be that Socrates did not exist (even though an irrational number’s being named ‘Socrates’ got the whole story going). Whether we know this apriori, or know it only by knowing something about the historical lore surrounding the name ‘Socrates’, is open to question. For discussion of effects of accumulated lore on reference, in some cases, see Evans (1973). 33. It is important to remember here that being apriori warranted is not equivalent to being invulnerable to empirical counterconsiderations. Apriori warrant concerns the source of positive support, not sources of possible
332
Tyler Burge
overthrow. Thus a belief can be apriori warranted even though it is vulnerable to possible empirical overthrow. The mere fact that someone theorizes that there is no space or time and that it is epistemically possible that this may turn out to be empirically supported does not show that our warrant for certain applications of spatial concepts is not apriori. 34. Normore speculates that I would say that in the “Ur demon world” all our thoughts are about the demon or about ourselves. He holds that Descartes would say that we have no general thoughts at all because there are no natures (p. 9). What I would say depends on a more detailed account of the relation between the demon and us and of how the demon purportedly thinks. I am not committed to disagreeing with Descartes on this matter. 35. For a recent discussion of empirical aspects of this point, see Akins (1996). 36. I take it that Normore is right that simple natures, for Descartes, do not include human or goat bodies. Although the ontological status of ordinary bodies is obscure, it seems to me that Descartes’s satyr example in Meditation I is meant to exemplify a primary sort of error. Satyr representations are made up of parts that veridically apply to the simpler natures. The simpler elements one most immediately thinks of (human heads and torsos, goat legs) are not genuinely natures, but can be regarded as such for the sake of illustration. Ultimately the real simple natures are parts of extension. 37. I cite such a case in my “Other Bodies” (1982). 38. One might maintain that with such a simple view of elementary natures and of the ways they relate to one another to form complexes (basically part–whole ways), Descartes can afford to rely on his simple account of reference. Everyone has had causal relations to chunks of matter and to part–whole relations. Assuming that all geometrically possible combinations are innately available to our mathematical intuition, perhaps Descartes can hold that it is safe to assume that we bear causal relations to all the genuine constituents and have access to all the genuine relations needed to form all kinds that we in fact have ideas of. Then my objection would be to his ontology, which I shall discuss shortly. 39. Shuster, (1980); Gaukroger, “Descartes’ Project for a Mathematical Physics,” in Gaukroger (1980). To Descartes’s credit, he came to place less and less emphasis on the role of images in mathematical thinking. 40. Descartes, Conversations with Burman, AT X, 160; (1976), p. 23. Cf. Meditations VI, AT VIII, 79–80. 41. I believe that representation of most mathematical reality is in principle independent of perception not only for its justification but also for individuation of its content. This is a complex issue that I will not pursue here. 42. See Hacking (1973). 43. Brown (1992), pp. 167–168. Elsewhere Professor Brown has provided evidence that Descartes faked his death in Sweden, emigrated to San Diego for a warmer climate, left there because of an inhospitable intellectual environment in the philosophy department, and moved north to Santa Monica, becoming very fond of beach life. The circumstances in which he allegedly wrote his letter to me are somewhat obscure. Something like the point about the ideas of the sun can be found in Meditation III, though I am not confident that the doctrines of the letter and those in the Meditations are precisely the same. 44. See my (1986a), where I use the term ‘normative characterizations’ for conceptions, and my (1990). 45. See my (1982). Even if the perceptual concepts were to specify the sun more richly than I think they do, I would see no reason not to say that they are the same. As applied on earth, the visual representation and visual concept would provide false information about the size of the referent; as applied on Twin Earth, it would produce either true information or less false information about its referent. But the information would seem to be the same. The difference in visual referent and the consequent differences in truth-value would depend on indexical differences, not conceptual or perceptual differences. 46. See my (1973). If one thought, as I do not, that proper names of different objects are mere homonyms, then the difference in name-concepts would trivially derive from the difference in their individual referents. Again, no associated descriptional material is encoded in the concept associated with the name. 47. Descartes’s views of reasoning strongly suggest that he thinks that clear and distinct ideas are not complex. At any rate he seems to make little use of formal compositional relations in his account of scientific or mathematical reasoning. And he thinks that reasoning that hinges on formal relations among thoughts plays no significant or essential role in scientific discovery, even in the mathematical sciences. This suggests that structural relations among ideas as well as propositions are irrelevant to reasoning, hence not a significant part of his theory.
Reply to Normore
333
Cf. Ian Hacking (1973). Thus Descartes is in a position to take a sympathetic view of my conception of most concepts. The differences lie in his view that perceptual concepts, and indeed most of the other ordinary macroconcepts of the world, are confused or infected with material error, and thus not a proper basis for scientific inference. This view, as I have said, seems a piece of hubris with no sound grounding. 48. I wrote in the my first “premise” (in 1986b), “our perceptual experience represents or is about objects, properties, and relations that are objective. That is to say, their nature (or essential character) is independent of any one person’s actions, dispositions, or mental phenomena.” Later I glossed this premise as “we make veridical perceptual reference to objective entities.” I wrote in my second “premise,” “Representations specify such objective entities as blobs, bars, boundaries, convexity, cones, rough texturedness, being farther from x than from y; and they specify them as blobs, bars, boundaries, and so on.” Both of these points assume that perceptual reference is veridical and is to mind-independent entities. 49. See e.g. my (1982). As I indicate in “Some Remarks about Scepticism: Reply to Stroud,” this is an issue that I hope to say more about. 50. One can use ‘perceptual representation’ to apply only to representations that have sometimes succeeded in applying to kinds or properties that they purport to apply to. We could perhaps mistake a representation as a perceptual representation of something outside us when the kind or property represented is not outside us. We could mistake a presentation of a sensory quality for a perceptual representation. If the term ‘perceptual representation’ is used in the way just indicated, then the issue for the skeptic is whether we know which representations are perceptual and which perceptual representations are of objective kinds. This is why I shall use ‘purported perceptual representation’ in what follows. Of course, throughout, I am not discussing scepticism about particular perceptual experiences, but about the applicability of perceptual representations to corresponding kinds. 51. See my (1996), note 11. 52. A natural variant of the thought experiment I give regarding sofas in my (1986a) shows this. The individual could wonder whether sofas are really religious artifacts rather than articles to be sat on even if she relied on no one to tell her the true “nature” of sofas. 53. See my (1979), and (1989). 54. I make this point in my (1989).
References Akins, Kathleen. 1996. Of Sensory Systems and the “Aboutness” of Mental States. Journal of Philosophy 93: 337–372. Alanen, Lilli. 1986. On Descartes’ Argument for Dualism and the Distinction between Different Kinds of Dualism. In The Logic of Being, Knuuttila and Hintikka (eds.). Dordrecht: Reidel. Brown, Deborah. 1992. “Swampman of La Mancha,” and Other Tales about Representation. Doctoral Dissertation, University of Toronto. Burge, Tyler. 1973. Reference and Proper Names. Journal of Philosophy 70(14): 425–439. ———. 1977. Belief De Re. Journal of Philosophy 74: 338–362. ———. 1979. Individualism and the Mental. Midwest Studies in Philosophy 4: 73–121. ———. 1982. Other Bodies. In Thought and Object, A. Woodfield (ed.). Oxford: Oxford University Press. ———. 1983. Russell’s Problem and Intentional Identity. In Agent, Language, and the Structure of the World, Tomberlin (ed.), pp. 79–110. Indianapolis: Hackett. ———. 1986a. Intellectual Norms and Foundations of Mind. Journal of Philosophy 83(12): 697–720. ———. 1986b. Cartesian Error and the Objectivity of Perception. In Subject, Thought, and Context, McDowell and Pettit (eds.). Oxford: Oxford University Press. ———. 1986c. Individualism and Psychology. Philosophical Review 95: 3–45. ———. 1988. Individualism and Self-Knowledge. Journal of Philosophy 85(11): 649–663.
334
Tyler Burge
———. 1989. Wherein Is Language Social? In Reflections on Chomsky, A. George (ed.). Oxford: Blackwell. ———. 1990. Frege on Sense and Linguistic Meaning. In The Analytic Tradition, Bell and Cooper (eds.). Oxford: Blackwell. ———. 1991. Vision and Intentional Content. In John Searle and His Critics, pp. 195–213. Ernest Lepore and Robert Van Gulick (eds.). Oxford: Blackwell. ———. 1996. Our Entitlement to Self-Knowledge. Proceedings of the Aristotelian Society 96: 91–116. Descartes, René. 1964–1976. Oeuvres de Descartes. Charles Adam and Paul Tannery (eds.). Paris: Vrin. ———. 1971. The Principles of Philosophy. In Descartes: Philosophical Writings, G. E. M. Anscombe and P. T. Geach (eds.). Indianapolis: Bobbs-Merrill. ———. 1976. Descartes’ Conversations with Burman. Cottingham (ed.). Oxford: Clarendon Press. ———. 1981. Descartes’ Philosophical Letters. Kenny (trans.). Minneapolis: University of Minnesota Press. ———. 1988. The Philosophical Writings of Descartes. John Cottingham, Robert Stoothoff, and Dugald Murdoch (eds.). Cambridge: Cambridge University Press. Donnellan, Keith. 1966. Reference and Definite Descriptions. Philosophical Review 75: 281–304. ———. 1970. Proper Names and Identifying Descriptions. Synthese 21: 335–358. Evans, Gareth. 1973. The Causal Theory of Names. Proceedings of the Aristotelian Society 47 (suppl.). Reprinted in Evans’s Collected Papers. ———. 1982. The Varieties of Reference. Oxford: Oxford University Press. Gaukroger, Stephen (ed.). 1980. Descartes: Philosophy, Mathematics, and Physics. Sussex: The Harvester Press. Hacking, Ian. 1973. Leibniz and Descartes: Proof and Eternal Truths. Proceedings of the British Academy 59: 4–16. Kaplan, David. 1989. Afterthoughts. In Themes from Kaplan, J. Almog, J. Perry and H. Wettstein (eds.). Oxford: Oxford University Press. Kripke, Saul. 1980. Naming and Necessity. Cambridge, MA: Harvard University Press. Marr, David. 1982. Vision. San Francisco: W. H. Freeman. Michotte, A. 1963. The Perception of Causality. Miles (trans.). London: Methuen (Translation of 1946 French edition). Normore, Calvin. 1986. Meaning and Objective Being: Descartes and His Sources. In Essays on Descartes’ Meditations, Amelie Rorty (ed.). Berkeley: University of California Press. Putnam, Hilary. 1962. It Ain’t Necessarily So. The Journal of Philosophy 59. Reprinted in Putnam’s Philosophical Papers, volume 1. Cambridge: Cambridge University Press. Rozemond, Marleen. 1994. Evans on De Re Thought. Philosophia 22: 275–298. ———. 1998. Descartes’ Dualism. Cambridge, MA: Harvard University Press. Searle, John. 1983. Intentionality. Cambridge: Cambridge University Press. Sellars, Wilfrid. 1970. Presidential Address. Proceedings of the American Philosophical Association 44: 5–34. Shuster, John A. 1980. Descartes’ Mathesis Universalis 1619–28. In Gaukroger (1980). Sutherland, N. S. 1960. Visual Discrimination of Orientation by Octopus: Mirror Images. British Journal of Psychology 51: 9–18. Sutherland, N. S. and N. J. Mackintosh 1971. Mechanisms of Animal Discrimination Learning. New York: Academic.
14
Some Reflections on Scepticism: Reply to Stroud
Barry Stroud’s interesting paper raises fundamental issues about the relation between antiindividualism, principles of charity, and antisceptical arguments. These issues, particularly those concerning scepticism, are too large to make significant progress on here. But I can think of no honorable way to finesse Stroud’s challenge to use the occasion for what it is. So I try to navigate between timorousness and saying more than can reasonably said in so short a space. I Stroud gives an accurate statement of the general idea of my anti-individualism about perception: Most perceptual representations represent what, in some complex sense of ‘normally’, they normally are caused by and are applied to. Perceptual representations of something as a physical feature are, relative to a complex standard of normality, caused by and veridically apply to such physical features. This statement can be misleading, however. For the complexity in ‘normally’ is important and difficult to get right. As Stroud recognizes, an individual perceiver, under special circumstances, could be mostly wrong. The individual could be lifted out of the environment that its perceptual system was fashioned to connect with. Some of its perceptual categories could be innate, but triggered by nonstandard stimuli. Or they could be learned and (as a result of the displacement) triggered by abnormal but indistinguishable stimuli. Such an individual’s perceptual system would still have to have been fashioned—say, in evolution or through synthetic copying—through interaction with the types of things that its perceptions represent. The perceptual system might have interacted with the relevant types of things only when it was instantiated in other previous individuals. So some veridicality is implicit in the relevant conception of normality, but not in a way that guarantees that an individual has any veridical perceptions at all. Stroud quotes several times, somewhat misleadingly, my remark that we are “nearly immune from error in asserting the existence of instances of our perceptual kinds.” The remark does perhaps carry some broad epistemic reassurance. But this assurance should not be overrated. The remark is immediately followed in the same passage by an allowance of the possibility of individuals’ being under induced massive perceptual hallucination or having no regularity in interaction between themselves and their environment—say, in certain cases of being transferred into a brain in vat situation—so that their perceptual experiences are systematically false.1 I do think, however, that, with further qualifications that I will not discuss here, an individual could presume with near immunity from error that some individual or other has (at some time) perceived instances of some of the perceptual kinds or types that his perceptual system represents things as having.
336
Tyler Burge
Even laying this point about individuals aside, ‘normal’ cannot be glossed as ‘usually veridical’, at least not for each type of perceptual representation—even if one is considering a perceptual system’s representations rather than those of an individual perceiver. Some perceptual systems are fashioned in such a way as to produce a predominance of false positives for some perceptions. A rabbit’s system may misperceive objects as predators, even in its normal environment, more often than not. Or a whale’s perceptual system could in principle misperceive something as krill or as edible more often than not, although this is surely not the way whales actually operate. In being oriented to survival, the perceptual system may in certain cases scrimp on veridicality. But even in these cases, there must be a base of veridical applications interacting with some types of things being represented, if representation and error are to be possible. There are further complexities in the account of anti-individualist principles governing perception. But it does seem impossible that no perceptual representations as of the physical world ever are or have been veridical representations. This impossibility is itself a necessary truth. Stroud notes similarities between some of my views about perception and the principles of charity invoked by Quine and Davidson. He raises a question about what I meant when I wrote that they have sometimes used “with insufficient discrimination” the slogan that error presupposes a background of veridicality (Burge 1986, p. 131). Quine, as Davidson has noted, sometimes writes as if the slogan applies only to observation sentences and logical truths. It seems to me that it applies in more detailed ways than that. And with respect to observation, the qualifications indicated above have not been stated by either Quine or Davidson. Davidson sometimes writes as if the slogan can be glossed as ‘most of our beliefs are (or must be) true’. Because of problems about counting beliefs he has come to prefer ‘there is a presumption in favor of the truth of a belief that coheres with a significant mass of belief’. But in his actual arguments against scepticism, he relies, I think illegitimately, on the former, stronger formulations in terms of necessary veridicality (Davidson 1986). The problem with the first formulation—that most beliefs have to be true—is not just one about counting beliefs. As I have indicated above, the principles of antiindividualism by themselves, on any natural construal, do not allow us to know apriori that “most” of every individual’s beliefs must be true. It is possible for there to be massive systematic error on the part of an individual, even about the concurrent existence of putatively perceived objects. Further, the formulation (and Davidson’s supporting argument) in terms of coherence seems to me to gloss too generally over the specifics of how beliefs support one another. Some humanly common types of inferences are unreasonable. People are capable of wildly mistaken theories about quite a lot of things. Coherence with such bodies of belief confers no warrant. People are more likely to be right in some areas
Reply to Stroud
337
than in others. Davidson knows this, and has emphasized it on some occasions. But sometimes the points are glossed over in ways that matter to the argumentative context. More fundamentally, Davidson’s emphasis on coherence and his basic claim that warrant can derive only from beliefs seem to me to overlook that we are entitled to certain beliefs not because they are justified by other beliefs, or because they are self-justifying, but because of our competence with a representational (e.g., perceptual) system that bears certain types of connection to the world. The most salient type of entitlements are those attaching to perceptual beliefs.2 Neither sensations nor perceptions provide a justification, though they play a role in the account of the warrant—the entitlement. And coherence with other beliefs does not provide all the warrant, or even the primary warrant, for perceptual beliefs. In fact, I think that coherence plays very little role in the warrant for our more primitive perceptual beliefs. Rather, perceptual beliefs are often warranted through entitlements that do not involve justifications available to the individual at all. Not all warrant is justification. Not all warrant is inferential, or evidential, or derivative from within the individual’s system of belief. Most perceptual warrant is not. I do believe, however, that there is a prima facie, general warrant in favor of the veridicality of belief as well as perception. This warrant derives from the connections, first, between belief and rationality—or more broadly epistemic warrant—and, second, between rationality and truth. Certain extremely abstract and restricted elements from antiindividualism can play a role in the latter of these two connections. I further believe that, properly qualified to allow for such points as those about individuals, systems, and survival that I have just made, the contents of many of our empirical beliefs must apply to sorts of things that normally cause them. So, I think that individuals can be systematically mistaken in their perceptions and perceptual beliefs. But the perceptual representations of a perceptual system, as opposed to an individual’s actual perceptions, must have a core of veridical perceptions. I believe with Quine and common sense that beliefs causally and epistemically less directly tied to perception—more theoretical beliefs—are susceptible to error in more ways than beliefs more directly tied to perception. (We can understand ‘directly tied’ in terms of causal distance from sensation, in the learning and normal application of the beliefs.) I think with Davidson that there is a general prima facie presumption in favor of the truth of all beliefs. But I am not sure that I motivate my view in quite the way that he does his. Stroud is right that I do not envisage a distinction between perceptual and theoretical belief that makes the sceptical scenarios impossible for the former but not for the latter. Dramatic error is possible in both cases. And I think that a generalized prima facie warrant backs theoretical as well as perceptual belief because anti-individualist principles cover both, and because a presumption of rationality applies to any system of belief. Stroud’s Kantian (Critique of Pure Reason A226/B273) suggestion that perception is in principle
338
Tyler Burge
extendable ever further into empirical reality is a salutary reminder. But I do not think that it is the fundamental point. It seems likely, and certainly possible, that some empirically knowable entities are in principle (say, by physical law) imperceptible. The fundamental point is that principles about content formation and about rationality carry the argument that there is necessarily a base of veridicality in any intentional system. Principles about perception are important special cases. I believe that the general presumptive warrant in favor of one’s beliefs and perceptions is qualified, and overridden, in numerous ways that bear on the particularities of an individual’s forms of reasoning. And I believe with Stroud that this warrant is in principle (“metaphysically” or “logically”) compatible with any individual’s being massively mistaken about what the world around him is like. There are important differences between my, Quine’s, Davidson’s, and Stroud’s conceptions of the basis for the idea that error presupposes a core of veridicality that I shall return to. II Stroud raises some profound questions about the relation between anti-individualism and scepticism. I have supported a qualified version of the idea that error presupposes a background of veridicality. I have also endorsed a version of the idea that our perceptual systems’ representations must bear certain relations to types of entities causing veridical perceptions in at least some individuals. Implicit in these ideas is some seeming reassurance against scepticism about the physical world. Stroud asks whether these ideas provide an adequate reassurance against traditional scepticism about perception, and more generally about belief regarding the physical world. Of course, as Stroud is well aware, there are different types of scepticism. So the question is a schema for a number of different questions. But I give a schematic negative answer to all of them. There is a simple and uninteresting reason for this answer: Scepticism is about knowledge or justification. Neither antiindividualism nor the slogan that error presupposes veridicality says a word about knowledge or justification. So something more needs to be said even to speak to scepticism, much less answer it. Stroud anticipates this point, and later I will say more about what he does with it. But there is a further point in favor of a schematic negative answer. Much of the difficulty in answering scepticism lies in avoiding begging the question, or in deciding what legitimate questions the sceptic raises that should not be begged. Antiindividualism is a metaphysical view about the nature of certain mental states—what having such states necessarily presupposes. Merely asserting it, even with justification, does no more to answer scepticism than Moore’s asserting ‘Here is a hand’. Moore’s assertion is warranted and true. But his mere assertion hardly provides a non-question-begging
Reply to Stroud
339
answer to the worries that the sceptic is developing. Anti-individualism may be a component in, or a clue to, an answer to scepticism. But simply asserting it may beg the question. How is the assertion justified? Do we know it on any less question-begging grounds than we know ‘Of course there are physical objects. I’m looking at this hand full on in a good light’? Such claims might be made with full assurance. In fact, I think that it is a deep point, which Moore rightly played on, that they should be made with full assurance, even in the light of the sceptic’s challenge. But the interest in answering scepticism lies not in displaying or bolstering our assurance—which is surely sufficiently high even after hearing the sceptic out. It lies in explaining the nature of reason and epistemic warrant. The fact that anti-individualism is supposed to be asserting necessary truths may seem to differentiate it from Moore’s answer. But there remains the task of explaining the epistemic status of anti-individualism and showing that that status is not such as to beg some important and legitimate question that the sceptic is asking. Asserting anti-individualism may give clues, components, assurance. But it does not constitute an answer. What justification for our beliefs can be associated with the putative fact that our having attitudes with the relevant contents about the world necessarily derives from some interaction between individuals with attitudes and the world? Do we know this account of content-possession, and whatever claims we make about warrant for it, in a way that derives from assumptions whose justification is not brought as much into question by the sceptic as that of the straight-out claims about the physical world that we began with? The antisceptical interest of anti-individualism stems partly from the view that it might be known through apriori reflection. Traditionally the sceptic is supposed to concede that invocation of principles known from apriori reflection is legitimate in providing an answer. For it was assumed, at least on some traditional conceptions, that knowledge from reflection was knowledge backed by reason alone. And the scepticisms that I regard as interesting purport to be reasonable. But the precise epistemic status of anti-individualism—or of various aspects of it—is not obvious. It seems to begin by reflecting on the relation between mental states and a physical or social environment that is simply assumed to exist. Moreover, certain features of anti-individualism—such as our reflections on the role of evolution or the differences between individuals’ perceptions and perceptual systems—seem to owe something to empirical background knowledge. Emphatically, I do not assume that these points decisively undermine all antisceptical strategies that make use of anti-individualism. But the dangers of begging the sceptic’s question are too evident to glide over quickly and without comment. Both Davidson and Putnam provide stimulating responses to scepticism that are grounded in anti-individualist considerations. But neither so much as considers the
340
Tyler Burge
question of how the premises of his antisceptical argument are justified, much less whether the premises might beg the question.3 As an example of how Davidson’s argument threatens to beg the question in ways that he does not address, consider this remark: “Belief can be seen to be veridical by considering what determines the existence and contents of a belief. Belief . . . is supervenient on facts of various sorts, behavioral, neuro-physiological, biological and physical” (Davidson 1986, p. 314). The sceptic can reasonably ask what justifies this invocation of behavioral, neurophysiological, biological, and physical facts, and why it does not beg the question. I also think that the relations between reliable connection to the world and rational coherence of a belief system are not sufficiently distinguished in Davidson’s argument, opening it to Stroud’s charge that mere appeal to coherence can make the position look covertly idealist. An adequate response must, I think, consider those issues in some depth. I believe that anti-individualism can be a component in antisceptical initiatives. There are positive as well as negative things to be learned from the initiatives already proposed. But just invoking anti-individualism to show that our having beliefs about the physical world presupposes interaction with the physical world—even if it is on the right track—is too quick a response to provide lasting illumination on the issues raised by scepticism. The key issue with scepticism is to understand what are legitimate starting points for reason. III Transcendental arguments, understood as arguments from the conditions necessary for aspects of our cognitive or mental lives, can provide at most general responses to the most general scepticisms. The general response might be at the level of “we know that there (tenseless) are and have been physical objects.” I think that non-question-begging responses of this sort are imminently worth having and, for all their elusiveness, not beyond getting. But the fact, noted above, that anti-individualism about perception allows that an individual could be massively mistaken about his immediate environment raises obvious difficulties with expecting too much of anti-individualism in dealing with all the traditional scepticisms. I think that anti-individualism can contribute to general antisceptical responses, and in more limited ways, to more specific responses. But I agree with Stroud that no argument from anti-individualistic premises alone can provide a worthwhile response to the sceptic. As I indicated, I think that general antisceptical arguments are worth striving for. Providing a powerful general antisceptical transcendental argument would be a major philosophical achievement. But Stroud is right in conjecturing that I would not be satisfied (although I would be thrilled) by an adequate general antisceptical response—a guaran-
Reply to Stroud
341
tee of the reliability of perceptual systems in their representations of physical objects, or of the mere existence of physical objects. I do not go so far as to hope that we could prove reliability of perception or the existence of physical objects. But I do think it philosophically worthwhile and possible to show with some rational force, against the most interesting forms of scepticism, that we have such knowledge. I also think that we can show that we know some particular specified propositions about our own environment. For example, I think that it is possible to provide an antisceptical argument in defense of the claim that there is a hand here (looking at one in a good light). I think it is worthwhile and possible to argue rationally that no demon is systematically fooling us and that we are not brains in vats. Such arguments would have to allow that individual claims to knowledge, even reasonable claims, are fallible, for ordinary, nonsceptical reasons. That is one reason I am leery of the language of proof. I think that we can infer the negative propositions about demons and vats from our ordinary perceptual knowledge claims. Stroud correctly attributes this view to me. I do not purport to have provided philosophical arguments for the view; I only assert optimism that it can be done. The chief difficulties are supporting the inferential principles and supporting those ordinary claims themselves in a way that would be illuminating in the light of the standard sceptical challenges. There is no easy argument, either here or in the more general case. I hope to pursue these matters, despite Stroud’s evident belief, certainly backed by considerable apparent inductive support, that such pursuit is quixotic.4 There are important differences in the ways Stroud and I see the prospects of dealing with scepticism. Some of these differences seem to derive from differences in the ways we view the relation between anti-individualism and the charity principles that Quine and Davidson have brought to prominence. Stroud states: “Davidson holds on antiindividualist grounds that most of our beliefs are true, or that they are by and large or for the most part true. The anti-individualist grounds he relies on are to be found in the conditions of what he calls ‘interpretation’—one person’s understanding and communicating with another. . . .” He continues, “Burge’s anti-individualism . . . is also a theory of our practices of attributing psychological states and attitudes with determinate contents to people in a world we inhabit. The thought experiments he appeals to to support his antiindividualism turn on and exploit our capacities of belief- and attitude-attribution as they actually are.” These remarks may reveal a large difference between my view and Stroud’s. I do not accept the idea that anti-individualism has its grounds in interpretation, or that antiindividualism is itself a theory about our practices of attitude attribution. Rather, I think that our practices of attribution have their ground (partly) in anti-individualism. Antiindividualism is not about those practices—though it can be loosely seen as implying or
342
Tyler Burge
requiring conditions on a theory about them. Anti-individualism is about the nature of certain types of belief and other propositional attitudes, and about the nature of concept possession. Of course, our views about belief are partly informed by our practices of attribution of belief to others. There is no way to avoid being in some respects influenced by the ways we attribute beliefs (our own or others’) that we actually have. But antiindividualism is not a second-level theory about our views about belief or about our ways of detecting belief—much less about the language of belief-attribution (as some, not Stroud, have claimed). It is a theory about belief and about possession of intentional or representational content. I see anti-individualism as no more about our ways of attributing or detecting belief than mathematics is about the ways we think about or find out about the numbers and quantities. The view that we detect beliefs about the physical world by considering their physical causes does not by itself entail anti-individualism. It lacks the needed premise that our modes of detection are correct. And it lacks the modality associated with antiindividualism: Relevant beliefs about the environment are by their nature necessarily related to the environment. These differences entail a difference in the way I approach the problem of scepticism. For someone who starts with a theory about our practices of attitude attribution, one must defend some non-question-begging principle that our practices are correct and rational. It is usually not noticed that one must also show that we know what our practices are in a way that itself does not beg the question against the sceptic. As Stroud has explained in his wonderful classic work on transcendental arguments, such defenses frequently fall into verificationism or idealism (Stroud 1968). By contrast, my problem is to justify in a non-question-begging way my claim to know something about the nature and individuating conditions of certain beliefs. The problem is still difficult. But it does not involve a step from putative knowledge about our practices of attribution—or about beliefs about belief, or about detecting belief—to knowledge about belief itself.5 I believe that the theory of interpretation is grounded partly in anti-individualism and partly in a (broader) theory of reason, rationality, and epistemic warrant. We can get clues to both of these latter, more basic theories from reflecting on our practices of linguistic and psychological interpretation. But this is an order of discovery, not an order of justification. I think that the more metalevel theory of interpretation, insofar as it has any claims to be apriori or necessary—as opposed to a sociological theory of our actual practices— is grounded in prior principles about attitudes, epistemic warrant, and reason. Anti-individualism and principles about reason and epistemic warrant can contribute in different ways to addressing the antisceptical problem. I think that one can defend a rational presumption that belief generally can be rationally presumed to be prima facie veridi-
Reply to Stroud
343
cal. On this matter, I side with Davidson against Stroud. But I would motivate such a presumption not merely from very abstract features of anti-individualism. (If I understand Davidson correctly he motivates the view from an account of our practices of interpretation.) I would motivate it also from a theory of epistemic warrant and reason. The bearing of such an argument on particular beliefs—say, about the existence of physical objects— would require supplementation. Anti-individualism can be used to create some rational presumption that certain particular types of beliefs could not be held if the individual who had them bore no appropriate (perhaps indirect) relation to some relevant physical environment. And from this one can derive some limited conclusions about veridicality. In this domain, Stroud seems to me right in insisting that the argument for veridicality derive from what a person believes, not from the nature of belief in general. Anti-individualism has primarily to do with how having particular types of intentional contents depends on relations to an environment beyond the mind or skin of the individual. A full antisceptical theory must separate the different strands—considerations about reason and epistemic warrant from antiindividualist considerations. IV I would like to make some brief remarks about Stroud’s own approach to scepticism. He claims that we do not have to “prove” scepticism to be false in order to oppose it on antiindividualistic grounds. I have indicated some unease about this conception of proof. Perhaps Stroud and I share the unease. But I think that traditional positive responses to scepticism need not be construed in terms of proof. I think that one might refute scepticism by showing in a non-question-begging way that it is, all things considered, unreasonable to doubt or suspend belief about (say) the existence of the physical world. Such a refutation need not portray its reasons, much less those associated with ordinary beliefs, as infallible or incontrovertible. It is sufficient that they be, and show themselves to be, substantially more reasonable than the sceptic’s putative reasons to doubt. Or it is sufficient to show that there is something unreasonable about the sceptic’s putative reasons to doubt in light of the nature of certain reasons for believing. I agree with Stroud that we should reject the “stronger thesis” that it is “logically” necessary that most beliefs and perceptions are veridical—or that any “large and reasonably comprehensive set of beliefs” is for the most part veridical. But I think that what he calls the “weaker thesis” also deserves scrutiny. The weaker thesis is that anyone who thinks that people have determinate beliefs, and thinks so on anti-individualist grounds, will regard those beliefs as for the most part true; for the interpreter will not see the people he
344
Tyler Burge
interprets as holding the beliefs he attributes unless he sees them as sharing his beliefs to a large extent, hence as having mostly true beliefs. I doubt this weaker thesis, as stated. An interpreter can, I think, find another person wildly mistaken, for example, if the other person has been transferred from a normal environment into a new abnormal environment (for example, a controlled vat). So anti-individualism does not block one from finding another person to be in at least some of the sceptic’s scenarios for at least some periods of time. Sharing a single world and using that world to interpret others does not block one from attributing wildly mistaken beliefs, because the environment by reference to which the preponderance of the belief-contents were fixed can change or be destroyed. So even if the weaker thesis could be shown to rest on non-question-begging grounds, I do not believe that it can be relied on to deal with all sceptical scenarios. There remains a yet weaker thesis that might be proposed from Stroud’s point of view: Our considering the particular set of perceptions and beliefs that we and others have guarantees that we find, in accordance with anti-individualism, that some of them are grounded in relations to an environment—even if that environment is no longer the one in which we are situated. This may yield the result that we must find veridical a limited set of beliefs about the (tenseless or past tense) existence of kinds indicated by some of our intentional contents. This thesis is the meta-analogue of the “general” response against scepticism that we considered before. It is not the same, for it centers not on the necessary veridicality of some of our beliefs but on our necessarily finding some of our beliefs veridical. I am not hostile to this weaker thesis, modulo certain questions about ‘must find’. But I think that it too needs further argument to engage the sceptic. Wherein is anti-individualism justified? Wherein are we warranted in determining what perceptions and beliefs we have? Do such justification and warrant not beg the question against the sceptic? If these questions are answered satisfactorily, then one has shown that our practices are reasonable, and one has answered the sceptic. If they are not addressed, then it seems to me that the yet weaker thesis remains under the suspicion that it begs the sceptic’s question. These points do not touch the first-person version of the weaker thesis, and the analogy to Moore’s paradox. No one of us can consistently believe that the beliefs he or she now has are false. But this point applies to irrational as well as rational beliefs, and to revisable beliefs (such as that it is raining) as well as intractable ones. And we can certainly find later that beliefs we once had are wildly false or thoroughly irrational. So this analogy alone might offer us no reassurance in response to the sceptic. Stroud seems, however, to envisage that the anti-sceptical theses that he favors are not only beyond consistent rejection, in the Moore sense, but are things that we necessarily believe, once we have concepts of belief at all. I think that this is an interesting strategy. It links up with the ideas that I discussed in the previous paragraph.
Reply to Stroud
345
I suspect, however, that the strategy is unstable.6 If we can get that far, we will have to go further. I do not think that we should restrict ourselves to responses to scepticism that maintain that necessarily we find it incredible. I think that it is possible and worthwhile to show in non-question-begging ways that it is reasonable to believe what the sceptic doubts—reasonable to believe that there is a physical world, for example. As noted, my basic starting point, as well as my aimed-for conclusion, differs from Stroud’s. I start not with a theory of attribution of belief, but with a theory of belief—particularly certain types of belief. And I aim for the conclusion that it is unreasonable (not only locally incredible) to suspend belief about the physical world. But both strategies must answer the questions asked two paragraphs back. The issue about begging the question is, I think, the central one in discussions of scepticism. For it goes to the issue of the legitimate starting points for reason and epistemic warrant. One can beg the sceptic’s question or ignore it. Those are perhaps reasonable procedures, though they do not seem to me to be philosophically illuminating procedures. How much philosophical, rational reassurance should this last yet weaker thesis yield, apart from the epistemic investigations that I have recommended? I think that this is an interesting and subtle question. One can perhaps with perfectly reasonable assurance ignore the sceptic’s question. But in light of the sceptic’s question, does the yet weaker thesis provide philosophically illuminating rational reassurance? The claim that we cannot do otherwise than to find some of our beliefs true, or some carefully hedged existential commitments veridical, may seem not to count for much reassurance. The sceptic can heartily agree with the claim and say that that just brings out our pitiful, irrational plight. It seems to me that reassurance may require some account of wherein there is reason in the necessity of finding the world to answer to at least some of our, and others’, beliefs. And if we give such an account, it seems that we must once again confront whether we are begging some legitimate question that the sceptic is asking. I would like to know what Stroud thinks about this. In any case, my own strategy in dealing with the sceptic through reflections on antiindividualism does not go through the route of a claim about the necessity of our following our practices of belief-attribution. I concentrate directly on the implications of having certain beliefs. Moreover, I hope for something more nearly like what Stroud calls a direct refutation of the sceptic. I do not think that it has been shown that we cannot explain, in light of the sceptic’s question, reasonably and in a non-question-begging way, wherein we have knowledge and even wherein we can know we have knowledge. This route may not be any easier than the traditional routes that Stroud has convincingly criticized. But I believe that it is different.
346
Tyler Burge
Notes 1. See my (1986), p. 131. 2. For discussions of entitlement, see Burge (1993), (1995), (1997), (1999), and Perceptual Entitlement, forthcoming in Philosophy and Phenomenological Research. 3. See Davidson (1986); and Putnam (1991), chapter 1. 4. My Locke Lectures 1993 at Oxford were devoted to this topic. Since the issues are so difficult, I have published some of those lectures only with the antiskeptical objectives deleted. I hope eventually to provide a fuller treatment of the subject. 5. As I note in my reply to Loar, I did use one argument for anti-individualism from our practices of beliefattribution. But it was never my primary or only argument. The primary argument is from judgments, based on cases, about conditions under which it is metaphysically possible or impossible to have certain beliefs. The conclusion of the arguments was never primarily about our practices but always about the nature of belief. This point is quite explicit even in Burge (1979). See also Hahn’s contribution to this volume, and my reply. 6. See Brueckner (1996).
References Brueckner, Anthony. 1996. Modest Transcendental Arguments. Philosophical Perspectives. 10:265–280. Burge, Tyler. 1979. Individualism and the Mental. Midwest Studies in Philosophy 4: 73–121. ———. 1986. Cartesian Error and the Objectivity of Perception. In Subject, Thought, and Context, McDowell and Pettit (eds.). New York: Oxford University Press. ———. 1993. Content Preservation. Philosophical Review 102: 457–488. ———. 1995. Reply: Intentional Properties and Causation. In Philosophy of Psychology: Debates on Psychological Explanation, Cynthia and Graham MacDonald (eds.), pp. 226–235. Oxford: Blackwell. ———. 1997. Interlocution, Perception, and Memory. Philosophical Studies 86: 21–47. ———. 1999. Comprehension and Interpretation. In The Philosophy of Donald Davidson, Lewis Edwin Hahn (ed.). Chicago: Open Court. Davidson, Donald. 1986. A Coherence Theory of Truth and Knowledge. In Truth and Interpretation, E. Lepore (ed.). Oxford: Blackwell. Putnam, Hilary. 1991. Reason, Truth, and History. Cambridge: Cambridge University Press. Stroud, Barry. 1968. Transcendental Arguments. Journal of Philosophy 65: 241–256.
15
Davidson and Forms of Anti-Individualism: Reply to Hahn
Martin Hahn’s essay is a fresh and illuminating discussion of differences between my views and those of Donald Davidson on issues associated with anti-individualism. I think that Hahn is right to emphasize differences, and right about where most of the fundamental differences lie. In some places I cannot answer questions he raises about Davidson’s views. I will try mainly to answer for my own. First, I want to make some general points about anti-individualism. I Hahn is right to see the relevance of “Belief De Re” to my arguments for anti-individualism (Burge 1977, pp. 338–362). That article is the main conceptual forerunner in my work to “Individualism and the Mental” (1979) and “Other Bodies”(1982). Hahn is also right that I do not argue for or even advocate anti-individualism in “Belief De Re.” There are two theses in that article that are relevant to the development of antiindividualism. One is the thesis that having de re attitudes is necessary for having any propositional attitudes at all. Hence they are necessary for having de dicto attitudes.1 This thesis hinges on the claim that for a creature to have propositional attitudes, with their attendent intentionality, it (or, I would now add, some progenitor) must engage in activity or enter into relations that connect those attitudes with their subject matter. This activity and these relations must not be merely relations of conceiving of the subject matter. Some must be nonconceptual, and some must even be nonintentional. Demonstrative applications of intentional types to individual objects are the primary sort of intentional but nonconceptual relations. Causal relations in ordinary perception are paradigmatic of relevant nonintentional relations in actual cognitive activity. But I indicated that they are not the only relevant relations. I avoided claiming that ordinary perception of objects in the physical world is metaphysically necessary for intentionality. The key idea of this thesis is that the intentionality of all of an individual’s conceptual states depends on further relations, including nonconceptual and even nonintentional relations, between some of the individual’s states, on one hand, and entities that are referential objects of the intentional states, on the other. These relations underlie and make possible de re attitudes. De re attitudes make possible de dicto attitudes.2 This is a completely general thesis that makes no commitment to what the subject matter of the de re attitudes has to be. The thesis makes no special mention of the environment beyond the individual’s body.3 Thus Hahn overstates what I argued when he writes that for me “De Re beliefs are irreducibly relational (acts of application of a conceptual content to an object in the environment)” (p. 34). I think that they are, in a certain sense, irreducibly relational. I think that
348
Tyler Burge
perceptual relations to objects in an environment are paradigmatic in the actual circumstances in which human beings operate. But I did not maintain that de re attitudes necessarily involve relations to an environment beyond the individual, nor therefore that having propositional attitudes necessarily requires bearing de re relations to such an environment. Thus Hahn is somewhat misleading in his remarks on p. 52 of this volume. These remarks suggest that his explanation as to why “Belief De Re” does not involve commitment to anti-individualism is that anti-individualism is a thesis about only our de dicto attitudes. On this interpretation, a theory about de re attitudes would not even be relevant to antiindividualism. But my anti-individualism is a thesis about the individuation of both de re and de dicto attitudes. The element of truth in Hahn’s remark is that I wanted to emphasize in “Individualism and the Mental” and “Other Bodies” two closely related points: First, I wanted to emphasize that anti-individualism is not merely the thesis that the environmental referents of de re attitudes play a role in individuating the attitudes considered as relations to the environment. I regarded this thesis, though in a certain sense anti-individualist, as almost trivial. The thesis avoids triviality inasmuch as it presupposes that de re attitudes cannot be reduced to de dicto attitudes. But taking the relevant de re attitudes to be relations is, I think (and thought), trading on a somewhat contrived conception of propositional attitude. I think that the attitudes themselves (as opposed to our attribution of them, or relations between them and the environment) are states of the individual completely fixed by their intentional content, including the nonconceptual, applicational (or token-demonstrative) elements in the content. The attitudes bear relations to an environment but are not themselves relations to an environment.4 These relations are not reducible to de dicto conceiving-of relations. The attitudes’ being de re lies in their being in certain referential relations through the demonstrative applications. The de re attitude itself is a mental state involving a nonconceptual demonstrative application token, or abstraction from a token. It is not itself a relation that includes the object. It is important to take the relation seriously. It enters into various psychological explanations. The state must, however, be distinguished from the relation for other psychological purposes and for many of the purposes of epistemology and theory of rationality. The de re relation is a success relation. It is to the incompletely conceptualized, demonstrative-involving mental state as veridical perception is to the perceptual state. Second, I wanted to emphasize that anti-individualism applies to the conceptual elements in propositional attitudes, whether these be de re attitudes or de dicto attitudes. It was a much more substantial thesis that conceptual, standing, intentional aspects of the attitudes are nonindividualistically individuated than that the token-demonstrative elements are. (The associated but even weaker thesis that de re attitudes that are about objects
Reply to Hahn
349
in the environment involve, insofar as they are de re, relations to the environment is virtually trivial.) So I highlighted the conceptual “de dicto” elements in attitudes, whether or not the whole attitude was de re or de dicto. Thus, the reason that “Belief De Re” does not argue for or advocate anti-individualism is not because it makes a claim about de re attitudes, whereas anti-individualism is a thesis about only de dicto attitudes. The reason is that “Belief De Re” does not claim that any propositional attitudes are essentially individuated in terms of relations to an environment beyond the individual. It does not single out de re attitudes about the environment. And it does not specifically claim that the intentional components of mental states that are in fact de re are ever individuated in terms of relations to the environment. The relations that “Belief De Re” requires de re attitudes to have are relations between the attitudes and (usually nonintentional) objects of those attitudes—where no requirements are placed on the nature of the objects. When I wrote “Belief De Re,” I did believe that mental states that involve de re applications to objects in the environment are partly individuated in terms of nonintentional relations, paradigmatically, causal-perceptual relations, to those objects. These relations support and help individuate the intentional but nonconceptual demonstrative applications. Applications to different actual objects count as different applications, whereas different application events that are anaphoric or otherwise tied together can count as the same application for purposes of logical form. This view is anti-individualist with respect to the de re intentional aspects of the de re attitudes—the demonstrative-like applications. Although this view is anti-individualist, it was not asserted or argued for in the article, though certain claims in the article do presuppose or entail it. The view is at most a relatively small and noncentral aspect of anti-individualism. Anti-individualism is most interesting in its claim about the individuation of conceptual elements in a wide range of propositional attitudes, both de re and de dicto attitudes. I did not make such a claim until “Individualism and the Mental.” Whether the thesis of “Belief De Re” that all attitudes presuppose de re attitudes can be strengthened so as to require de re attitudes about the environment is an extremely difficult issue. The issue is close to those that have exercised philosophers at least since Kant’s “Refutation of Idealism” (Kant 1787 B274–279). Some philosophers, following Kant, have maintained that to have propositional attitudes about sensations and about mathematics, one must have attitudes about physical objects in the environment. I find this thesis plausible. I think that it is almost surely a deep truth about the psychology of human beings. But I am not satisfied with apriori arguments for it as a necessary truth. I find neither Kant’s “Refutation of Idealism” nor private language arguments with this sort of conclusion compelling as they stand. I have so far not myself argued for the view as an apriori, necessary truth.
350
Tyler Burge
The second thesis in “Belief De Re” relevant to anti-individualism is that de re attitudes are attitudes that are incompletely conceptualized. The attitudes essentially involve n onconceptual intentional elements, principally token-applications of demonstrative or indexical elements attached to concepts. They are associated with and supported by nonintentional relations between the individual and the objects referred to in de re attitudes. The thesis about incomplete conceptualization was a development of the points about reference brilliantly initiated by Kripke and Donnellan. The development consisted in associating those points with demonstrative elements in the logical form of thoughts, and in elaborating their points about linguistic reference in the context of individuating propositional attitudes. These two theses of “Belief De Re” help explain, as Hahn notes, anti-individualism. It is natural to see the nature of empirical beliefs about objects in the environment as depending on de re beliefs about objects, not necessarily the same objects, in the environment. This is a specialization of the first thesis of “Belief De Re.” But the arguments for it are not to be found in that article. It is also natural to see the nonconceptual and even intentional relations between thinkings and referents of their representations, which are required by the second thesis, as helping to explain the specific content of conceptual elements in propositional attitudes. But again the arguments for this view are developed only later. Finally, the idea of the second thesis that a person can think and make reference to individual objects with incomplete mastery of conceptual methods that individuate the objects suggests the further point, developed only later, that a person can think with concepts that he incompletely understands. A person incompletely understands a concept he uses, in the relevant sense, if he cannot explicate it sufficiently thoroughly and accurately to fix, through those explications, essential features of the concept’s applications to kinds or properties. II Hahn several times characterizes my anti-individualistic view as the claim that for a large class of propositional attitudes, the intentional contents of the attitudes, hence the attitudes themselves, are in part individuated by the nature of their referents or extensions. I want to enter two qualifications on this characterization. First, anti-individualism is, as Hahn is aware, a thesis not primarily about concepts but about intentional mental states. It is about the constitutive or individuating conditions of propositional attitudes, perceptual states, the having of concepts, the applications of demonstratives, and so on. Anti-individualism is relevant to the nature of concepts themselves, but the arguments for anti-individualism all have to do with the individuation of
Reply to Hahn
351
mental states, which are typed partly in terms of concepts. The view is intended to be compatible with a variety of accounts of the nature and ontology of concepts.5 When one says that the view applies to lots of different kinds of empirical concepts, this should be taken as shorthand for saying that it applies to lots of different kinds of mental states typed, of course, by their intentional contents, including concepts. Second, the remark that attitudes are individuated in part by the nature of their referents or extensions needs elaboration if it is not to be misleading. For most of the contents of attitudes I discuss, the concepts logically determine their extensions or referents, in the sense that a relevant given concept has one and only one extension or referent. Given the actual world, the concept water applies to one and only one extension, all the water (both small and large portions) that there is. Moreover, the concept water applies only to water in any possible circumstances. In my arguments I sometimes use this point by showing that under certain different circumstances, certain persons have different referents for the concepts in their attitudes, and hence have different attitudes. But the point that if referents are different for nonindexical intentional contents of propositional attitudes, the attitudes are different (and thus that referents play a role in the individuation of attitudes), is not itself an anti-individualist point. It is purely a “logical” point about intentional contents and reference. Anti-individualism involves two further commitments. One is that a condition on the individuation of relevant intentional states is that the individuation presupposes a background of referential success. A representational state could not be what it is if it were not associated with the representational success of some relevantly associated representational states. That is, individuation of a representational state presupposes successful applications of relevantly associated intentional contents, or veridical representations containing relevantly associated contents. The associated contents are commonly the contents of the state, but they need not be. The possible sorts of association are various. This point about the role of successful reference and veridicality as a condition on individuation is a development of the adage that error presupposes a background of success. The second further commitment of anti-individualism is that attitudes are in part individuated in terms of nonintentional relations—or relations partly individuated independently of reference—that the individual bears to objects, properties, or relations in the environment. For individual attitudes, the account specifies particular kinds of entities in the environment. The nature of the nonindexical aspect of the attitude determines—in the sense explained above—the extension of that aspect of the attitude. So the attitude’s intentional, referential relation to the extension is not more basic than the attitude itself. The two go lock-step together. Certain objects, properties, relations bear a certain complex of nonintentional relations to the individual. Partly by virtue of standing in those relations, the
352
Tyler Burge
individual has attitudes with intentional contents that can include some of those entities in the attitudes’ extensions. Anti-individualism is a view about how mental states are constitutively individuated. The individuation helps account both for what extensions the attitudes have and for what their intentional contents are. So one cannot appeal to the extension of the attitude as such in explaining how—or in citing a constraint on how— the attitude is individuated even though successful reference is a condition on any intentionality. The relations that the account appeals to are nonintentional relations between the individual and entities in the environment. These relations help (individuatively) establish relevant entities as falling in the extensions or being referents of the attitudes. Since relevant aspects of the attitudes determine extension and reference, the relations constrain the nature of the mental states. I believe that the nonintentional relations that figure in the individuation of many intentional mental states cannot be specified in such a way as to produce a reductive explanation of their intentionality. Rather the presence of such relations is a metaphysically necessary constraint on intentionality and on the contents of particular sorts of attitudes—at least for reference and purported reference to objects in the environment cognized by empirical means. Because of the enormously complex variety of ways in which intentional content and reference can be constrained by the individual’s relations to an environment, I have tried to stay open and flexible in specification of the relevant relations. I do not, however, think that just any relations will do. The environment’s being a certain way must somehow connect to the individual’s being a certain way. An individual thinker’s mere presence in an environment has little individuative relevance. I believe that most relations to future events are of little relevance.6 It is clear that causal relations to past and present events play a large and central role. However, the ways that causal relations play a role in individuation are extremely complex. Even in the individuation of perceptual states, the relations can be very indirect. For example, relevant causal relations can be between objects and ancestors of an individual’s perceptual system, far back in the evolution of the system, rather than within the individual’s learning experience. The relations are compatible, in some instances, with a predominance of “false positives.” For example, the perceptual signal predator there may be caused, even in evolutionary history, more often by non-predators than by predators (see Burge 1996, note 11). Causal relations can go through other individuals in acquisition of concepts through interlocution. They can be mediated by empirical theory. They can be supported by imaginative projection from perceived instances of other kinds. Hahn is certainly right (p. 51) that I do not hold that beliefs about cows must be the normal causal products of interaction with cows. Someone could have beliefs about cows without ever having interacted with one—if, for example, he or she constructed a correct
Reply to Hahn
353
biological theory of cows and imagined what they would look like.7 Such a theory would, of course, contain other notions that had more direct causal relations to their referents. My reticence about talk of causal relations in discussions of anti-individualism has been motivated by an appreciation of the complexity of the matter and a determination to avoid reductionism. The exact details of relevant nonintentional relations interest me less than the broad categories of relations that play roles in constraining the individuation of various types of attitudes. My reticence does not stem from any doubt that causal relations are among—even central to—the nonintentional relations that connect the individual to an environment in such a way as to help individuate the individual’s attitudes. I also want to emphasize that relations to the environment are not the whole story about the individuation of attitudes or of meaning. Internal relations among attitudes—for example, inferential relations—also play a large role. I think that that role had been overemphasized in theories of meaning in the first two-thirds of the twentieth century.8 That is not to say that mental content can be individuated purely in terms of relations to an environment. Having any belief requires mastering some inferential relationships involving it. III Hahn raises some interesting issues about relations between Davidson’s views and mine. I agree with Hahn in being unaware of a clear statement of an anti-individualist thesis in Davidson’s writings prior to the 1980s. I also agree with Hahn that Davidson’s later advocacy of the view derives from premises that I do not share and that have, I think, less appeal than the basis on which I (and with qualifications, Putnam) argued for the view. In particular, Davidson’s important and sound points about the role of distal stimulants as evidence for interpretation do not nearly suffice by themselves to establish antiindividualism.9 Unless these points are conjoined with his interpretationism or with other strong premises, they do not yield any conclusion about the nature or constitution of propositional attitudes. The interpretationism—the idea that propositional attitudes like belief are constitutively dependent on relations to an interpreter—seems to me quite implausible. I do not find Davidson’s arguments for this view compelling or even attractive.10 It is certainly a relatively special philosophical position. There are many aspects of the view and its application, some of which Hahn probes, that I do not understand. I do believe that Davidson gives the empirical world an important role in his account of interpretation; and I am not sure that Hahn gives this role sufficient weight in his discussion. I leave these matters to others, however. Whether and in what sense Davidson sees the kinds in the empirical world as dependent on us and our practices is another matter I am not clear about. I would certainly
354
Tyler Burge
oppose any such generalization about empirical kinds. As Hahn emphasizes, I do not believe that our activities of “carving” the world into similarity classes (see Hahn’s note 10) play any substantial role in constituting empirical natural kinds. Our language and concepts develop in ways that take account of differences and similarities we come to appreciate in the world. The idealism suggested by the carving metaphor, or by interpretationist accounts of natural kinds, plays no role in the sort of anti-individualism that I maintain. I agree with Hahn that such a view would, at the very least, attenuate the meaning of anti-individualism (or externalism) about the role of the physical world in individuating mental states. Let us turn to Davidson’s Swampman case. The case involves many complexities and cannot be covered in a few words. I will not discuss whether Swampman could have thoughts at all. I will confine myself to the question of whether it could have thoughts about its physical environment. I stipulate that my discussion concerns Swampman at the moment that its synthesis is completed, but before it interacts with its environment. I believe that once Swampman interacts with objects in the world (through stimulation of sense organs or through operating on those objects), there begin to be the sorts of relations that support intentional content.11 Hahn is strictly speaking correct that my anti-individualism does not entail a position on the case. To claim that many mental states are individuated by relations between the individual and the environment is not by itself to say that Swampman lacks mental states, or even lacks any particular mental states. On the other hand, as Hahn points out (in his note 11), it is hard to understand how a creature like Swampman (at least before it starts interacting with its environment) could have intentional content about that environment. The ‘could’ here is, in my view, not merely causal. It is metaphysical. Anti-individualism suggests a range of ways that an individual can have intentional content that makes reference to an environment. It suggests that intentional content is supported and made possible by a range of supplementary nonintentional relations between the individual and the environment.12 Swampman seems to be in no relevant relations. Merely happening to be in an environment where it will chug along does not suffice to explain individuation of its (putative) intentional mental states. My version of anti-individualism is open-ended about what could count as relevant relations. Still, it seems to me that Swampman almost surely lacks any relevant relation. So I think that there is reason to deny that it is metaphysically possible that Swampman could share any of its counterpart’s thoughts about the environment, at least not until it starts interacting with the environment. Here I am inclined to agree with Davidson. I believe that the range of possible things that Swampman’s supposed thoughts could be about, correlative to the range of possible environments that his physical dispositions could be adaptive to, is huge. Apart from the variety of natural-kind environments, there
Reply to Hahn
355
are indefinitely many possible envatted environments and environments involving laboratory induced variances on the source of the same proximal stimulations. It is probably so huge and varied that there is no relevant determinate intentional content about an environment associated with those dispositions and physico-chemical structures. The mere fact that Swampman is synthesized by quantum accident in a certain environment is not sufficient to provide it with the sorts of relations to the environment that would make thoughts about that environment possible. Nor do I think that Swampman’s future relations to the environment suffice—at the time of the synthesis, before there are causal interactions with an environment that might be referred to.13 I see no other constraints that would associate Swampman’s physical dispositions and chemical structures with referents in any one environment. If all this is correct, there is nothing in Swampman’s physico-chemical structures and dispositions, taken in themselves, that makes them referentially relevant to the properties of the environment they are in, rather than to any number of other properties in other possible environments.14 So I believe that Hahn somewhat understates the relevance of my anti-individualism to the case. Although my central arguments do not deal with the case and do not entail a commitment on it, the natural lessons to be drawn from the arguments suggest, I think, that Swampman lacks any thoughts about its environment. Hahn is certainly right that my views stem from reflection on constraints on the having of propositional attitudes, not from reflection on the nature of interpretation. Hahn is also right in conjecturing that learning plays no essential role in my view, although learning does play an important role in the actual determination of much of our intentional content. I think that there is no necessity that a thinker learn a language, or even have a language. There is no metaphysical impossibility in the idea that an individual’s language and repertoire of beliefs could be entirely innate. An individual might have innate concepts or percepts about the environment. Some of our intentional content is innate. And I think it is an empirical fact, well established since the demise of behaviorism in psychology, that animals have propositional attitudes, but lack language. However, even innate intentional content must be supported by relations other than referential relations to the subject matter. Swampman seems to lack any relevant supporting relations (see note 2). So Swampman at the time of its synthetic birth seems to lack relations that would allow any of its mental states to be about the specific environment that it happens to coalesce in. I think it certain that Swampman cannot share all of its chemical counterpart’s empirical thoughts about the environment. Assuming that Swampman’s chemical counterpart is not scientifically omniscient, there are some aspects of our environment that we can already, with certainty, know Swampman’s physical structures are not necessarily uniquely correlated with—natural kinds, for example.
356
Tyler Burge
I want to caution against one line of interpretation in Hahn’s discussion of the case. My anti-individualist arguments are not supposed to be applicable merely to beings “similar to us.” There is no such parameter in the arguments. They are meant to provide insight into conditions for individuating certain sorts of mental states under certain conditions of limited background knowledge, full stop. Swampman is not exempt from necessary conditions on individuation by virtue of its being so unlike us. Mental states are not special to human beings. What matter are the specific ways that Swampman is unlike us in regard to relations relevant to individuating mental states. The problem in dealing with Swampman is that the anti-individualist arguments are example-based and are somewhat open-ended about the most general anti-individualist principles of individuation. For Swampman to have thoughts about its environment, there must be nonintentional relations that help make such reference possible. I see no likely candidate relations. Still, there is no infallible guarantee that one has not overlooked some important relation that would connect Swampman to specific aspects of its environment so as to make intentional reference to those aspects comprehensible and nonmagical. IV Hahn emphasizes similarities in my and Davidson’s social anti-individualisms (or externalisms). It is true that in each case individuation of at least some mental states is held to be dependent on relations the individual bears to other individual thinkers. But there are fundamental differences. Hahn formulates the point of agreement thus: “in order to have thoughts or language we need other people whose perceptual and concept-forming abilities and dispositions are innately substantially similar to ours” (p. 54). This formulation is potentially misleading. It is insufficiently specific in the use of ‘in order to’ and ‘need’. If the point is just that to learn language, we need relations to others, as a psychological necessity, then there is surely agreement. But this point is not anti-individualistic. If the point is that, as a constitutive matter, having thoughts or language is dependent on relations to other people, then although it constitutes a strong form of social anti-individualism, there is no agreement. Davidson holds this view. I reject it. Hahn recognizes this difference, but it is somewhat elided in his formulation of the point of agreement. Both Davidson’s and my social anti-individualisms are supported by apriori arguments. My conclusions are meant to be metaphysically necessary truths. Davidson’s are supposed to be whatever approximation to metaphysical necessity he allows. The arguments and conclusions are, however, at different levels of generality. Davidson’s conclusion is supposed to give constituent conditions on what it is to have language or thought. My con-
Reply to Hahn
357
clusion gives constituent conditions on having certain particular types of thoughts, given that certain contingent parameters are fixed. For example, I think that it is metaphysically contingent that an individual is dependent on others for his acquisition of words. Similarly, it is a contingent matter that an individual allows that norms governing the use of his words can be partly dependent on others’ usage. It is a contingent matter that an individual lacks sufficient background knowledge to distinguish—by perceptual, explicational, or theoretical means—arthritis from other rheumatoidal diseases. The apriori argument has two conclusions. First, it is possible under these conditions to have a concept of arthritis and to think specifically about arthritis. Second, given that these contingent matters are in place, and given the particular type of notion arthritis is—for example, that it is nonindexical and not a natural kind notion—it is necessary in order to have a thought about arthritis that one be in certain relations to others who are in a better position to specify the disease. I make similar arguments for a wide variety of other notions. But I do not take social relations to be constitutively necessary for having language or thought. So my apriori arguments for necessities of social anti-individualism presuppose that certain contingent parameters are fixed. The arguments are not as global as Davidson’s. Despite the fact that my arguments are more modest (and I think more concrete and solid) than Davidson’s, he rejects them. One ground Davidson gives for rejection is that the incomplete understanding that they invoke—the inability to distinguish arthritis from other rheumatoidal diseases—is supposed to conflict with first-person authority (Davidson 1987, p. 449). Since the supposed conflict is never explained, and since, as Hahn notes, my approach to understanding first-person authority’s compatibility with antiindividualism is similar, in broad outline, to Davidson’s, I have been puzzled by this ground. I suspect that it rests on a failure to distinguish knowing what one’s thoughts are in the sense of being able to understand them well enough to think them and self-attribute them, and knowing what one’s thoughts are in the sense of being able to give correct explications of them.15 Davidson’s indiscriminate claim that on my view the relevant agents do not know what they mean and think suggests this conflation (Davidson 1987, p. 449). As I have argued in various places, we are authoritative about the former sort of understanding and about certain self-attributions that rely on that understanding; but we are not authoritative about explications. Incomplete understanding involves inability to provide full and correct conceptual explications. That is not a distinctively first-person ability. Although I believe that it can have a special type of apriori warrant, the warrant is not that of first-person authority. Another ground Davidson gives for his rejection is the proposal of a different interpretation of the particular cases. Where I hold that a person could believe that arthritis occurs
358
Tyler Burge
in the thigh, Davidson maintains that any such interpretation of a person’s belief could not be right. He appeals to holism about belief and the uncontroversial point that such a person would associate arthritis with different background beliefs and inferences from someone who knows that arthritis can occur only in joints. He maintains that the error is a metalinguistic one about the dictionary meaning of the word ‘arthritis’. I do not, as Davidson charged, insist that “we are bound to give a person’s words the meaning they have in his linguistic community” (Davidson 1987, p. 449). Rather, my account is based on details particular to the case. I tried to build into the case various facts that make the metalinguistic move unacceptable. Davidson does not discuss these. Moreover, the generalized appeal to holism to support rejection of the arthritis case seems to me vague and ineffectual. Much of the import of the work of Kripke, Donnellan, and Putnam has been that in establishing meaning and reference of a term (and by extension, the reference of concepts in thought), inferential connections and background knowledge are not the only factors and are sometimes not decisive. Often there are person–world relations that play a large role, even when the person is ignorant or mistaken about the referent. The arthritis case is such a case. I continue to believe that the doctor and patient can share beliefs like the belief that arthritis is a painful disease, and thus can share a concept of arthritis. They can do this even though the patient is mistaken about some fundamental features of arthritis and has vastly less background knowledge than the doctor. I believe that there is a complex network of principles, with many escape clauses, that carry a bias in favor of preservation of meaning and concepts between people in the same community who communicate with one another. I have developed this position in a series of papers about interlocution.16 There remains a great deal to be done if we are to understand the phenomenon in depth. The main thrust of the original social anti-individualist thought experiments does not, however, depend on my view that there are principles governing communication that defeasibly enforce a bias toward preservation of conceptual content and that can override differences in explication and conception associated with the concepts. Although I claimed that the doctor and patient share a notion of arthritis, the outcome of the thought experiments does not depend on this claim. As Hahn points out, I argue in a later article that thought experiments work on the mere assumption of shared reference. The doctor and the patient need only have concepts that apply to arthritis. The doctor might well have a different concept of arthritis by virtue of his or her superior background knowledge. It is enough that they share a reference to arthritis. The idea that people can share a reference while differing widely in their background knowledge and assumptions about the referent is fundamental to the work of Kripke, Donnellan, and Putnam. I regard it as well established. People allow the referents or extensions of their concepts and words to be partly dependent on chains that pass through other people.
Reply to Hahn
359
The thought experiment can make use of this phenomenon to ensure that the patient relies on others to help determine the referent of the concept expressed in the patient’s idiolect by the patient’s term ‘arthritis’. I do not think the doctor and patient have to fail to share any concept of arthritis. But even if the patient’s concept were to differ from any concept of arthritis that the doctor has, we can safely assume that there are cases where the referent is preserved. Then we conceive of another circumstance in which a physically similar patient is connected through another communal network to a different syndrome of diseases—perhaps all rheumatoidal ailments. Since the relevant concepts are, or can be, nonindexical, the difference in referents in the two circumstances entails that the respective patients have different concepts about these referents. So they have different beliefs with different conceptual contents. The differences clearly depend on differences in their respective social circumstances, external to their physical make-ups, considered individualistically. These differences in social circumstances ultimately connect the individuals with different diseases or groups of disease-instances. The different social networks connecting the patients to different syndromes of disease necessitate that the patients have different beliefs with different conceptual contents.17 This argument depends on the contingent fact that the patients are ignorant of certain facts. It depends on the contingent fact that they have nonindexical concepts for the relevant diseases. It depends on the contingent fact that the patients rely, perhaps intentionally, on others for the fixing of the referents of their concepts. Social interdependence is a deep and pervasive but, I think, contingent fact about language users and thinkers. I do not see it as a metaphysical necessity for language or thought. Given that social interdependence is assumed to be in place, however, the argument claims apriori that under the circumstances it is necessary that the respective patients have different thoughts. Thus it seems to me that Davidson’s social anti-individualism is at least as different from mine as his physical anti-individualism is from mine. What we have in common is an appreciation of ways in which individuals are dependent on matters outside them for many of their thoughts. The structure of this dependence still invites exploration. Notes 1. The latter thesis is initially stated: If a creature lacks de re attitudes, we would not attribute any attitudes at all (Burge 1977, p. 347). Hahn is correct in pointing to the oddity of the formulation in terms of attribution. The intent is that having de re attitudes is necessary not just for attributing attitudes to a creature, but for its having propositional attitudes—as subsequent discussion (e.g., ibid., p. 348) indicates. My misleading formulation is a product of excessive focus on language and on attribution through language. I believe that this is the residue of my initial entry into these issues through a fairly compartmentalized view of the philosophy of language. There are numerous passages in the paper that constitute the beginning of my shifting my focus to the nature of the attitudes expressed or attributed by the language, rather than the language used in making the attributions. But this and a few subsequent articles still involve some awkward straddling of the distinction. As Hahn recognizes, I was not maintaining the Davidsonian interpretationist view (that is, the view that an interpreter’s interpretation
360
Tyler Burge
helps constitute the linguistic and mental content of the creature being interpreted). But the formulations indicate Quine–Davidson influence in considering matters from the standpoint of an interpreter. 2. On this view, even innate structures come to be intentional only by being triggered when the individual engages in some demonstrative-like relations to some referential objects. But I think that a somewhat less committed position would serve my primary purposes. One might allow that an individual could have innate structures with intentionality even before these structures are triggered through activity of or in the individual, but maintain that the intentionality of the innate structures is possible only because it is appropriately connected to prior de re attitudes by others. For example, God’s de re attitudes in fashioning the innate structures (on a traditional rationalist line) could play this role. Or the individual’s evolutionary ancestors’ de re relations in the evolution of the individual’s innate perceptual and other categories could play this role. In both cases the de re relations would help give the innate structures their intentionality and help determine their specific referential characteristics by helping to give them a certain function in relation to the relevant subject matter. Then the less committed thesis would be that in order for a creature to have any attitudes at all, the creature’s attitudes must presuppose and be appropriately connected to de re attitudes—either by that individual or by individuals on whom the creature’s attitude-structures depend. 3. I explicitly included sensations and mathematical objects as possible objects of de re attitudes. See my (1977), pp. 352–353, 361–362. 4. This is why I said above that de re attitudes are “in a certain sense” irreducibly relational. To be de re, an attitude must be in relation to an object that it is about. But on my view, the attitudes that are in fact de re, are not relations. What makes de re attitudes irreducible to de dicto attitudes, where we are considering the attitudes themselves, is not their bearing relations to the res. It is their being not completely conceptual. They involve irreducibly applicational or token-demonstrative elements. Some attitudes that are not de re, in the sense that they fail to be about an object, are irreducible to de dicto attitudes for the same reason. Although they fail to pick out an object, they are irreducibly demonstrative or applicational. So, for example, a demonstrative thought applied under an illusion, where no object is referred to, is just as irreducibly non–de dicto, because it is not completely conceptualized, as a de re thought. The key distinction on my view is between what is standing, sayable, or conceptual content and what is involved in showing or demonstrating or applying the standing content on particular occasions. 5. I think of concepts as abstract entities. But for purposes of my arguments, it is not crucial that one think of concepts this way. The term ‘concept’ marks a place for individuating aspects of propositional attitudes. The aspects are the formally distinguishable aspects of the abilities associated with propositional attitudes—abilities like recognition and inference. To type and explain these abilities, one must speak of such formal aspects. For example, one must link a belief that humans are mortal with a belief that Socrates is human. The concept human marks something common to these two attitudes that ties them together in inference. The thought experiments are compatible with just about any ontological (or anti-ontological) position on concepts. I do think it appropriate to quantify over concepts and take them as objects of reference. And I have views about the ontologies of various types of concepts. These views are, in some cases, related to my anti-individualism. But these positions are not presupposed in the anti-individualist thought experiments. 6. I mean by this that the mere fact that one will interact with an object in the future is of little individuative relevance. But intentions, beliefs, desires that are involved in making objects (such as artifacts) might be partly individuated in terms of the objects made, especially with regard to de re applications in the intendings. 7. I made this sort of point with respect to beliefs about water, in my (1982), note 18. 8. I believe that Davidson overrates them, as will emerge below. 9. These points, which constituted a break with Quine on stimulus meaning, are of great importance. They certainly influenced my nascent thinking about these matters. But Davidson’s explanations of externalism and his arguments that it is true seem sometimes to present this point about evidence for interpreting attitudes as sufficient for the externalism. This leads him, I think, to underrate the role of counterfactuals in establishing a thesis about the nature and constitution of the attitudes. See Davidson (1987, p. 450). 10. This is not to say that Davidson’s account of the nature of interpretation is unattractive. I think that his contributions in understanding radical interpretation are brilliant and in many fundamental respects true. It is the further view that mental states are constituted by a relation to an interpreter that I regard as mistaken.
Reply to Hahn
361
11. I leave open how quickly such relations become “normal” and begin to make content possible. In some cases, I conjecture, relevant relations are established through a single instance of interaction. How content spreads through the cognitive system is a matter I will not try to adjudicate. Since it is not a central case and may involve fuzzy boundaries, I am not sure that it is important to have a definite view about the matter. 12. The view of “Belief De Re” also requires that to have intentional content, the individual must bear some nonintentional relations to a subject matter. Swampman appears to bear no relevant relations to its environment. Whether there are some de re relations to some other subject matter that might make possible some content would depend partly on a fuller description of the case. 13. On relations to the future, see my (1998b). 14. This argument from the variety of possible matches is important in my reasoning. If I were persuaded that Swampman’s physical structures and dispositions were uniquely fitted to some aspects of his environment—so that no metaphysically possible substitute for those aspects would make any sense of its chemical structures and responses—then I would think it worth considering whether, despite the lack of historical connection to the environment, the physical structures had the function of responding to just such environmental aspects. But even this argument would be a difficult one to make. I doubt that function is ascribable to chemical as opposed to biological structures. And it is debatable whether Swampman has any biological functions, at the time of its synthesis, in relation to its environment. Moreover, the notion of function, or some related notion, would have to be motivated as applied to this product of accidental synthesis. All this aside, as noted in the text, it seems to me very doubtful that there is a metaphysically necessary unique fit between Swampman’s physical structures and any aspect of the objective environment. It seems to me that any environmental property that is perceivable has possible lookalikes that are indiscernible (under sufficiently odd but metaphysically possible circumstances) given the actual dispositions of a given nonomniscient individual. These are complex matters, however. Although I am strongly inclined to think that Swampman can have no thoughts about its empirical surroundings (until it opens its “eyes” and starts interacting with those surroundings), the matter may bear further discussion. 15. I emphasize this distinction in various places. See my (1988), p. 662. 16. See my (1993), pp. 457–488; (1997), pp. 21–47; (1998a), pp. 1–37; and (1999). 17. The argument is from “Wherein Is Language Social?” (1989).
References Burge, Tyler. 1977. Belief De Re. Journal of Philosophy 74: 338–362. ———. 1979. Individualism and the Mental. Midwest Studies in Philosophy 4: 73–121. ———. 1982. Other Bodies. In Thought and Object, A. Woodfield (ed.). Oxford: Oxford University Press. ———. 1988. Individualism and Self-Knowledge. Journal of Philosophy 85(11): 649–663. ———. 1989. Wherein Is Language Social? In Reflections on Chomsky, A. George (ed.). Oxford: Blackwell. ———. 1993. Content Preservation. The Philosophical Review 102: 457–488. ———. 1996. Our Entitlement to Self-Knowledge. Proceedings of the Aristotelian Society 96: 91–116. ———. 1997. Interlocution, Perception, and Memory. Philosophical Studies 86: 21–47. ———. 1998a. Computer Proof, Apriori Knowledge, and Other Minds. Philosophical Perspectives 12: 1–37. ———. 1998b. Memory and Self-Knowledge. In Externalism and Self-Knowledge, Ludlow and Martin (eds.). Stanford: CSLI Publications. ———. 1999. Comprehension and Interpretation. In The Philosophy of Donald Davidson, Lewis Edwin Hahn (ed.). Chicago: Open Court. Davidson, Donald. 1987. Knowing One’s Own Mind. Proceedings and Addresses of the American Philosophical Association 60: 441–458. Kant, Immannel. 1787. Critigure of Pure Reason, second edition. Guyer and Woods (eds.). Cambridge: Cambridge University Press.
16
The Thought Experiments: Reply to Donnellan
Keith Donnellan’s essay raises important questions about the nature of the Twin Earth thought experiments. The relations between Putnam’s thought experiment in “The Meaning of ‘Meaning’ ” and mine in “Individualism and the Mental” are complex. The obvious similarity is the use of the Twin Earth methodology, for which I am indebted to Putnam. Putnam focused on language, not mind. He did not initially argue for antiindividualism about mental states at all. In fact, he assumed that the psychological states of the twins are the same. I criticized this construal of his thought experiments. I took his thought experiments regarding natural kinds, properly interpreted, to provide a different type of support for the most general anti-individualistic conclusion of my arthritis thought experiment. The most general conclusion is that the nature of many mental states is such that being in those mental states necessarily requires bearing certain relations to one’s environment. Putnam and I are now in agreement on these matters (Burge 1982; Putnam 1996). There are other similarities between my arthritis and water thought experiments. Both depend on the fact that we are not omniscient and the fact that there is possible slack between what we know descriptively about the referents, or correct applications, of our concepts and what their referents are. Since the sort of incomplete linguistic understanding that enters into my thought experiments is equally an incomplete knowledge of the way “the world” is—or what is the true character of what is objectively indicated by our thought and language—there is a sense in which both thought experiments depend on ways that “the world” determines our thought. I I think that Donnellan is incorrect in suggesting that I have conflated the two thought experiments, and in suggesting that I have argued that mine is a variant of Putnam’s. I shall try to explain this as matters proceed. Donnellan claims four differences between the thought experiments. The first is that the relevant environment in my thought experiments is social or linguistic, whereas the relevant environment in Putnam’s thought experiment is the nature of the physical world. I have, of course, highlighted this point. But I reemphasize that the understanding of language is not sharply separable from knowledge of the world. Not understanding that ‘arthritis’ applies (“by definition”) only to diseases in joints is equally not knowing that arthritis occurs only in joints. I believe that the thought experiments that show that perception and ordinary thought about the physical world are anti-individualistically individuated are in their pervasive application and ontogenetic primacy more fundamental than the social anti-individualistic thought experiments.
364
Tyler Burge
Moreover, I think that it is a mistake to hold, as Donnellan appears to, that sofas will have no place in a thought experiment where the relevantly different environments are nonlinguistic. In Burge (1986b), I showed that thought experiments can be constructed wherever there is scope for radical theory about the true character of a type of entity. There is such scope even for sofas, as I specifically argue in that paper. I agree with Donnellan’s account of a second difference between the thought experiments—his account of the difference in the ways experts enter into the two thought experiments. I want, however, to enter a qualification. It is not essential to the “social” thought experiments that the individual rely on “experts” or others with more competence for fixing the linguistic intentional meaning of his terms. One can develop the case in such a way that the patient does not rely on the doctor’s meaning. The patient and the doctor can have different meanings for the term ‘arthritis’, in their respective idiolects, in the actual situation. I think that it is natural and correct in many instances to take them to have the same meaning. But there can be such differences of idiolect, without those differences’ undermining the main anti-individualistic conclusion of the thought experiment (Burge 1989). What is crucial is that the referential application of the patient’s term be dependent on the social environment for its connection to the world. The patient’s term and the patient’s concept can attach to arthritis, even if he and the doctor conceive arthritis differently and mean different things by ‘arthritis’. What makes the thought experiment work is that the patient has mistaken beliefs about the disease arthritis, which the doctor’s superior expertise reveals to him. In the twin situation, the patient’s concept attaches to a quite different syndrome of diseases, as does the doctor’s. Given the nonindexical nature of the respective patients’ intentional contents or concepts, those contents’ having different referents individuates them as different contents. This difference entails that the patients in the different environments are in different mental states. The difference depends on differences in their social environments even though the patients in the two cases do not share concepts with the experts in their environments. They share only referents with those experts. Or at any rate their referents are sufficiently affected by their dependence on others to ensure that the respective patients have different referents from one another. I am more doubtful about Donnellan’s account of a third difference. I do not see any reason why my thought experiments are not parallel to Putnam’s in suggesting a “semantical rule,” though a different one, that describes the social thought experiments. Something like: “Under such and such conditions, the referential application (or the sense) of the patient’s term ‘arthritis’ is the same as the referential application (sense) of the term as used by others on whom the patient depends in such and such a way.” The referential application of the term as used by others will depend on the term and the others. I will return to the bearing of such semantical rules on anti-individualism.
Reply to Donnellan
365
Donnellan’s fourth point concerns an indexicality that he sees as central to Putnam’s thought experiment but absent in mine. There are several things that Donnellan says on this topic that I believe to be mistaken, or at best misleading. Donnellan agrees in locating indexicality not in the terms themselves. He places it in a semantical rule governing them. He admits, however, that the rule itself does not have to involve indexicals. It could be formulated: “Water is whatever has the important underlying physical characteristics of the stuff users of the language, which itself could be specified nonindexically, call ‘water’.”1 We could replace ‘users of the language’ with ‘we’, as Donnellan proposes; or we could use a definite description. There are other nonindexical possibilities. Why does Donnellan feature the indexical formulations? He writes, “We want [them] in part at least in order to ensure that in saying that the psychological states of counterparts on the two planets can be the same. . . .” He is not fully explicit why “we” want such a thing. Sometimes it appears that it is because Donnellan is trying to be faithful to Putnam’s original interpretation of the thought experiments. On that original interpretation, the twins were in the same psychological states, but differed in the meanings of their linguistic terms. But I have shown this interpretation to be mistaken. It is now rejected by Putnam as well. The relevant psychological states of the twins are not the same. Sometimes it appears that Donnellan writes what I have just quoted because he himself wants to hold that Putnam’s thought experiments do not support anti-individualism. For if the psychological (mental?) states are the same between actual and twin situations, the thought experiment does not support the view that the nature of the individual’s mental and psychological states is dependent on relations to the physical environment.2 Donnellan claims that we can imagine the twin’s “having the same history and (in an important sense) psychological states.” But he never gives any argument for this contentious point. I believe that the twins cannot have all the same psychological states, unless ‘psychological’ is given a narrow meaning that neither common sense nor empirical science actually uses. I have argued this in various publications, and Putnam has argued similarly. Suppose that it could be shown empirically—what I doubt—that all relevant individuals have such a rule in their minds or heads. The implausibility of using this line to try to save individualism is highlighted by the fact that perceptual categories, in animals and young children, are subject to relevantly similar thought experiments (Burge 1986a,c). Such thought experiments do not depend on an underlying nature but merely on an objective, mind-independent perceptual kind. But perceivers do not carry around in their minds an analogue of the indexical semantical rule governing the application of their perceptual categories. I find such a supposition about an internal semantical rule doubtful even for the case of natural kind concepts. I think it empirically doubtful that young children, or indeed many adults, who have a distinctive use for natural kind words, and thus a sense for the
366
Tyler Burge
appearance–reality distinction, have metaconcepts for words and for calling something by a word. I see no reason why they must have such metaresources in order to use natural kind terms in the way we do. Spelling out the rule in detail—specifying what stuff it is that is being called ‘water’—requires a conceptual-descriptional knowledge that I think most users of the term or concept simply do not have. The information is often stored perceptually, not conceptually. (For more on this, see my reply to Peacocke in this volume.) I believe at the very least that individuals need not have conceptualized their usage in order for that usage to be correctly described by such rules. Let us waive all these doubts, however. Suppose that the rule is formulated indexically and that the indexical formulation is common to all relevant twins. Then it still does not follow that the individuals are in the same psychological states. For the words and concepts that the indexically formulated rules govern are not indexical. The semantical rule is not synonymous with or conceptually identical with the words or concepts being semantically explicated. The sentences and thoughts using these words and concepts get their intentional content, not just their referents, from interaction with the actual environment that they are in. Since the environments are relevantly different, the intentional contents are different. The intentional contents are distinct because of the difference in the environments that the twins are related to. The original point I made in “Other Bodies” that the relevant concepts and words are not indexical is not merely a quick, shallow point. It goes to the heart of the difficulty for defending individualism along these lines. Donnellan gives no grounds for believing that the point applies only to language and not to mind. II Donnellan’s widget example is nicely posed and contains more than immediately meets the eye. He imagines that we introduce the word ‘widget’ as applying to any physical object that has all three physical properties that one finds by repeating the following procedure three times: One opens the Shorter Oxford English Dictionary at random and looks (presumably scanning through entries alphabetically) for the first word that designates a property a physical object can have. I think that Donnellan is clearly right to reason that the relevant semantic features, or at least semantic rules like those he gives, of terms explained in this way are not sufficient for a term to be a natural kind term. For one thing, there must be a unified underlying nature—something that is not guaranteed by the widget rule, or by any apriori rule for ordinary macroterms like ‘water’. For another, there must be more to understanding and using the term than mastering its syntactic namelike properties combined with a widgetlike semantic rule. With natural kind terms like ‘water’ there is also a background of apply-
Reply to Donnellan
367
ing a term according to relatively unified identified macroproperties. These, not a random procedure, guide one to an underlying nature. I think that Putnam never did show that “the meaning of a term [like ‘water’] cannot be both in the head and determine the extension of the term.” The proposed semantical rules for ‘water’ and ‘widget’ are not their meanings. ‘Water’ is not synonymous with Donnellan’s semantical rule for it (Burge 1982; Putnam 1996). I do think that the semantical rule for ‘widget’ is closer to being its meaning, although here too the meanings will differ. The term ‘widget’ functions as a namelike term that is rigid (assuming that its reference is fixed by one sequence of three dictionary look-ups) in a way that the descriptive meaning of the semantical rule is not. But understanding the term requires mastering nothing further. Imagine a twin who carries out the same procedure that you do in establishing an extension for ‘widget’. Imagine that the twin randomly arrives at different properties and a different extension. Would the twin think a different thought when using the term ‘widget’ than you would? I do not think that the answer matters, because of the artificiality of the term and the fact that it is importantly different from the terms discussed in the original thought experiments. Because the procedure is random, probably most would say that at least before an extension is established, the terms in the twin situations express the same thought-components and meaning. After the procedure is carried out, perhaps intuitions will differ. I am inclined to hold that the twins will express different meanings and thoughtcomponents (assuming that the different extensions are fixed by one sequence of dictionary look-ups in each situation). But I would not want to rest a theory on the point. III I would like to turn now to some comments that Donnellan makes on my social thought experiments. He characterizes them as resting on three premises. The first claims the possibility of incomplete mastery of intentional contents or concepts. The second is that the standard of complete, or greater, mastery resides in the usage of the individual’s speech community. The third is that twins have incomplete mastery of different concepts because they are in different communities that set different standards. I would like to emphasize (as I do in my reply to Loar) that the thought experiments invite judgments about particular cases. They do not rest on general premises—although the judgments may be guided by principles that can be discerned retrospectively by reflecting on the judgments about cases. I think that all of the premises Donnellan cites are probably true. The actual presentation of the thought experiments in “Individualism and the Mental” does accord with Donnellan’s premises, as far as I can see,3 though I definitely
368
Tyler Burge
do not argue from those premises (Burge 1979). Similar but different thought experiments dispense with incomplete linguistic understanding and locate the individual’s idiosyncrasy in criticism of the communal meaning or communal understanding (Burge 1986b). And as I have indicated, the community need not provide a standard for the individual’s understanding or linguistic meaning, except insofar as it provides a link to the referent of the individual’s term (Burge 1989). I agree with Donnellan’s association of incomplete understanding with Plato, and more broadly with the rationalist tradition. I think that he is also correct in seeing categorial notions like justice—I would add logical and mathematical notions—as raising complications about the relevant standard for complete mastery. I believe that social thought experiments can apply to terms like these, but the issues are complex, because of the possibility of a fundamental or dominant concept or meaning that has a certain priority in rational thinking. I hope to address these issues elsewhere. In the last section of his essay Donnellan invokes the metalinguistic strategy to defend individualism against the social thought experiments. I agree with his description of the case that involves total failure of understanding of French words on a menu. I do not quite agree with his description of what is involved in the passage of a name. I think that names are closely associated with metalinguistic paraphrases, but are nevertheless different (Burge 1973, pp. 425–439).4 In cases of passage of a name, one has a cognitive marker that associates the name with a mental file that tracks the communicative source. I do not, however, differ fundamentally with Donnellan about this case either. On the other hand, Donnellan’s appeal to a metalinguistic paraphrase to account fully for the thoughts the patient expresses with the word ‘arthritis’ is, I think, unacceptable. According to Donnellan, “what he really believes is only that he has in his thigh the condition called ‘arthritis’ or something like that.” In “Individualism and the Mental” I argued in some detail that the case as I arranged it could not be credibly deflated in these metalinguistic terms (Burge 1979, IIIC). The individual can be quite explicitly not focused on words, and explanations of his behavior will not go through beliefs about words. I have also emphasized the enormous extent of the phenomenon of incomplete understanding, or reliance on others for meaning and reference. A significant portion of our vocabularies involves terms that we have an imperfect grip on. I find it incredible that metalinguistic beliefs—beliefs analogous to those infected by foreign terms that are not understood at all—underlie the huge range of cases in which what I have called incomplete understanding occurs. We are not cognitively so massively fixed on our own languages. Some of our nonindividualistically individuated beliefs are prelinguistic. Our cognition is largely directed toward the world. The differences between the ordinary cases I discuss and cases in which one uses a foreign word that one does not understand are palpable. Donnellan provides no answer to
Reply to Donnellan
369
the detailed discussion in my (1979) of the metalinguistic move. So I will not try to take the matter further here. (For more, see my replies to Owens, Peacocke, and Loar.) I think that Donnellan is right in holding that Putnam’s thought experiments, properly understood, help provide new insight into the way natural kind terms are understood: A complete grasp of the concept is still beholden to the empirical nature of the empirical world for determining what is grasped. My social thought experiments bring out how the nature of our thoughts depends on the ways our relation to the physical world is shaped and mediated by our relations to other people and their linguistic and conceptual relations to the same world. But anti-individualism is illustrated in a wide range of cases that are not easily assimilable either to Putnam’s natural kind example or to my example of incomplete understanding of terms whose meaning is set by social norms. The slack between an individual’s perspectives, physical processes, and individual behavior, on one hand, and the objective environment beyond the individual, on the other, takes many forms. The individual’s intentional content or perspective cannot be individuated apart from the objective world that it interacts with and purports to characterize. Notes 1. The semantical rule needs a fall-back position if there are no unified, underlying physical characteristics. For we cannot know apriori whether a term is a natural kind term. But I omit to enter this into the rule, for the sake of simplicity. 2. I do not see that Donnellan is suggesting a distinction between psychological and mental states. I believe that such a distinction would be pointless and unattractive. 3. A minor exception is that in my examples, it is not obvious that the twin, as opposed to the patient in the original “actual” situation, has an incomplete mastery of his term. 4. I do not regard the predicative aspect of the name ‘Aristotle’ as synonymous with ‘is called “Aristotle” ’, but I think that the latter explains the core semantical conditions for applying the predicative aspect of the former.
References Burge, Tyler. 1973. Reference and Proper Names. Journal of Philosophy 70: 425–439. ———. 1979. Individualism and the Mental. Midwest Studies in Philosophy 4: 73–121. ———. 1982. Other Bodies. In Thought and Object, A. Woodfield (ed.). Oxford: Oxford University Press. ———. 1986a. Individualism and Psychology. Philosophical Review 95: 3–45. ———. 1986b. Intellectual Norms and Foundations of Mind. Journal of Philosophy 83(12): 697–720. ———. 1986c. Cartesian Error and the Objectivity of Perception. In Subject, Thought, and Context, McDowell and Pettit (eds.). New York: Oxford University Press. ———. 1989. Wherein Is Language Social? In Reflections on Chomsky, A. George (ed.). Oxford: Blackwell. Putnam, Hilary. 1996. Introduction. In The Twin Earth Chronicles: Twenty Years of Reflection on Hilary Putnam’s “The Meaning of ‘Meaning’,” Andrew Pessin and Sanford Goldberg (eds.). New York: M. E. Sharpe.
17
The Indexical Strategy: Reply to Owens
I am in broad agreement with Joseph Owens’s fine essay. My reply will be relatively brief even though the issues that the paper raises are complex. Contrary to Quine’s epigram, agreement, at least in philosophy, hardly goes without saying. But owing to our contentious natures as philosophers, agreement seems to breed brevity. I think that Owens is right in his rejection of the indexical strategy for blocking antiindividualist conclusions. I think that he is also right in his conclusion that the aspects of psychological states typed by character in Kaplan’s sense—or alternatively, typed by the intentional content of indexical expression types—are themselves individuated in nonindividualistic ways. I want to enter a few qualifications and elaborations on Owens’s basic points. My central claim in the Twin Earth thought experiments is not that the twins entertain different propositions and so warrant different belief characterizations. The notions of proposition and belief characterization have so many uses and construals that my view would be compatible with the sorts of trivializations that Owens argues against if it were not more specific. The central claim is that a wide range of mental states—considered in the ordinary nonrelational way as states in the mind or brain of the individual—are dependent for their natures and identities on relations between the individual and a physical and social environment beyond the skin of the individual. The view does not depend on appeal to singular propositions that include as constituents physical objects located beyond the individual’s bodily surfaces. It also does not depend on thinking of mental states as themselves relations between an individual and elements in his environment. I find both of these conceptions unnatural and insufficient for understanding mental states and epistemology, though useful for limited purposes. Where they are applicable, these notions are applicable in virtue of the sorts of mental states that are commonly recognized. For example, one believes a singular proposition only in virtue of intentional representations (concepts, percepts, or demonstrative representations) that indicate the objects and properties. One cannot relate to an object in thought except by indicating the object in a way that is part of some perspective on it. One bears de re mental relations to objects in the environment only by being in propositional states that are typed in terms of intentional contents that indicate, in a context-dependent way, the relevant res. Anti-individualism is also not the claim that mental states fail to supervene on nonintentionally individuated physical states within the individual’s body. I believe that certain failures of supervenience do point to the truth of anti-individualism. And I argued for such failures en route to arguing for anti-individualism. But supervenience failure is logically compatible with individualism. In the context of a “Cartesian” claim of the total independence of the mental from the physical and from any surrounding environment, one could consistently maintain supervenience failure together with individualism.
372
Tyler Burge
Supervenience failure and anti-individualism are arguably logically independent in the other direction as well. The supervenience of mental states on underlying physical states for an individual is logically compatible in some instances with the truth of antiindividualism. If, for example, differences in the normal distal causes of perceptual states that are sufficient to individuate the perceptual states differently were also sufficient metaphysically to necessitate differences in the internal physical states, then the perceptual states might supervene on the internal physical states. I think that, generalized, such a view is metaphysically unattractive. I think that it is also empirically unwarranted as an account of the individuation of fundamental and even certain higher-level internal physical states. In such a case, both the underlying chemical states and the perceptual states would be nonindividualistically individuated—through their environmental relations; and they would vary together with variations in causal antecedents. We do not individuate chemical events in that way. But the point is that if supervenience did hold in these cases, that would not itself support individualism about perceptual states or perceptual beliefs. Antiindividualism is a claim about individuation, not a claim about possible alternative relations of internal, nonmental states to the environment. To put the point another way, suppose that it were true, as Davidson holds that it is true, that any difference in the causal antecedent is necessarily sufficient for a difference in the identity of the effect (Davidson 1980, 2001). Then any difference in the antecedent causes of a mental state between the twin worlds would necessitate a difference in the internal physical states of the two individuals. So the Twin Earth cases would never illustrate a case in which the internal physical states of the twins would be the same while the mental states differed. But anti-individualism would remain true. For the mental states of the individuals would depend for their natures on relations to the individuals’ environments. It is just that their physical states would be anti-individualistically individuated in an exactly parallel way. So supervenience would hold. I believe that Davidson’s generalized doctrine about events is not true. But the issue here is not the general metaphysics of event individuation. My point is just that there is a notional difference between the doctrines of local supervenience-failure and anti-individualism. Owens’s criticism of the indexical strategy seems to me to be to the point. The strategy is almost always thinly supported and often rests on a misreading of analogies or a misjudgment of the thought experiments. Owens brings out an elementary version of this tendency in his criticism of taking the twins to have the same words in a single language. Moreover, the strategy distorts the logic and truth conditions of our language and thought. Almost any speaker/thinker can, through counterfactual cases, be brought to recognize that in his idiolect/psychology a term/concept like aluminum works very differently from any indexical expression. An indexical shifts its referent with possible contexts; the term/concept does not. The description-governed indexical applies to things in some pos-
Reply to Owens
373
sible circumstances that the term/concept does not. Even with respect to actual cases, the individual will often recognize that the descriptions that he has available for picking out instances of the candidate term/concept may not always suffice to isolate what the candidate term/concept applies to. By contrast, a term/concept like aluminum suffices to apply exactly to aluminum and nothing else. Moreover, the individual will commonly recognize that his term/concept can be shared with others even though they associate different description-governed indexicals with it. The individual will also commonly recognize that the term/concept can remain intentionally and referentially the same even while his own means of identifying the referent may shift over time. The individual will commonly speak and think in accord with the recognition. Some philosophers have taken over the anti-individualist account of how the applications of the candidate terms/concepts are fixed and have placed them in the psychology of the individual and claimed that this complex anti-individualist set of directions is constant across the twins and does all the work of the candidate terms/concepts. The problems with this strategy are similar. Children and unsophisticated adults need not have internalized all the rules that govern the referential workings of their terms and concepts. Even where the account is subliminally or unconsciously internalized, the complex description does not have the same logic or psychological structure of the ground-level terms/concepts whose intentionality the complex describes. How the term works depends on the environment’s fixing certain parameters that the complex set of directions only generalize over. But terms/concepts work with these parameters already fixed, whether the individual knows how they are fixed or not. So the account in terms of metalevel antiindividualist directions does not supplant or exhaust the candidate term/concept in the language/psychology of the individual—even if and when those directions are internalized by the individual. The content of the individual’s object-level term/concept is still fixed by factors outside the individual. Owens points toward a further problem for the indexical strategy. The character, or alternatively the intentional content of ordinary indexicals, is itself subject to antiindividualist strictures, insofar as it types mental states. As applied to the strategy of internalizing the whole anti-individualist account, the point is that the terms used in the directions for fixing reference are, in most cases, no less subject to the anti-individualist argument than the candidate term/concept. If one tries to apply the strategy to all relevant terms/concepts at once, then, quite apart from the other difficulties that I have alluded to, one is left with an account of the language/psychology of the individual that is too inspecific to our actual environment to be credible. It leaves us absurdly without any cognitive states that specify kinds as such in our environment, and thereby underrates our epistemological powers. It makes psychological explanation independent of specific knowledge and belief about our actual environment, in a way that seriously distorts its character.1
374
Tyler Burge
Owens attacks one root of the indexical analogy by arguing that our understanding of the Kaplan-character (or ordinary constant linguistic meaning) of actual indexicals like ‘today’ is normally or frequently itself nonindividualistically individuated.2 He points out that “twins” could be in different communities where the demarcation of the beginnings and ends of days are slightly different. He then proceeds to imagine that the referents of the twins’ respective uses of ‘today’ could differ because the rule for determining what counts as within a given day would differ. So the character (or ordinary constant linguistic meaning) of the word-shape ‘today’ would also differ. All this could occur without the individuals’ differing in their internal physical states. I accept the conclusion of Owens’s argument about ‘today’. I think, however, that the case is underdescribed as he puts it. The mere fact that Alf, the first twin, lives in a community in which ‘today’ is commonly applied to days beginning at midnight does not commit Alf to the same usage. But I think that Owens is right in holding that there is no reason to think that Alf’s deviant belief (that one o’clock is the end of the day, or the second of two twelve-hour periods) must affect the meaning or the character of the term in his idiolect. The twins could be implicitly committed to the standards of their communities. If Alf himself is disposed to regard others’ usage as a norm for correcting his belief, or if he thinks that when the day begins is not a matter of how he stipulates or takes it to begin but an objectively determinable matter, then lacking some further, very subtle empirical consideration, it is reasonable to take Alf’s belief as mistaken. Alf himself may come to regard it as having been mistaken. And it is reasonable to take the meaning, or at least the rule for referential application of his indexical, to be the same as everyone else’s. With this caveat, I think that Owens’s argument is sound. The same sort of caveat is needed in Owens’s whale-fish example. Merely being in a community in which whales are not counted as fishes by the cognoscenti does not ensure that one’s own word ‘fish’ (or one’s corresponding concept) does not apply to whales. One’s usage has to be appropriately tied to that of one’s surrounding community. (See also my reply to Chomsky in this volume.) Our understanding of the character or intentional content of indexicals is antiindividualistically individuated in another way as well. Understanding the content of indexicals that deal with space or place (‘here’) depends on systematic perceptual and conceptual interaction between the individual—or at least the perceptual system inherited from others—and spatial locations. No thought or use of a sentence could have spatial meaning unless there were systematic causal and perceptual interactions between the individual or his system and specific locations in physical space. I think that we know this apriori, by reflection on cases. The status of understanding of the temporal indexicals, with respect to time—as opposed to specific temporal boundaries as in Owens’s discussion—is less clear-cut. It seems to
Reply to Owens
375
me that our actual grasp of time is interwoven with our experience of change or motion in physical space. Insofar as this is so, the considerations of the previous paragraph reapply. It may be, however, that some aspects of our conception of time could “in principle” be developed from reflection on a procession of thoughts or sensations. For example, it may be that rudimentary conceptions of before and after could be developed purely from reflecting on change within one’s phenomenal consciousness. I am more doubtful that a temporal metric could be developed in this way. It seems to me, for example, that our concept of a day, regardless of exactly when the day is taken to begin or end, is wrapped up with our experience of the trajectory, disappearance and reappearance of the sun, or changes in light. Having temporally specific concepts like today, yesterday, last year, the previous hour, requires some sense and conception of physical changes in space, such as the movement of the sun or changes associated with the seasons. Having such specific senses and conceptions seems constitutively dependent on interaction with a physical environment. Thus having either of the twins’ concepts (of day and day*) inherits the anti-individualistic character of perceptual and conceptual representation of the physical world, quite apart from the more specific differences in their concepts. So the anti-individualistic nature of their concepts seems affected by the physical environment even more fundamentally than by the social environment. The physical dependence would remain even if each individual somehow avoided commitment to his community’s usage. More generally, I think that some of our sense of time depends on innate sensitivities to circadian and seasonal periods, built into our cognitive systems in something like the way it is built into the systems of insects, birds, and other animals. These innate sensitivities are surely anti-individualistically individuated, inasmuch as the cognitive sensitivities to temporal periods are evolutionarily tied to physical changes (day-night changes, the seasons, and so on) in the environment.3 Issues like these seem to me worth developing to determine the scope and form of anti-individualism. The status of our understanding of the first-person indexical and other personal indexicals is more complex than that of our understanding of temporal indexicals. There is, of course, a long tradition, beginning with Kant’s Refutation of Idealism, that maintains that understanding of a self, including temporal properties of oneself, depends on some anchoring of the understanding in experience of spatial objects. Whether, and where, this tradition is stating necessary truths is a question of continuing interest. The nature of our grasp of temporal concepts and the nature of our grasp of the first-person concept are in need of deeper investigation. What seems to me valuable about Owens’s essay is his eliciting some of the ways in which anti-individualism goes deeper than the standard Twin Earth cases may suggest.
376
Tyler Burge
Notes 1. For more on this, see my reply to Loar in this volume. 2. I prefer to concentrate on our understanding of the intentional content of the indexicals, their cognitive or conceptual content. I believe that Kaplan, Owens, and I now all agree that character is not to be identified with cognitive value. It is coarser-grained. 3. For a discussion of these matters, see Gallistel (1990), chapters 7–9.
References Davidson, Donald. 1980, 2001. The Individuation of Events. In Essays on Action and Events. Oxford: Clarendon Press. Gallistel, C. R. 1990. The Organization of Learning. Cambridge, MA: MIT Press.
18
Tracking Perspectives: Reply to Higginbotham
Jim Higginbotham’s elegant and interesting essay deals with some fundamental problems in the philosophy of language, attractively disguised through the medium of low-key technical issues. I am sympathetic with most of what he says. My comments will mainly probe basic terminology and implications for understanding the relation between language, meaning, and thought. Higginbotham’s initial “impressionistic” terminology makes me uneasy. The problem is said to be that if we preserve reference of demonstratives or indexicals, we “lose” or “fail to preserve” the perspective of others’ utterances reported in indirect discourse. Late in the essay he claims that we can “convey” but not “express” the perspective of others in indirect speech. But he does not explicate these terms. What he is invoking with this language is the fact that the reporter just cannot pick out the referent that the initial speaker intended by referentially using the same indexicals that the original speaker did—or indexicals with the same standing meaning or character. And in some cases, understanding the perspective of the speaker requires mastery of contextual matters that goes beyond anything codified in the language. But it is part of Higginbotham’s persuasive point that we are often able to understand and even articulate through our use of language some important aspects of the perspective that the original speaker had on the referents. As Higginbotham adds, understanding such things is part of standing linguistic competence. In that sense the perspectives seem not to get lost. They are preserved, at least marked, through the reporter’s report. Moreover, insofar as the reporting language enables one to keep track of and think, ascriptively, the content of the original speaker’s thoughts—as I will argue that it can in some cases—it might be said, in a so far nontechnical sense, to “express” the original perspective. It might be said to do so even if it cannot “take up” and use that original perspective. Some of this is just terminology, involving no substantive disagreement. But the impressionistic terminology encourages ways of construing the applications of the notions of reference, truth-conditions, meanings, and rules of use that deserve scrutiny. One issue that I would like to reflect on arises in connection with the normal form that Higginbotham proposes. I am glad to see that he has found some use for my old truthconditional explications of demonstratives and indexicals. I would not be surprised if he understands the implications of the proposal better than I did. What I am presently interested in is the relation between the normal form for statements of truth-conditions, on one hand, and the notions of meaning and rules of use, on the other. It is certainly possible to state truth-conditions, as Higginbotham does, in a way that leaves out (on the right side of the biconditional) everything but the referent of the demonstrative or indexical. But it seems to me questionable whether such a statement of truth conditions will coincide with “meaning,” on any normal or theoretically promising construal. For then ‘she is a cat’ and ‘that is a cat’ will have the same meaning, if ‘she’ and ‘that’ refer to the same thing. I am
378
Tyler Burge
aware that some philosophers would embrace this conclusion with enthusiasm. But I do not find it attractive. I do not think that the conclusion will hold up over time. Of course, one can say that the two demonstratives have the same meaning and different “rules of use” (or perhaps better, “rules of contextual reference”). But I am doubtful about this idea. I certainly want to agree that the rule for contextual reference for ‘I’ (or ‘she’) does not have the same meaning as, is not synonymous with, ‘I’ (or ‘she’). The rule is a metastatement about the word; the meaning is not. (This point is relevant to some passages in my reply to Donnellan.) Similarly, I agree that what Higginbotham illustrates as rules of use are not part of the truth-conditions of an utterance—at least insofar as truthconditions are what are stated on the right sides of truth-theoretic biconditionals in “normal form.” I might possibly be brought to agree that ‘this’ and ‘that’ have the same meaning— either generally, or in given instances where they have the same referent—but slightly different rules for contextual reference. But I see no plausibility in claiming that particular uses of ‘that’ and ‘she’, or ‘I’ and ‘she’ (given sameness of reference) have the same meaning (or no meaning at all)—regardless of what one says about rules of contextual reference. This is to say that I am very doubtful about Higginbotham’s recommendation that we “split the lexicon into items with meanings and items with rules of use.” It seems to me that ‘she’ has both a meaning (the quasi-predicational aspect that requires application only to females would presumably be included) and a rule of use, or rule of contextual reference. (There seems to be an analogue of the referential-attributive distinction, in Donnellan’s sense, even with indexicals like ‘she’.) Thus, insofar as statements of truth-conditions are supposed to give statements of meaning, I would think it better not to leave everything out of the demonstrative but its referent. At least it would be better not to do so in cases like ‘she’ or present tense (and unlike ‘this’) where the demonstrative, apart from any particular use, places a requirement on the nature of the referent. This line is, of course, compatible with still associating bound variables, somehow restricted, with the demonstratives or indexicals in the truth-theoretic normal forms. There are fine-grained issues here about whether elements of meaning include only requirements on the nature of the referent or subject matter, or also structural, perspectival requirements on the use-occurrence of the demonstrative or indexical. For example, with present tense (or ‘now’), there is the requirement that the indexical concern times. That is a requirement on the nature of the subject matter that seems to be part of its meaning. Even if ‘this’ refers, on an occasion, to a time, it is not part of its meaning that it do so. With present tense (or ‘now’) there is also the perspectival requirement in most uses that the time be broadly contemporaneous with the use-occurrence of the indexical. I am inclined to think that both types of requirement go into “meaning.” I am inclined to mean by ‘meaning’ what is noninferentially understood through the expression, apart from special context of use of the expression, by any competent speaker.
Reply to Higginbotham
379
Perhaps one should agree that there is more to literal or standing meaning than what is stated as a truth-condition in normal form on the right side of the truth-theoretic biconditionals. Or perhaps one should associate with the bound variables on the right side more of the restrictions on reference or on perspective, which seem to be part of the meaning. If neither of these strategies is attractive to Higginbotham, then I would like a better understanding of his theoretical understanding of the notion of ‘meaning’. As I have noted, it seems to me that things go into the setting up conditions or into the rules of contextual reference that are not part of the meaning or truth-conditions. Such rules make reference to the utterance or to the speaker or to the relation of reference itself, whereas the meaning does not. But I think it mistaken to hold that no aspects of the rules of contextual reference for demonstratives or indexicals are part of their ‘meaning’. This issue spills over into the understanding of the ‘content’ of complement clauses in intensional propositional attitude contexts. It seems to me that these wonderings are different: (1)
Is that an adult female?
(2)
Is she an adult female?
where the demonstratives in (1) and (2) refer to the same individual. In reporting these wonderings in indirect discourse, we should be able to distinguish between the two. (2) is a wondering about developmental stage, but not about gender; (1) could be about either or both. So the complement clause in an indirect report of these wonderings, if it is to remain maximally faithful to the originals, needs to make the distinction. It seems to me that one can wash out the distinction in a report, but one should be able to avoid doing so. It is not clear to me how Higginbotham would respond to these points. Higginbotham’s discussion of Evans and Perry and transformations between coordinate systems seems to me persuasive. I do not see any basis for postulating Evans’s completing senses. But I think that there are distinctions between rules of use (or rules of contextual reference), the indexicals’ (or demonstratives’) standing meanings, and conceptual perspectives backing demonstratives. The conceptual perspectives go beyond anything contained in conventional or standing word-meaning or in the rules of contextual reference. Each of these three elements plays a role in fixing or constituting the elements of coordinate systems. Although rules of contextual reference describe the way that people refer with indexicals, I am not convinced by Higginbotham’s claim that one must understand them to understand the indexicals. Children may not have mastered the metaconcepts needed to understand the rules, but they could still use language in accord with them and count as understanding (comprehending) the terms in a nontheorerdical sense of ‘understanding’.
380
Tyler Burge
The discussion of the puzzles of transformations also seems to be on the right track. The exposition here is masterly. The strategy that tracks perspective seems to me to have an advantage over the strategy that tracks only the possible worlds in which the speaker’s perspective is true. Modally equivalent truth-conditions are likely to be too coarse-grained to capture all there is to perspectives. Contrary to Higginbotham’s suggestion, I doubt whether we can in general construct a perspective out of the reference of an indexical (or especially a demonstrative) together with the relevant possible worlds. But these are technical issues to be worked out on their own. Higginbotham seems to me right in claiming that sometimes the original perspective is lost but unneeded, and that (almost) always the perspective of the reporter is sufficiently different from that of the original speaker that the demonstratives and indexicals the reporter uses will not have the same standing meaning and will not, as used referentially, express the same perspective as that of the original speaker. But as Higginbotham emphasizes, “Fregean elements” remain. Even in these cases, especially with demonstratives, there will always be cognitively relevant conceptual perspectives, modes of presentation, for both speaker and reporter that are associated with the demonstratives and that go beyond both their standing linguistic meaning and their rules of contextual reference. So in using a demonstrative ‘this’ on a particular occasion, one will associate perhaps a perceptual concept with the object purportedly demonstrated. Such conceptual perspectives are necessary for using demonstratives, but they are too variable with context and user to be part of the standing meaning of the demonstratives. Although they are not always “completing,” in the sense that they do not always suffice of themselves to fix a referent (they too rely on nonconceptual relations, and singular applications, to fix their objects), they are richer than the standing linguistic meaning, or the rules of contextual reference. Although they are normally not “completing,” they are otherwise very close to what Frege meant by the senses of context-dependent devices.1 Perhaps Higginbotham is right to balk at calling them “Fregean senses.” But Frege certainly included more in sense than contemporary philosophers include in meaning. I think that this difference does not merely derive from errors in his theory. It derives from his conception of language as primarily a vehicle for the expression of thought. He was not much interested in the lowest common denominator understood among competent speakers—which is approximately what modern theories of linguistic meaning focus on. Frege could (and perhaps did) allow for perceptual as well as linguistic items to be thought components, which are unlike senses only in that they are not expressed in language. They play the same role in his theory of thought. I do not know what it would mean for a sense to “involve objects intrinsically.” I think that some senses for Frege necessarily have referents. On the other hand, I do not think it reasonable to mix up typical referents of conceptual perspectives with the perspectives themselves; only the latter are intentional. The
Reply to Higginbotham
381
objects of de re reference or attitudes are over and above the senses. Moreover, the contextual character of the association between demonstratives and conceptual perspectives is no bar to counting the latter senses. Frege was not primarily interested in conventional linguistic meaning. He was interested in the relation between language and thought. Where language expresses thought only contextually, rather than in a standing or conventional way, the thought can still be the sense of the language. Let me return to Higginbotham’s point that often the reporter does not lose track of the original speaker’s perspective. What seems to me most interesting about his discussion is his claim that some aspects of our keeping track, in favorable circumstances, are part of our linguistic competence. The reporter’s competence in using language to “convey” (to use Higginbotham’s term) the meanings and even conceptual perspectives associated with the original speaker’s demonstratives and indexicals suggests that a theory of language should look more closely at this “conveying” relation. This brings me back to the issue about terminology that I began with. In some cases the relevant competence will involve substantial reliance on nonlinguistic background knowledge. For example, conveying the perceptual perspective of a demonstrative will not flow merely from application of linguistic competence to a few definite and recurrent contextual parameters. In such cases associating the conceptual perspective, or Fregean sense, of the original speaker’s utterance with the report will depend on generalized intelligence. Perhaps giving a systematic account of how such perspectives relate and are associated with language is beyond the scope of semantic theory, as it is now conceived. But in the cases concerning tense or indexicals like ‘yesterday’ that Higginbotham discusses, it seems to me that associating at least some aspects of the original speaker’s perspective with the language of the report—in particular the meaning of the reporter’s indexicals—may not derive from much beyond applying a specifically linguistic competence to a relatively standard, known contextual parameter—the time of the original utterance. Then, as Higginbotham intimates, semantic theory should concern itself with the relation between the language and the perspective. But contrary to Higginbotham’s terminology, the perspective (at least the character or standing meaning of the indexicals) is linguistically preserved. And this fact supports a sense in which they are contextually “expressed” (not merely “shown”) by the reporter. The information can be recovered purely by application of linguistic competence. In such circumstances I think that there is something to be said for counting the ordinary, referentially used meaning of the original speaker’s indexicals as the ascriptive meaning or ascriptive content of the reporter’s indexicals (which simultaneously have their own literal meaning, referentially used). It is an open question, I think, how much of the speaker’s perspective—how much of the concepts or Fregean senses thought through his
382
Tyler Burge
language—is to be treated by a semantic theory that concentrates on truth-conditions and specifically linguistic understanding. I think that a satisfying investigation of language must sometimes go beyond the bounds of linguistic meaning, or linguistic understanding, conservatively construed. Taking Higginbotham’s discussion in the direction that it naturally points suggests to me that more should be treated than recent semantic theory has allowed itself. Notes 1. For some earlier discussion of these issues, see my (1977) and (1983).
References Burge, Tyler. 1977. Belief De Re. Journal of Philosophy 74: 338–362. ———. 1983. Russell’s Problem and Intentional Identity. In Agent, Language, and the Structure of the World, Tomberlin (ed.), pp. 79–110. Indianapolis: Hackett.
19
Concepts, Conceptions, Reflective Understanding: Reply to Peacocke
Chris Peacocke’s interesting essay is broadly persuasive and broadly congenial to things I have emphasized about incomplete understanding and rationality.1 It takes these matters in a direction that I have not carried them. It is also congenial to things that I have written about the way the nature of a person’s mental states can depend on his relations to his environment. Peacocke’s essay investigates the explanation of reflective understanding. He asks how we are to explain an individual’s improvement on his understanding of terms or concepts like chair or limit. He connects these issues in original ways to questions about how we are to explain the dawning of understanding of simple logical or mathematical truths that rests on the basis of an understanding of their component concepts. This is a rich, provocative essay. I will not be able to do justice to it. What I want to do is endorse the spirit of the main proposals about explanation, support the letter of some of them, and raise some questions about points that I have doubts about or that I think may need further development. I What seems to me right about the spirit of the proposals is the idea that one can explain many cases of reflective acceptance of “conceptual truths,” and the application of incompletely understood concepts, in mentalistic terms. The proposals are also attractive in their appeal to unconscious, “implicit” mental structures—though I will return to the question of what “implicit” should mean here. The idea that these explanations are not only mentalistic explanations but are explanations that bear on understanding our rationality also seems fundamentally on the right track. Finally, the idea that not all implicit conceptions that explain the use of concepts are correct conceptions seems to me fundamentally right and a consequence of the principles that underlie anti-individualism. What is an implicit conception? I take it that implicit conceptions are unconscious psychological conceptual structures that explain our ability to apply concepts to cases, or to realize that principles involving concepts are true. They explain such applications and principles even though we cannot easily explain to ourselves how we recognize the examples as instances of a concept, or how we explain the dawning of realization that such principles are true. I take it that implicitness is supposed to be compatible with, and in fact demand, psychological reality. And implicitness of a conception is compatible with its being either explicitly or implicitly represented (p. 137). I am not fully clear about what explicit and implicit representations are. I suppose that the issue turns on whether there are neurally realized syntactically structured tokens whose syntax corresponds to the form of conceptual structures that implicit conceptions have. If something like this is what is meant by explicit representation of an implicit conception, it seems to me correct that whether implicit conceptions are explicitly (though unconsciously) represented can be left open.
384
Tyler Burge
But I am still not clear what the implicitness of an implicit conception is supposed to consist in. Peacocke seems to indicate that implicit conceptions are conceptions at the subpersonal level (p. 384). I take the subpersonal level to be a level that is not only not conscious, but is not accessible to introspective or reflective consciousness and must be gotten at only theoretically. This is true of the basic grammatical structures underlying our linguistic competence and the information-processing structures underlying our perceptual experience. But elsewhere Peacocke takes implicit conceptions to be difficult but not impossible to make explicit through reflection. This makes it look as if implicit conceptions are real personal-level conceptions, just ones that are unconscious and relatively difficult to articulate in consciously available judgments. My guess is that both sorts of “implicitness” might be relevant to different aspects of explanations of the psychological facts. Perhaps there are even more than two types of unconscious structure here. A further question I have about the psychological aspect of Peacocke’s proposal centers not on the appeal to implicitness but on the presumption that the mentalistic elements in the relevant psychological explanations are always conceptions associated with the relevant concepts. It seems to me likely that in many cases, the relevant explicit explicative judgments—and improvements through reflection on previous explicative judgments— will be derived through unconscious mentalistic processes that do not use an unconscious explicative conception as material in their transformations. Rather, for example, they may use perceptually stored material, which has not been unified even at any “implicit” level under some explicative, conceptualized principle. The storage may be in the form of purely perceptual judgments about cases. Such perceptually stored material might be used together with certain (“implicit”) inductive principles or principles governing relevant similarities to form explicative judgments at the explicit level. Take the chair case as an example. I discussed this case at length in my (1986). On Peacocke’s account, those who can arrive at an explicitly articulated definition of ‘chair’ carry around an implicit “definition” of ‘chair’ that is often presumably fully correct. I am inclined to think that often we lack such a definition not only at the explicit level of readily accessible conscious judgment, but at any implicit level (both the truly subpersonal level and the unconscious, personal, hard-to-access levels). What enables us to arrive at correct definition is partly the memory of many instances that we have judged to be chairs, or at any rate, dispositions derived from such judgments. So we have stored hard-to-access memories of perceptual judgments (or dispositions to new judgments based on these prior judgments) of things as ski-lift chairs, deck chairs, living room chairs, and so on. One might ask how we made these initial judgments (unifying the different perceived examples under the concept chair), if we did not have a guiding definition. The answer is that usually we are just told that a ski-lift chair is a chair (at some time in the dim past), or that a deck chair is a chair, without being guided by some antecedent conception of
Reply to Peacocke
385
what makes them all chairs. We may never have used even implicitly a conception to include the ski-lift chair under the concept—though we do include such chairs under the concept. It does not seem necessary that once we have judged a ski-lift chair to be a chair, we already find (implicitly) a unifying explicative conception that explains what is essentially chairlike in both living room chairs and ski-lift chairs. We may carry only the notion that both are to be sat on, and are to be differentiated from stools, benches, love seats, and sofas—but no conception that yields necessary and sufficient conditions. We may, it seems to me, simply carry unconsciously the perceptual memory of the look of a ski-lift chair together with the unconscious memory that we categorized it as a chair, or simply the present disposition to categorize it as a chair. When we try to form an explicit reflective explicative conception of what chairs are, we may simply use offline—that is, unconscious—“implicit,” inductive principles to arrive at our explicit explicative conception of our concept chair. We may remember the ski-lift chair and realize that it is a counterexample to an explicit explicative conception that held that chairs must have legs. I think it unlikely that such an explicative conception must always be already formed at some unconscious level. So in such cases, it is not conceptions that are “implicit.” Rather, it is inductive principles together with a range of examples that are unified under the concept— though not under any conceptualization, or explication, of the concept. We need not, of course, always work with memories of instances that have actually been categorized under a concept, or with dispositions associated with past categorizations. We may be driven by general unconscious similarity principles from actually categorized cases to include other merely hypothetical cases as well, without having—even at some implicit level—a unifying conceptualization of the cases that is specific to the concept being explicated or conceptualized. An explicative conceptualization may first emerge explicitly, as a product of unconscious processes, at either subpersonal or hard-to-access personal levels, which make use of intentional material that is both more specific and more general than the explicative conceptualization specific to the relevant concept. One could project from one jade sample to the next. One might remain open to the idea of a unifying account of the similarity, and lack any general defining conception. Such a conception might become available only through empirical research, by geologists or philosophers. Here the limits may be set partly by what people have actually applied the jade concept to. Possible similarly looking and feeling minerals, other than jadeite and nephrite, may not count as jade just because they are not in the appropriate actual sample classes. I find Peacocke’s account of the standard model of arithmetic very attractive. But a point similar to the one just made may apply, in more complex form, even to the Leibniz/Newton limit case. It seems to me a stretch—and at any rate, not obviously correct—to think that Leibniz and Newton shared as a stable part of their unconscious repertoire an implicit but fully formed version of the Weierstrass explication. What seems to me more likely is that
386
Tyler Burge
they had a mastery of the basic calculus techniques, aided perhaps by some geometrical sense of approaching a limit on a line, which yielded correct answers in specific applications. Again, they may have had a not fully conceptualized sense of mathematical similarity, which may or may not be conceptualized into a principle, even implicitly, that explains their ability to project to the cases. Whether the implicit conception is fully formed seems to me open to investigation, even assuming that a definite concept, that of limit, is sharply grasped. The concept is sharply grasped insofar as one applies the concept to exactly the right cases. But what guides the application of the concept might not be purely an implicit conception or rule, but a combination of rules of thumb, paradigm cases, and a sense of mathematical similarities. That is, incomplete conceptualization of a definite concept that is being thought with may be present at both explicit and implicit levels. The individual’s ability to get the examples right may be explained by a combination of mental abilities that do not fall at just the level of a correct conceptualization, implicit or explicit. It seems to me doubtful that implicit conceptions explain all the phenomena that Peacocke is concerned to explain. The individual also may not, at the explicit level, make all the right judgments about examples, yet may still grasp a definite concept that includes those examples. It may be that general principles of mathematical practice and rationality can be seen, retrospectively, to warrant inclusion of certain cases under a given concept, even though the individual expert may be disposed to misjudge those cases in individual instances. I think that this case may be illustrated in the early history of the concept of set, when limitative prejudices blocked natural generalizations for at least some experts. I think that the early disputes over the axiom of choice can be seen, at least in some instances, in this light. It is certain that some of those disputes derived from mathematicians having different concepts, while using the same term ‘set’ to express them. Some had the modern iterative concept of set. Some had a concept closer to the modern concept of class. But some of the disputes seem to have stemmed from objections to the axiom of choice that were driven not by a noniterative concept of set, but by a sense of a need to limit the proliferation of sets by a closer epistemic control on their postulation than the axiom of choice provided.2 I think that mathematicians with such philosophical views as these had the iterative concept of set, but made mistaken judgments about what counted as a set (rejecting consequences of the axiom of choice), because of philosophical prejudices that interfered with what has come to be seen as sound mathematical practice. At any rate, it seems to me that it would be a mistake to think that the implicit mental structures that explain explicit judgments must themselves always be complete. Sometimes we depend on others. Sometimes we depend on a combination of examples, an unconceptualized sense of similarity, and principles at the wrong levels to count as conceptions associated with the concept at issue. Sometimes the limits of a concept are determined
Reply to Peacocke
387
partly by similarity principles but partly by whatever happens in actual fact to account for the types of samples that are actually counted as examples. This anti-individualist element in our mastery of concepts combines with the points about the role of nonconceptualized psychological elements in our use of concepts to indicate that the actual correct application of concepts we have is guided by more than implicit conceptions that we associate with the concepts. II This psychological point seems to me to bear significantly on our understanding of the epistemology of reflection. Most traditional accounts of a priori reflection have, like Peacocke’s, assumed that a conception of the rule associated with application of the concept is—at some implicit and unconscious, or subliminally conscious level—fully formed in any individual that has the concept. Reflection was seen as just a matter of bringing to consciousness and fully articulating a conception or rule that is already present in the mind. Peacocke notes, as traditional rationalists tended not to, that the implicit conception will sometimes be at a subpersonal level. So it will not be accessible to reflection or personlevel inference. This seems to me correct. But I think that the distance from traditional conceptions of implicit mastery of concepts goes further. Kant sometimes writes as if he identifies concepts with functions of unity, or rules, for holding cases and subordinate concepts together.3 He saw the rules as produced and, at some level, as grasped by the understanding. But the twentieth century has seen an emphasis on the role of instances or examples in individuating concepts. Wittgenstein, Kripke, and Putnam, in their different ways, have indicated that concepts are not entirely fixed by background rules, principles, or descriptions that the individual has grasped. I have tried to develop this idea in my own ways. Implicit in this emphasis on the role of instances or examples in individuating concepts is, I think, a recognition that some of the ways that we have for projecting from examples are stored nonconceptually. The perceptual system and nonconceptualized senses of similarity may guide our projection from central instances to which a concept applies, to further instances. Only with reflective conceptualization of rules that codify these lower-level abilities do we arrive at conceptions that are adequate to explain our application of certain concepts. Suppose that concepts are not always backed by implicit conceptions—conceptualizations—that explain our application. So reflection on the nature of our concepts is not always a matter of bringing to consciousness a conceptualized rule that guides their application. It is part of the formation of such a rule. But such formation cannot be seen as formation of the concept. For the concept is already fully formed, thought with, and even
388
Tyler Burge
correctly applied. To think with the concept and even to have a sharp grasp of it, in the sense discussed earlier, it is not necessary that one have an associated descriptive rule for its application. This point is relevant to understanding philosophical thought experiments. For example, in the Twin Earth cases it is commonly assumed that we all have, unconsciously in mind, descriptions, rules, or principles that guide our use of such concepts as arthritis, chair, sofa, edge, water, and so on. It is conceded that these are hard to formulate. Still, it is often assumed that they are always somehow implicitly complete and present in the individual’s mind. I believe that this assumption is mistaken. Some philosophers who have sought to refute the thought experiments have added to this view the further mistake of identifying concepts with the supposed underlying descriptions, rules, or principles. They presume that their formulations are themselves environmentally independent. (This presumption is itself unargued and in many cases unconvincing.) They then maintain that these descriptions, rules, or principles guide the individual’s use of a term or concept, regardless of the environment. This line commonly makes further mistakes. But the one that interests me here is the assumption that when an individual has a concept, there is always in the individual an associated, fully formed implicit conception that explains the application of the concept and applies to the same instances that it applies to. Reflection on the nature and application of our concepts seems to me a more complex enterprise than making conscious certain conceptualizations, rules, principles, descriptions, or definitions that are already implicitly in the mind and associated with the concept. The epistemology of reflection is, I think, correspondingly more complex than traditional philosophy has represented it. It is natural and traditional to see reflection on the nature of concepts as warranted apriori. Let us suppose that ‘apriori’ means ‘independent in justificational force from sense-perception or sense-perceptual belief’. Suppose that we are reflecting on the nature of our concept chair. Suppose that we recognize that the concept applies—and long has applied—to ski-lift chairs, without legs; but we have no conceptualized principle, even implicitly, for projecting from our standard cases of chairs to these special cases. Thus the connection between the standard cases of chairs and the ski-lift chairs is simply that we have stored perceptual similarities between the cases and have accepted long ago someone’s calling a ski-lift chair (or ski-lift chairs) a chair. Thus we are supposing that when we recognize that a ski-lift chair is a chair we do not derive this recognition from a principle or conception that includes specification of the properties of ski-lift chairs that make them chairs. Rather, we make use of a memory of the case and a confirmatory sense of perceptual similarity and generalized conception of functional similarity between the cases. Thus, we note that the ski-lift chairs have a flat seat, accommodate one or two persons, and function to be sat on. But our acceptance of the case is
Reply to Peacocke
389
driven not by a generalization but by our memory that that sort of object—or perhaps even a particular remembered object—counts as a chair. Is such a memory warranted apriori? That depends on the nature of the warrant for present tense claims like that sort of object, which is used in ski-lifts, counts as a type of chair and that object counts as a chair, where the claims are taken to have a role in specifying or teaching the nature of the concept, not merely ordinary statements of fact.4 These questions are complex and multifaceted, and I will not try to answer them here. I want simply to raise them. I do not think that these questions arise for all cases of reflection on the nature of our understanding of concepts. It seems to me that sometimes Peacocke’s account of implicit conceptions is correct and fully adequate. I just want to point to what I regard as further complexities that warrant our attention. III Let me turn now to issues about explaining rationality, with particular reference to Peacocke’s historical points. His invocation of Leibniz and the rationalist tradition seems to me entirely appropriate. I agree with most of what he says in this section. I have here three reservations, two of them of perhaps only minor significance. As Peacocke notes, Leibniz holds that axioms are evident as soon as their terms are understood. He objects that Leibniz here overstates the ease of the discovery of axioms. This may be so. But it is important to remember that Leibniz meant by ‘axiom’ not just any proposition that might be taken as a starting point for an axiomatic theory. He had the old Euclidean conception of axioms as truths that are basic in a justificational order and that are sufficiently simple that, assuming they are fully understood, there is no need to argue for them or derive them from anything else. I think such a conception has more to be said for it than most modern philosophers presume. Moreover, I think that Leibniz set a very high standard for understanding of the terms. I think that he meant complete, explicit understanding, not merely the sort of understanding sufficient to use the terms and reason with them. So it is not clear to me that there is any mistake in Leibniz’s view that given full understanding, one finds basic truths evident. I think the view is virtually definitional of the traditional conception of an axiom. It seems to me that Leibniz is right about a narrow class of truths that might be counted axioms in the old sense. Simple truths of logic and arithmetic seem to me to be so basic that if one understands them, one realizes that they are true. No argument for them could provide them with a justification that adds force that is not already present in understanding them. The second reservation concerns Peacocke’s account of what it is to be clear but not distinct. He notes that according to the traditional view, an idea or concept is clear for a
390
Tyler Burge
person if the person can use it to recognize instances of the concept. He quotes Leibniz’s remark that having a distinct idea lies in the ability to enumerate separately the essential characteristics that distinguish the things the idea applies to from other things. He explicates distinctness in his framework. He claims that a thinker with a distinct idea is one who has succeeded in achieving an explicit formulation of the implicit conception he had when he had only a clear but indistinct idea. This claim seems only approximately true. For as Peacocke himself remarks, implicit conceptions can be incorrect explications of an individual’s idea or concept. In such cases, making them explicit would not be making them distinct. Further, suppose that I am right that even implicit conceptions that are correct as far as they go may not have conceptualized the full correct explication of a concept. So merely making such incomplete implicit conceptions explicit will not suffice to make a concept distinct in the old fashioned sense— that is, give it a full, correct explication. Incompleteness of explicational understanding, or indistinctness, can hold at the implicit as well as the explicit level. A third reservation, the one that interests me most, has to do with Peacocke’s discussion of Frege and the rationality of accepting logical axioms. He points out that Frege gave arguments for his axioms from semantical-looking background assumptions for the truth of most of his axioms. Yet he regarded the axioms as self-evident—that is, recognizable as true independently of justifying them through derivation from other truths. In my view, although Frege did not philosophize about this apparent oddity, he knew exactly what he was doing. What is more important, there is a philosophically tenable resolution of the apparent conflict.5 First, it should be noticed that the arguments Frege gives that have his axioms as conclusions are fully explicit. So the apparent conflict arises independently of any distinction between implicit and explicit levels. Second, it is important to distinguish between justification of a sentence’s expression of an axiom and justification of the axiom itself. Frege believed—and I agree—that the fundamental truths of logic are not strings of symbols, even though strings of symbols express such truths. Frege is interested primarily in the truths, but he is simultaneously setting out and justifying his logical symbolism by showing its adequacy to express the underlying truths. The arguments in question, in Basic Laws, bear on both the symbolism and on the logical truths, but in different ways. Close analysis can separate out these points. But Frege is fairly loose in his book about slipping back and forth from semantical discussion about symbols to substantive exposition of his truths. Peacocke also writes sometimes of derivations of sentences (p. 146) and other times of derivations of the logical truths themselves (p. 146). Of course, both are at issue. But the bearing of the semantical arguments is different in the two cases. It seems to me that the semantical arguments do provide, in a straightforward way, a justification for axiom-expressions and for formal symbolic expressions of the rules of
Reply to Peacocke
391
inference. They show rigorously that the symbols are adequate to express what we recognize as axiomatic logical truths and valid inference rules. But the bearing of the arguments is different on the logical truths and inference rules themselves. It seems to me incontestable that Frege would not have regarded any arguments from language as being capable of justifying language-independent logical truths and rules of inference. In fact, however, most of Frege’s semantical-looking arguments for the axioms make no essential reference to language at all. Still, they are arguments from truth-conditions (associated with thought contents) to the language-independent axioms. And the very fact that they are arguments with the axioms as conclusions is already puzzling, since the axioms are supposed to be self-evident and not in need of proof. What is going on, and what can be shown from comparing several of Frege’s texts, is that Frege regards the arguments not as justifying the conclusion but as articulating the content of the conclusion. The arguments are not intended to provide justification for belief in the conclusion by deriving it from premises belief in which is antecedently justified. For, as he says, the conclusions are not in need of proof or justification. Frege means by ‘proof’ a deductive argument that provides justification from self-evident basic truths as premises.6 The articulation of content that the arguments provide is simply an articulation of understanding of the conclusion. So it remains possible for him to hold that the content of the conclusion carries all the evidence needed to recognize the conclusion as true: That is, the conclusions of the arguments, the axioms, are self-evident. Understanding the axioms justifies one in believing them; but full, explicit understanding itself requires an ability to articulate the truth-conditions of the contents that are understood. I want to elaborate this point a bit, since I think it correct. Three background points are important. One is that any understanding, even understanding of simple logical truths, requires mastery of complex inferential connections. This is one of Frege’s greatest contributions to philosophy, and something not present with anything like the same clarity in Leibniz. The point requires that understanding of both terms and propositions is not independent of acceptance of principles and inferential connections. So understanding a logical truth is associated with arguments using the terms or concepts embedded in the truth. Second, as Frege also famously maintained, what is understood places conditions on truth. So understanding the sense of a sentence, or understanding (grasping) a thought, requires understanding its truth-conditions. Putting the two points together, understanding a logical truth in a fully articulate way requires an ability to articulate through argument its sense or truth-conditions. Third, it is doubtful that arguments to at least some very simple logical truths—and at least some simple rules of inference—from their truth-conditions are arguments that provide any additional justificational force to that already involved in really
392
Tyler Burge
understanding those logical truths or rules of inferences. The premises and rules of inference used in the arguments from the lines of a truth table are not any more strongly justified, or more evident, in at least many simple cases, than the logical axioms that they are used to derive. Whatever role Frege’s arguments from truth-conditions play, they do not provide any extra rational support or warrant for their conclusions beyond what is involved in understanding the conclusions. The conclusions are, in these cases, at least as rationally evident as the premises. I believe that in Frege’s senses of ‘justification’ and ‘proof’, the relevant arguments are not justifications or proofs of their conclusion. Some things in Peacocke’s exposition are congenial with this point. He writes of the arguments as explaining our rationality in accepting the premises. Such explanation might be distinguished from justification that adds justificational force to our warrant in accepting the conclusion. He also writes of explaining the evidentness of the axiom. Again, such explanations might be seen as articulations of our understanding, not as justifications from more basic premises. Moreover, with certain qualifications I will not try to state here, I join Peacocke and follow Frege in holding that what is understood is to be explicated in terms of truth-conditions. But I do not believe that Frege’s arguments for the simplest logical axioms, where these axioms are understood to be thought contents or propositions, were meant as justifications of them. The arguments do not add any justificational force not already involved in complete understanding of the content of the propositions in question. Frege was surely fully aware of this fact. I think that the view that Frege seems to have held is correct: Understanding the axioms requires an ability to give the sorts of arguments from truth-conditions that he gives in articulating the intentional content of the axioms. But it is the understanding of the axioms themselves, not a justification of them from antecedently understood principles governing truth-conditions, that is fully sufficient for being warranted in believing them. In this sense the axioms are evident in themselves and not in need of justification or proof from other truths. Frege’s great contribution is to indicate that because of the dependence of understanding on mastering inferential connections, “in themselves” is a more subtle and complex notion than most traditional philosophers realized. A thought has a definite content, but that content is logically connected to other contents. And thinking with the content necessitates being able to make some of the logical connections. Understanding the content (whether minimally or in some deeper way) requires understanding some of the inferential connections. But once understanding is achieved, once one has and understands the relevant contents, justification does not require: deriving it from other contents from which it inherits justification. The position seems to me to derive support from the fact that the semantical arguments seem intuitively to add no warrant to full understand-
Reply to Peacocke
393
ing of the conclusion. In fact, they seem to be just argumentative articulations of such understanding. It is important to bear in mind that there are different possible arguments with the axioms as conclusions that articulate understanding of the axioms. These differences reflect the fact that no one argument is necessary for understanding. Let us consider A Æ (B Æ A). There is, for example, the type of articulation Peacocke outlines. This type appeals to a metaperspective and explicitly uses a concept of truth that is a predicate of thoughts. Frege gives two other types of arguments, both in the object language: The first one starts with a step we would formalize as ~(A Æ (B Æ A)) Æ (A & (B & ~ A)). By commutativity and associativity of conjunction, conjunct-elimination, noncontradiction, modus tollens, and double-negative removal: A Æ (B Æ A). The second argument (which Frege gives in Begriffsschrift) is that if A, then A regardless of whether B, for any B. All of these arguments make use of an understanding of the truthconditions of the conditional. None seems to do more than articulate what is involved in full understanding of the axiom. And none seems to rest on principles that are clearly more fundamental, or more obviously true than the axiom, or are self-evident in a way that the axiom itself is not. I am happy to concede that in the case of each principle or rule there is an explanation of the rationality of accepting it. But I am not inclined to think that the most basic principles receive any genuine or needed epistemic warrant—or positive justificational force for believing them—through the arguments. The fundamental warrant for believing them lies in understanding their content. They are self-evident. It is just that any such understanding has to be accompanied by an ability to explain the rationality of accepting the proposition through discursive argument. I remain attracted to a conception of rational justification of the simplest principles of elementary logic and arithmetic that is broadly similar to the conception shared by Frege and Leibniz. According to this conception there are certain basic truths and rules of inference. Understanding these suffices to warrant belief in them. No argument for them can yield warrant for belief that adds force to the warrant already yielded by understanding them. Such warrant is maximal. The truths “do not need or admit of” any further justification. (See note 6.) In this sense they are basic and self-evident. Arguments of the sort
394
Tyler Burge
we have been discussing elicit the fact that there are more basic truths and rules of inference (all equally basic) than are needed to develop logic, at least first- and second-order logic. So the foundation is overdetermined. This feature of overdetermination makes contrasting this sort of foundationalism with a sophisticated coherentism a subtle and perhaps uninteresting terminological matter. I think, however, that warrant is best seen as lying in the understanding of the relevant truths. There is no need to rest the warrant on “coherence” with other truths. But the understanding unquestionably involves abilities to make inferential connections with other truths. And there is, of course, a necessary coherence among the logical and arithmetical truths. Like Frege, I combine this foundationalist justificational structure with a pluralistic, coherentist conception of understanding. Understanding requires an ability to make inferences to and from the understood intentional contents. Any of various inferential patterns of connections among thoughts can suffice to yield understanding of fundamental logical truths or rules of inference. For basic truths the connections hold both between thoughts and rules of inference that are equally fundamental, and between self-evident ones and some less basic ones. More needs to be said about the distinction between the role of understanding in rational acceptance of the principles and the role of argument in articulating the understanding. The naturalistic and holistic tendencies that we have inherited from Quine tend to ignore or blur such a distinction. But anyone who refuses simply to reduce understanding or grasp of a thought-content to some particular pattern of inferential abilities—while still holding that understanding requires some such pattern of inferential abilities—is in a position to draw it. More also needs to be said about the traditional notion of basic, self-evident truths. Many have doubted that there remains any use for the idea that some thoughts in logic and mathematics are, from an epistemological or psychological point of view, maximally basic. Even those who do not embrace the empiricist view that logic and arithmetic depend for their justification on their role in empirical science commonly emphasize that there are so many “axiomatizations” of formal theories that finding basic ones is a pointless exercise.7 This negative attitude often derives from mixing up the modern conception of axiom with the traditional one. The modern conception is centered on what is taken as basic in a particular presentation of a theory. Certainly the variety of possible “axioms” in this sense is endless. And certainly some things that are “axioms” in the modern sense are in no way self-evident or epistemically self-sufficient. Some of the axioms of high-level set theory, for example, are certainly not self-evident. They are not even derivable from self-
Reply to Peacocke
395
evident truths. Not all of mathematics is derivative from self-evident truths. It does not follow that there are no epistemically fundamental truths in logic or mathematics. In fact, it seems quite obvious that there are truths that are for us epistemically more basic than others. And it seems to me arguable and likely that there are truths that are maximally basic, in the sense that no argument for them could add to the warrant inherent in understanding them. Whether there are any basic truths that are basic for all finite rational beings is a further question. I am sympathetic to the idea that some truths and rules of inference are necessarily basic for every finite rational being that has the relevant concepts. But I leave this an open question. It is a question that would require extensive and subtle development. Whatever the answer to this question, it seems overwhelmingly likely that, as regards arithmetic and logic, there are broad similarities among human beings in what count for them as simpler and epistemically more fundamental truths, and in what count as more complex and epistemically derivative. It seems to me likely that the line between what is justificationally basic and what is justificationally derivative may be blurred in some cases, for some individuals. And it seems likely that where the line is drawn will vary with the individual. Some individuals may include more truths as basic than others, depending on the depth of their understanding. The old-fashioned picture of a rational order of truths is out of favor. There are certainly many obstacles to bringing such a picture back into focus. Even a rationalist picture that is much more modest and more qualified than the traditional versions faces numerous obstacles. I think, however, that such a picture is worth developing. Notes 1. Substantially this essay was given as a reply to an earlier version of Christopher Peacocke’s essay at the APA in Pittsburgh in April 1997. 2. For a detailed account of the controversies over the axiom of choice, see Gregory Moore (1982). 3. Kant, Critique of Pure Reason (B92–93, A103). Ultimately, Kant’s view is much closer to mine than these passages suggest. 4. Of course, I agree with Quine (1966) against conventionalists that it is both a claim of fact and a claim about meaning or concepts. Obviously these questions are close to issues that Kripke (1972) raises about the contingent apriori. 5. For a detailed discussion of these matters, emphasizing the historical point of view, see my (1998). 6. The phrase derives from Leibniz, New Essays in Human Understanding (1705, 1765, 1989), e.g. IV, ix, 2; 434; see also Frege, Foundations of Arithmetic, section 3. This notion of proof is discussed at some length in my (1998) and my (2000). 7. For a contrary view about all of this, see Cherniak (1986). None of what follows is meant to do justice to Cherniak’s position. I am not, however, persuaded by his arguments. For an interesting discussion of related issues, see Evnine (1999).
396
Tyler Burge
References Burge, Tyler. 1986. Intellectual Norms and Foundations of Mind. Journal of Philosophy 83(12): 697–720. ———. 1998. Frege on Knowing the Foundation. Mind 107: 305–347. ———. 2000. Frege on Apriority. In New Essays on the A Priori, Peacocke and Boghossian (eds.). Oxford: Oxford University Press. Cherniak, Christopher. 1986. Minimal Rationality. Cambridge, MA: MIT Press. Evnine, Simon. 1999. Believing Conjunctions. Synthese 118: 201–227. Kant, Immanuel. 1781, 1787. Critique of Pure Reason. Guyer and Wood (eds.). Cambridge: Cambridge University Press. Kripke, Saul. 1972. Naming and Necessity. In Semantics of Natural Language, Davidson and Harman (eds.), pp. 253–355. Dordrecht: Reidel. Leibniz, G. W. 1705, 1765, 1989. New Essays on Human Understanding. Peter Remnant and Jonathan Bennett, (trans. and eds.). Cambridge: Cambridge University Press. Moore, Gregory H. 1982. Zermelo’s Axiom of Choice: Its Origins, Development, and Influence. New York: Springer Verlag. Quine, W. V. O. 1966. The Ways of Paradox. New York: Random House.
20
Epiphenomenalism: Reply to Dretske
Fred Dretske’s essay presents a tantalizing and engaging combination of agreement and difference. We agree on anti-individualism, or what he and others call “externalism.” We agree on realism about intentional content, on the real existence of intentional states and events, and on the explanatory and causal relevance of such states and events, individuated nonindividualistically. I agree with most of his invocations of the relevance of function in intentional explanations. And I see his application of intentional explanation to animals that lack language, and his discussions of artifacts like gauges and computers, as congenial and illuminating. We differ over whether materialism is obviously true and whether naturalistic reductionism is a worthwhile enterprise. More to the point of the main lines of his essay, we differ about whether epiphenomenalism is a serious threat and how to go about showing its impotence. This difference over the status of a threat from epiphenomenalism is related to a lesser difference over sociology. Dretske writes that he thinks that he and I may be the only ones who believe that the content of intentional states can be both externally individuated and causally relevant to the explanation of behavior. I enjoy and value Fred’s company, but I do not feel nearly so isolated or beleaguered as he does. I think that, despite the widespread discussion of epiphenomenalism and the posing of “worries” about it, nearly everyone except for the occasional wide-eyed revisionist believes that intentional explanation is in part causal explanation and is relevant to the explanation of mental events and behavior. And I think that anti-individualism has carried the day among the vast majority of philosophers who have understood and reflected on it. The intersection of these groups constitutes a large majority of philosophers who have thought about these subjects. Of course the sociology of philosophy is of no great importance in the short run. But I believe both of these positions will prove very stable. This judgment is the historical analogue of my view that much of the recent discussion of epiphenomenalism, in the light of anti-individualism, is conducted in an atmosphere of make-believe—a mock phobia that is dramatized mainly to make some metaphysical model seem like desperately needed salvation from impending psychic or philosophical disaster. My view is not that philosophical issues about mental causation or mind–body causation are uninteresting or pointless. Dretske’s and others’ out-of-context citations sometimes make it appear that that is my view. But it isn’t, nor have I ever expressed such a view. There are difficult and worthwhile issues in this area. My negative attitude toward the ways these issues have been discussed has two components. The first concerns the antecedent make-believe posture toward the discussion, that treats epiphenomenalism as a serious threat rather than as a test or limit position that any sound view must avoid. The second concerns prevalent assumptions—centering on materialist metaphysics—about what sorts of considerations must be dominant in an explanation of why epiphenomenalism is false.
398
Tyler Burge
Let me begin with the point about make-believe. Dretske’s comparison of the problem to Zeno’s paradoxes seems to me to be in the right spirit. No serious person, even a philosopher, thinks that we cannot traverse a space. But there can be philosophical interest in explaining what is wrong with Zeno’s reasoning that says that we cannot. That is the spirit in which epiphenomenalist doubts should be treated. I think that despite his wonderfully relaxed philosophical manner, Dretske’s rhetoric of “worry” and “epiphobia” that needs to be “quieted,” seems to me to lose touch with the Zeno paradigm. I think that we know that beliefs and desires are causally efficacious as such with as approximately much certainty as we know that we can traverse a space. Even granting the materialistreductionist starting point that Dretske assumes, it seems to me that epiphenomenalism can be safely presumed to be false. The discussion should be pursued in that spirit, not in a spirit of angst, which I cannot help but find artificial and histrionic. Dretske produces quite a range of examples that are supposed to raise the epiphenomenalist worry. But taken as potential threats, none of them seems to me to be intellectually powerful. Relevant disanalogies with the mental state/event cases are in each instance immediately obvious. At most, the examples prod one to think more clearly about mental causation. Contrary to what Dretske suggests, we do not disagree about the causal relevance of the property of being an aunt or uncle. The comparison with the relational property of being an uncle suggests only that some relational properties are not causally relevant, not that all relational properties (much less all nonrelational properties whose nature presupposes the existence of relations) are causally irrelevant. Causally relevant relational properties are abundantly invoked in the natural sciences. No specific reason is given in this example for thinking that mental properties are not causally relevant. We have overwhelming grounds to presume that they are. Similarly, no one thinks that being a twenty dollar bill or being an H-particle or being a quarter is causally relevant to the transactions that are explained in terms of underlying physical properties that these properties happen to be correlated with. Such examples show at most (see below) that some higher-level properties are causally irrelevant, not that all are. Even assuming that all mental properties and events are correlated with specific underlying physical properties, which is highly doubtful, nothing specific to mentalistic causation has been cited that would suggest any error in the common presumption that mental states/events as such are causally relevant in a wide variety of causal transactions. Citing these cases is somewhat like saying that since a number cannot traverse a space, maybe physical bodies cannot either. Dretske notes that on the anti-individualist view it is assumed that at some level of description the twins, in the standard thought experiments, will be undergoing and even doing some things that are the same. One can use this as a ground to believe that antiindividualist explanation is causally irrelevant only if at least two conditions are met: First,
Reply to Dretske
399
explanations of the ways, particularly intentional and function ways, in which the twins are different are causally irrelevant. Second, where the twins do the same things, there can be no differences in mental causes due to differences in the causal structures and sequences in the two environments. I see no reason to think that either of these conditions is met. Dretske lays the first condition aside in his motivational section, without explaining why he does so. He does not discuss the second. But one can raise one’s arm for different reasons. It is not clear why the causes of raising twin arms in the different environments cannot also differ, given that the arm is raised for some purpose, whose eventual effects will have different properties. Assuming that these two conditions obtain really amounts to assuming that antiindividualism about mental states/events is incompatible with the causal relevance of intentional aspects of those events, which is exactly the point at issue. Dretske’s most specific formulation of his worry is as follows: “The worry is not that externally individuated beliefs do not play a role in the explanation of behavior. . . . The worry is that their role is dispensable—that the same explanatory results can always be achieved without mentioning those aspects of belief that fail to supervene on the believer” (p. 156). I think that this “worry” has only to be stated and reflected on to see how wildly implausible it is. No one has made a remotely persuasive case that nonmentalistic explanation, or some artificial form of explanation that postulates only “narrow” content, can achieve all the same explanatory results that mentalistic explanation does. Although in some particular cases, one can hive off particular external aspects of belief (e.g., the re in de re belief), no one has even begun to provide an explanatory replacement for the basic intentional explanatory elements in mentalistic explanation. Those elements and many of their explananda are individuated nonindividualistically. Nor is there strong reason to think that intentional, mentalistic explanation is, throughout, noncausal. Dretske has a tendency, in expounding the “worry,” to collapse the distinction between the mental states/events as individuated by their content—which are the explanatorily relevant properties—and the relational facts that are necessary to the having of that content— which are normally not themselves part of a causal explanation, or causally efficacious, at all. If this point is kept clearly in mind, the worry about the wholesale dispensability of mentalistic explanation seems to me vanishingly thin. There remains worthwhile inquiry into the specific ways that mentalistic causal explanation differs from underlying natural-scientific explanation, which prevent mentalistic causal explanation from being dispensable. Dretske has made substantial contributions to this inquiry. I think that his distinction between behavior, as a process consisting of a sequence of events, and motion is insightful and relevant. And I think that his appeal in his (1988) to structuring causes, though not in my view a full account of the causal relevance of reasons, is also a significant contribution. It is notable that despite the
400
Tyler Burge
materialist cast of his account, neither of these contributions depends essentially on his materialist framework. Even apart from Dretske’s particular account of behavior as a sequence or process, mentalistic explanation of simple events like raising an arm, which are not specified in intentional terms, is normally explanation of an act, not a mere happening. Even if the doing is to be counted identical, sometimes, with a happening, the aspect of the event that is explained by mentalistic explanation—even when it is not an aspect directly involving intentional content—is different from the aspect of the event (mere motion) that is explained by a physiological or purely physical explanation. Further, as we agree, often it is the intentional content of behavior that is explained. These reasons are just a small sampling of obstacles for physical-causal explanation’s replacing mentalistic explanation. Dretske makes much of a need to provide a metaphysical account of causation in light of the supposed unreliability of explanatory practice in signaling the causal relevance of mental properties. I do not deny that metaphysical models are worth having when we can get them. But I think that his motivation of a need for them by denigrating the role of explanatory practice in guiding us to the causal structures of the world is misconceived. It depends on too loose a conception of explanatory practice. Of course, there are scattered rule-of-thumb explanations in common sense that can be recognized on a moment’s reflection not to pick out the causally relevant properties, at least relative to certain explanatory enterprises. But I have in mind more fundamental aspects of explanatory practice in everyday life and in psychology. Consider the widely used Coke machine example. Being a quarter is epiphenomenal relative to the mechanical causation of a Coke machine. Recognition of this point is surely part of even moderately reflective aspects of explanatory practice. I think, however, that certain explanations of functional aspects of the “doings” of even a Coke machine can be causally relevant, along the lines of Dretske’s structuring causes. I agree with him that the functions of the machine are not autonomous and certainly not mentalistic in this case. I agree that the case is not a good model for the causal elements in mentalistic explanation. The issue of whether an object’s being a quarter is a causally relevant property does not seem to me very important. I am not convinced, however, that the property of being a quarter is properly seen as epiphenomenal with respect to all causal transactions. Suppose that we allow substitutes— counterfeit slugs—to count as having the causally relevant properties in particular cases, as most causal explanations of the functional aspects of events do. Such substitutes would succeed in causing particular instantiations of functional properties via other nonmonetary causally relevant properties. The slug would cause the Coke machine to deliver a Coke and change. The substitutes would succeed in explaining, relative to background assumptions about the functional relation, the same instantiations of functional properties as the
Reply to Dretske
401
“noncounterfeit” causes would. But if the substitutions were made wholesale from the beginning in all cases, the substitutes would not succeed in causing the functional aspects of the explanandum. If there were no normal causes—instances of noncounterfeit input coins—there would be no Coke machines yielding Cokes and change. One can explain any given motion of the machine without appealing to the property of being a quarter. But one cannot fully explain the functional properties of the motions—the fact that the machine provides a Coke and change—without appealing to monetary properties. This seems to me already to differentiate the Coke case from the standard epiphenomenalist model. But as noted, I agree with Dretske that the Coke machine example would be a poor model for mental causation. Dretske’s examples of a $20 bill’s effect on the cashier’s visual system are similar. It is easy to convince ourselves that the visual properties of the $20 bill, not its monetary properties, are what are causally relevant to purely physicalistic aspects of the causal transaction with the cashier’s visual system. The purely physically described transactions focused on do not need explanation through the higher-level properties. The explanations for higher-level properties that we are commonly invited to consider are offhand, rule-ofthumb explanations that are not embedded in a systematic explanatory system. It is easy to imagine dispensing with such explanations, at least relative to the aspects of the explananda that Dretske has in mind. But even in these cases, it is far from obvious that all causal explanation can dispense with the higher-level properties. To fully understand the situation, one needs to think about what the account of vision is supposed to do. What are the causally relevant properties? In one sense, anything that causes the same retinal stimulations. Any particular instance of a visual representation can, in a sense, be causally explained as well through appealing to a counterfeit bill, or a quantum-accident look-alike, as by the $20 bill. Considering only such cases would be to follow the method of Dretske’s argument “anything like the $20 bill that was sufficiently similar would have the same effects.” But we want to specify the cause as what is seen too. Which properties of what is seen are causally relevant? At a certain level, just the “visual” properties—properties of color, shape, texture, shading, distance, and so on. But there are perceptual concepts that derive from training. These include more than visual properties in the most psychologically conservative sense. People visually recognize horses, cancer, money. To account for the representation of these properties, we need to cite the properties at a certain level in the explanation of visual perception. Other lookalike properties would serve in the causal explanation of any given (false) visual representation of normally seen properties. But look-alike properties could not suffice to explain the acquisition of visual information about the original properties. Nor would they explain the function of the visual system, or higher-level representational systems, in representing the original properties. A systematic causal explanation of even individual instances
402
Tyler Burge
of perceptual representations will require appeal to higher-level properties in the generalizations, as normal causes. Moreover, mentalistic explanation that makes use of mental events as causes must be able to explain particular events. Dretske’s own account of mentalistic explanation pays too little attention, in my opinion, to the role of mentalistic properties in causal relations between particular events. Particular acts and events are caused by events with mental properties—sensations, perceptions, decisions, judgments. We have every empirical reason to think that such explanation is not only genuine but irreducible and ineliminable. Even the initial points made with respect to the Coke machine and cashier cases fail to apply in the case of mentalistic properties generally. We have no purely physical explanations of our detailed physical movements, much less of our intentional activity and behavior. We do not even have such explanations for the activity of the higher animals. We have massive empirical reason to believe that psychological causal explanation, with appeal to causal efficacy of instances of psychological kinds, is indispensable for explaining a wide range of important phenomena. A satisfying metaphysical model of mental causation would be welcome. But no such model is needed to show that such causation is real. An undercurrent difference between Dretske’s approach and mine lies in our attitudes toward the epistemic status of materialism. He begins with a materialist metaphysical model, assuming that as known. He holds that model fixed in trying to explain why epiphenomenalism is not true. I have indicated that I think that epiphenomenalism is no serious threat even given this starting point. I have no objections to metaphysical inquiry in this area. Where we can get deeper understanding from metaphysics, we should, of course, get it. I do not, however, accept as obvious or fundamental the assumption of the materialist and mechanist model that Dretske starts with. I think that whether materialism is a correct metaphysical view of ourselves and the mind depends on among other things exactly what the relations are between thoughts and the underlying physical material. The mere fact of some necessary dependency will not suffice for materialism. The relations must be identity, composition, or something similar. I think that we do not have satisfying accounts of exactly what these mind–body relations really are—in particular whether they are forms of material constitution. Just maintaining that mental events and properties need a material substrate—or need to supervene on such a substrate—does not suffice to support materialism. For materialism is committed to specific relations—identity or material-composition relations— between the mental items and the physical items. Assuming a metaphysical model with these commitments seems to me to go far beyond anything that our actual explanatory
Reply to Dretske
403
practice warrants. In fact, it seems to me that materialist metaphysics has succeeded in obscuring mind–body causation more than it has illuminated it. I have not been persuaded by the blizzard of reasons given to insist that mental events are material events. All the materialist accounts of phenomenological events and even of propositional attitude events seem to me to have a whistling-in-the-dark character. Materialist metaphysics will be and should be an ongoing enterprise. But it should be taken as experimental and conjectural, not as the philosophical starting point that it became for so many philosophers in the second half of the twentieth century. No particular form of materialism has earned a status of orthodoxy. And a generalized commitment to materialism seems to me to owe more to pious hope, wild inductive projections, and caricatures of nonmaterialist alternatives, than to rational argument or empirical support. I believe that the next century is ripe for fresh ideas on the mind–body problem. References Dretske, Fred. 1988. Explaining Behavior. Cambridge, MA: MIT Press/A Bradford Book.
21
Qualia and Intentional Content: Reply to Block
I agree with most of Ned Block’s illuminating and ingenious essay. I think that understanding consciousness and the nature of phenomenal aspects of experience is an important and difficult enterprise. It is not an enterprise to which I have contributed much. I hope that investigating the relations between Block’s projects and mine will be fruitful. I In my (1979) I held that color concept-possession is subject to nonindividualist thought experiments. I also maintained that in the thought experiments the individual’s “non-intentional phenomenal experience” remains the same (p. 78). I would now issue cautions about applying the term ‘experience’ to phenomenal awareness, and would prefer the phrase ‘nonintentional phenomenal aspects of our mental lives’. But with these qualifications, I still hold these positions. I think that having certain ordinary color concepts is constitutively dependent, partly, on bearing relations to the colors in a broader environment. And I think, with Block, that there are nonrepresentational qualitative mental properties. I did not argue for the latter view. The representationist view of sensations and their properties—the view that representational properties are the only properties of sensations1—has long seemed to me to resist belief. Somehow it has not resisted the belief of various other philosophers. But then philosophers sometimes overcome all natural resistance. Block’s ingenious and varied arguments against representationism clarify and deepen my sense of its extreme implausibility. A value of philosophy lies in scrutinizing and coming to understand what one tends to take for granted. Block’s work realizes this value very richly. What is surprising to me is that representationism has to be argued against so conscientiously. Not only does it seem to me beyond belief, I know of no forceful arguments for it. The defensive maneuvers that Block considers and criticizes seem to me to derive from philosophical strategy—an attempt to make a programmatic metaphysical ideology, encompassing a scheme for confronting the mind–body problem, “come out right”—rather than from insight into the matters at hand. I find it unfortunate that this habit, contracted in the hoary metaphysical and theological traditions, has reappeared—often joined with vague appeals to introspection—to mark currents in contemporary philosophy. The “argument” for representationism from the diaphanousness of introspection seems to me to have no force. As an argument meant to undermine what Block calls phenomenism, it is beside the point. As an argument for representationism, it is question begging. Of course our attention is normally focused on the redness of the tomato when we are seeing a tomato. That is because physical objects and their properties are the primary
406
Tyler Burge
objects of our perceptual systems. But no one, except a philosopher who held that we only “see” sense data and have to infer the existence of physical properties from these “data,” would hold otherwise. Hardly anyone nowadays who believes in qualia rests the view on the sort of flat-footed, vaguely specified introspection that the argument from diaphanousness invites us to engage in. Diaphanousness, properly understood, and the primacy of perception of physical properties can be granted by nearly everyone. Moreover, as Block points out, considering the situation when we close our eyes, or considering elements of awareness common to veridical perceptual experience and misperceptions or hallucinations, shifts the import of diaphanousness in ordinary perception. Diaphanousness is a normal but not ubiquitous phenomenon. The thought experiment that centers on diaphanousness does nothing to show that someone who can conceptually distinguish the tomato’s redness from the characteristic quality of his perceptual experience of the tomato’s redness is unable, with appropriate prompting, to attend to the latter. It seems to me that doing so is no great feat. Some philosophers, including defenders of qualia, have maintained that qualia should be seen as “postulated” by some philosophical or other explanatory theory.2 This view seems to me quite mistaken. The characteristic quality of the experience is not a “theoretical” postulation. We are directly aware of qualia. Even children and nonsophisticates are directly aware of them. Distinguishing them from the objects of perception, and attending to them, does depend on having and knowing how to apply immediately the relevant concepts of qualia. Such concepts almost always enter one’s conceptual repertoire after concepts of physical objects of perception. Probably young children lack such concepts. But the qualia themselves are present and are conscious sensational elements in perception all along. Children start with a phenomenal consciousness of qualia, and develop an ability to introspect them and discriminate them from the physical properties that they play a role in representating, once they gain the relevant concepts. It seems to me desperate and mistaken to maintain that qualia are theoretical postulates. Such a view seems almost as strained as representationism itself. Of course, attention is one thing; awareness, another. The argument that centers on attention gives no reason to think that the quality of the perceptual experience is not an element in one’s awareness when one is having an ordinary perception of a tomato. One need not have conceptualized the awareness, or be able to make it a discriminated object of attention, in order to have the awareness. But it seems quite possible to conceptualize it and make it an object of attention so as to discriminate it from the physical objects and properties that are the objects of vision. The representationist may disagree with all this. But the argument from diaphanousness provides not the slightest reason to do so. The argument from diaphanousness is hardly better than one that would claim that we cannot be perceptually aware of the surface of an object because no matter how hard we
Reply to Block
407
try to inspect the surface, we end up inspecting the object. Of course, there are disanalogies. Everyone agrees that we see, and have from the beginning different perceptual representations for, both surface and object. Nearly everyone agrees that we see physical color and do not see—or otherwise perceive in the ordinary sense of ‘perceive’—qualia. And we do not have perceptual or conceptual representations from the beginning that take the qualia as objects. We gain conceptual representations for our phenomenal awareness only later. But the issue is not about seeing or perception. It is about attention and awareness. The fact that the qualia and the redness of the tomato are indistinguishable to attention in an introspection that abstracts from conceptualizing a difference is not at all relevant to whether there are two properties one can be aware of—either nonconceptually, or through attentive, introspective, conceptually guided representation and discrimination. An individual can through perception be aware of an object’s surface without attending to the surface or distinguishing it from the object. Perhaps an individual could be aware of an object’s surface without even having conceptualized it as a surface. But in the absence of conceptualizing it, the individual is not going to be able to attend discriminatively to a difference between seeing the object and seeing its surface. Once the distinction is conceptualized, one can attend to one or the other property in visual experience. Thus the argument from diaphanousness does not come to grips with what Block calls phenomenism. It argues from premises that are, or should be, accepted by the opponent of representationism. It does not provide an illuminating argument from these premises. It simply takes them to defeat or embarrass phenomenism. The conclusion amounts to assuming that phenomenism is false. So it begs the question as a positive argument for representationism. I am sympathetic with Block’s belief that there are qualia that have no representative function. The representationist view of orgasm would be apt for future joke books about analytic philosophy. The less risqué example of visual blur—which has been discussed by others, including Block elsewhere, and Loar in this volume—also seems telling. The fuzziness in near-sighted perceptual experiences does not normally represent blur in the object. Nor need it represent anything else, such as some defect in the visual system. It is a defect or a noise in the medium of representation, not an application of a perceptual category by the perceptual system. It can be taken to indicate a defect in the medium. But it need not be so taken, or used in any way. That is, neither the system nor the individual need compensate for the noise or do anything with it. Insofar as this is so, I think that there is no case to be made that it must have representational content or function. So granting that such noise is an unattended element in one’s awareness does not require granting that the awareness need have a role or content in some representational system, much less be an object of representation. Even if it is used to indicate something about
408
Tyler Burge
the visual field, the phenomenological quality of blurriness is the relevant property that is represented as a feature of the visual field. Appealing to the representation of this property to support representationism is to give up opposition to the phenomenist. No phenomenist claims that qualitative properties cannot be represented, or that they are not properties of something—the visual field, sensory awareness, the medium of representation, the mind. Taking the awareness always to involve representation of some further property (e.g., some physical defect in the visual system or blur in the object) has no functional plausibility. For the distinction between the awareness of the noise and something that it represents is not one that always has a role in the cognitive economies of unsophisticated perceivers, especially those who do not have to compensate for the “noise.” It is simply not true that individuals with visual blur are always aware that their eyes are unfocused, or that there is a defect in their visual system. The blur might be an element in their phenomenal awareness, without their being aware of it as blur, much less having any awareness that, with respect to it. When one changes the angle of one’s vision as one looks through progressive bifocal lenses, one’s visual system does not represent a nearby object as changing from having clear markings to having blurred markings. Visual representation of blurred objects (where the blur is either in the objects or in the intervening physical conditions) is normally, I think, phenomenally as well as functionally different from blurred vision. One’s visual system represents the surface well and then badly. There is a particular phenomenal quality to the bad representation, which need have no intentional characteristic in itself. One has no inclination to find blurriness on the edges of the relevant object. It would be incorrect to take blurriness as some other objective property that happens not to be instantiated in any object. Nor is the blurriness normally about defects in the system. As I have noted, the individual may not make any use of the blur, even compensatory use. The blurredness is a feature that constitutes a subjective distortion in the mode of representation of objective properties. Awareness of such qualitative “noise” is not accounted for by regarding it as representation of something further. It is not accounted for by claiming that it has no other properties than its representational ones. Such awareness normally does not have representational properties, at least not representational properties of something further. Even when the qualitative character is noticed and taken to be noise, its representational properties tend to consist in its being used to represent itself. The normal reference is to the visual blur—the qualitative property itself. Only individuals who use the noise in some intentional way for some positive purpose (for example, compensating for the defect)— and not even all such individuals—use it as a way of referring to something beyond the qualitative feature.3
Reply to Block
409
II Block develops a number of ingenious inversion cases. I agree with him that the reply to the inversion cases that appeals to failures of memory is desperate and in effect question begging. I do not think that these implausible responses have an exclusive claim on the title “externalism about memory.” The content of a memory is normally fixed by the content of what the memory preserves. Normally, the states whose content memory preserves are externalistically individuated. I am broadly sympathetic with Block’s view that reflection on aspects of inversion cases tell against representationism. But I am interested in certain details of a particular case that he elaborates. I would like to discuss his response to the second main line of objection that he considers to the inverted earth case. This is the worry that the intentional content of the representations of color does not shift when the individual’s environment changes. Let us suppose that a person is transported to inverted earth with inverting lenses. Let us further suppose that no matter how long he stays, the person is correct in noticing and remembering no phenomenal change. The person’s original conceptual and perceptual representations of blue are causally tied to physical instances of blue in the person’s original environment. The causal ties are linked to recognitional abilities, which are cued by the qualitative nature of the experiences. That qualitative nature remains unchanged. And the person’s memories of blue objects in the original environment, and his recognitional abilities relative to the original environment, remain unchanged—even though the person is transported to a situation in which those abilities cannot without correction be successfully exercised. It seems to me therefore that it is doubtful that the person loses the original conceptual or perceptual representations of blue. Those representations make recognitional use of the phenomenal quality of the original experiences of blue. Indeed, this tie is almost surely innate in us. The qualitative nature of the experience probably gets its meaning and reference not through the individual’s experience of blue things but through the perceptual system’s having evolved to register blue in things. The system does this as a response to the occurrence of perceptual states with the relevant qualitative feature (or a limited range of qualitative features, which might differ among individuals with the same basic perceptual equipment), given the normal association of this feature with blue objects in the evolution of the system. If this is so, it seems doubtful that a change in the individual’s environment, even for a very long period, would suffice to change the normal environment that is relevant to fixing the representational content of the phenomenal aspects of his experience. The normal environment seems to be fixed by the evolution of the perceptual system that the individual has, not by the context of the individual’s learning history.
410
Tyler Burge
So it is not clear that normal observational representations of colors are conceptual or perceptual representations that can be lost through resettlement, no matter what inversions occur in the environment. Thus the yellow sky will continue to look blue and in a standard sense be represented as blue even for a transfer who remains on Inverted Earth and accommodates to it. In memories of how the sky used to look, for example, the imagined or occurrent perceptual experiences will still apply to blueness.4 Of course, the individual could learn by other means that the new sky is yellow, not blue. The individual could then learn to take his phenomenal experience to indicate yellowness. If this learning were embedded thoroughly enough in the individual’s automatic perceptual reactions, I suppose there might develop a sense in which sky “looked” yellow to the individual. But this sense would go through a route of inference—however automatically it might come to be sublimated and embedded in the individual’s cognitive procedures. So it would not have the same innate base purely in perceptual categories that underlies and warrants the perceptual use of the phenomenal experience to represent blue. So the cognitive way in which the sky looked blue would remain different from the way in which it “looked” yellow. I see no difficulty in allowing both sorts of representation in the same individual. Thus the relative constancy, under environmental changes, of the innate representational content of phenomenal types need not force a denial that the transfer will gain a new meaning and reference for experiences with the relevant qualitative character. The individual’s perceptual system could come to function in normal perceptual situations—as opposed to memory situations, or situations that made essential use of past recognitions of blue objects—to use perceptual states with the relevant phenomenal quality as indicators of yellowness, not blueness. The individual’s belief system could be geared to the “yellow” representational content, not the “blue” one. Whether yellow-representing or blue-representing states, or both sorts of states (all with the same phenomenal character) occur would, I think, depend on the details of the case. Here I think we should probably acknowledge perceptual and conceptual representations of yellowness in the shifted individual that are analogous to representations of color in a color-blind person. A red-green color-blind person frequently uses perceptual representations that are usually appropriate in normal beings for brown or gray to represent green. But they represent that color in what is, at least for the perceptual system that they share with other members of the species, a deviant way. They cannot discriminate the color purely by the look in a good light. They need to rely on supplementary, contextual cues. Suppose that the shifted individual gains a perceptual and conceptual representation of the color of the inverted-earth sky as yellow. His representation of yellow makes use of a phenomenal quality that is innately appropriate to—and still means, in some uses—blue. Then the representation of yellow has the same referent as, but I think different intentional
Reply to Block
411
content from, our perceptual and conceptual representations of yellow. His representations of yellow are analogous to color-blind representations of green. The individual may not always have to rely on contextual cues in quite the way color-blind people do. That is a disanalogy. But like the color-blind person, he will have had to rely on testimonial or inferential activity to establish the referent in the first place. These points do not undermine Block’s conclusions. There may be ways to produce examples in which the evolutionary ties are somehow broken between Earth and Inverted Earth. (And I think that it is an empirical hypothesis that the individuation of representational content of our color perception depends on evolutionary history.) But they seem to me very tough and durable ties. I am doubtful that merely adding that the person who shifts to Inverted Earth is afflicted with amnesia will suffice to deal with the objection. For the supposition is that the individuation of states representing color depends fundamentally on the evolutionary history of the perceptual system, not on anything about the individual’s memory. It is still well to reflect on constancies of phenomenal character despite shifts in representational content within these more complex cases, even though the cases do not involve simple inversions. The individual who gains representations that refer to yellow, but who represents yellow by having phenomenal elements innately associated with blue, has a different intentional content (a “color-blind” content) for those representations. His intentional content is different from the content that we have in our visual representations of yellow—even though he and we share a referent. It is different because it is fundamentally associated with different cognitive procedures, psychologically different modes of recognition. He has a different referent from our normal visual representations of blue. But it has the same phenomenal character. The shifted individual’s visual representations of yellow are associated with a different functional organization, inasmuch as the representation of yellow does not derive from simple exposure to yellowness through individual or evolutionary history, but through learning that the sky is yellow by other means and then using those means to affect his interpretation of his perceptual contents. Still, the example does seem to be one in which the same phenomenal qualitative features are associated with different reference and different intentional content. So, as Block claims, representational content changes without any corresponding change in phenomenal character. I see no plausibility in the idea that the phenomenal character originally associated with blue perceptions must undergo some change under the shifts. Block mentions the possibility of appealing to Swampmen analogues that lack any evolutionary history. So instead of moving a given person from Earth to Inverted Earth, one considers a person on Earth and a Swampman with compensating lenses on Inverted Earth. I think that such an appeal is legitimate, though I find that such cases are less compelling
412
Tyler Burge
to many people than cases involving ordinary persons. The idea underlying the appeal is that the phenomenal qualities of experience are more nearly dependent on their biochemical substrate, whereas the representational characteristics associated with phenomenal qualities are dependent on functional usage and on certain relations to an appropriate normal environment beyond this substrate. These particular relations are not relevant to biochemistry—whether or not biochemistry is itself nonindividualistic. I am sympathetic with Block’s conclusion for a variety of reasons. But I believe that the complexities that I have been discussing are interesting in themselves, quite apart from their role in arguments over representationism. The point I wish to emphasize is that the individual who represents a color (yellow) by having phenomenal qualia that are normally evolutionarily associated with blue would normally have a different intentional content from someone who represents the same color (yellow) by having phenomenal qualia that are normally evolutionarily associated with that color. The reference is the same, but the intentional visual content—the analogue of meaning or sense—is different. Since the individual’s recognitional ability is so closely tied to and cued by the qualitative character of the experience, and since mode-of-presentation commonly types such cognitive abilities as perhaps innate recognitional ability, the intentional content or mode of presentation will commonly differ, intrasubjectively, if the qualitative character of the experience differs. So in the cases of color representation, I grant that sameness of phenomenal character is compatible with difference in intentional content, and indeed difference in reference. But I doubt that sameness of intentional content—all the way down, at the finest intrasubjective level of individuative grain—is compatible with difference in phenomenal character. Only sameness of referent is compatible intrasubjectively with discernible difference of phenomenal character. The perceptual intentional content, the perceptual modes of presentation, will commonly be in some way, at some level, different, if phenomenal character is different. The reason for this correlation is not that phenomenal character is reducible to representational content. There is nothing essentially representational about phenomenal character. At least, there is nothing essentially representational as of anything beyond the phenomenal character itself. The reason that sameness of intentional representational content is not compatible, or at least is not obviously compatible, with difference in phenomenal character is that phenomenal character is commonly an individuating element in the individual’s recognitional ability. So it is an element in individuating the intentional representational content of an individual’s perceptions.5 A given phenomenal character could in principle have been associated with any of various intentional representational contents, inasmuch as it could be involved in any number of nonindexical recognitional representations and recognitional
Reply to Block
413
abilities with different referents. For nonindexical conceptual and perceptual representations, difference in reference entails difference in concept and percept. III Block’s suggestion that our phenomenal characters, in perception of a given color, vary widely, even among normal humans, is an interesting one. The empirical ground for his suggestion is at first blush striking. Suppose that his suggestion is correct. How might it affect individuation of the intentional aspects of perceptual representation? I think that one simply must bear in mind that representational content is a kind one of whose roles is that of typing psychological and other cognitive abilities. We find out what kinds there are by investigating and reasoning about different sorts of cognitive transactions and abilities. There are a variety of such transactions and abilities. I think that there is likely to be a variety of overlapping kinds of phenomenal character that individuate overlapping kinds of perceptual intentional content, especially when intersubjective typings are at issue. The finest differences of phenomenal character within an individual might be relevant to explaining different possible discriminations that the individual might make. Thus, at the most fine-grained level, there are as many perceptual contents as there are phenomenal discriminations—not necessarily pairwise discriminations—that the individual uses, or can use, in perception. But when it comes to normal perceptual tasks outside of discriminative tests in a matching exercise, some of these possible discriminations might be irrelevant. That is, different phenomenal “shades” might be grouped together and treated, in a given type of psychological activity by the individual, as the same phenomenal property for individuating a certain kind of psychological ability, hence intentional content. This happens in our perception of something as red, as opposed to a narrowly discriminated and quickly forgotten shade of red. The same point applies to individuation of intentional content across individuals. The most dramatic case of this sort is our treating the blind person and ourselves as sharing a concept of red. It is obvious that at a certain level, the psychological abilities of the blind person and of the normally sighted person differ in the representation of the redness of a flower. They have different concepts of red, inasmuch as some of the sighted person’s concepts are associated with recognitional abilities through perception. But there remains a sense in which the two can share a thought that the flower is red. In this case, there is a concept of red that does not depend on a visual phenomenal character at all, shareable by different individuals. The relevant shared psychological abilities are associated with a complex of shared background information about red things, and a shared network of
414
Tyler Burge
causal relations, through interlocution, back to red things. But the blind person lacks the sighted person’s perceptual ability to recognize red things. So there remains an intentional content, or better an array of intentional contents, that they do not share. I think that the same sort of point may well apply in cases of spectrum shift within a given community. Slight or even medium-level phenomenal differences among individuals might instantiate instances of the same phenomenal kind, where the kind is relevant to understanding shared perceptual-recognitional abilities, especially those that hinge in any way on communication or other shared activities. For example, suppose that qualia caused by seeing some standard sample of red differ between males and females. There will certainly remain a sense in which the sample looks red, even looks the same shade of red to the different individuals. Here ‘looks red’ is understood in terms of being in the same perceptual/representational state a normal human being is in when, in normal favorable circumstances, he or she is looking at a red object.6 Differences in the “shades” of qualia associated with the standard sample seem to be irrelevant for typing verbal and even discriminative similarities between the individuals. They are irrelevant to individuating this sense of ‘looks red’. Here our understanding of ‘looks red’ cuts through the presumed qualitative differences between the individuals’ qualia and counts them as instances of the same representative type.7 Whether they might count as instances of the same phenomenal type for purposes of intersubjective typing, in something like the way different shades count as falling within a given generic color, seems to me to be an open question. Groupings of phenomenal qualia seem to me to be open to psychological investigation. But I am sympathetic to the view that intersubjective typing might group together qualitatively different qualia as instances of single qualitative genera. It is, of course, not obvious that the groupings will, in every case, for every explanatory enterprise, closely parallel groupings of objective colors. Such coarse-grained kinds would be compatible with the finer-grained kinds relevant to explaining the narrowest intrasubjective individual discriminations. Insofar as one considers the finest-grained discriminations of phenomenal character, it might well be that there are some spectrum shifts among members of a given species, in regard to reference to a given objective color property. Block’s empirical evidence certainly suggests such shifts. Insofar as different individuals have different phenomenal characters—different qualia individuated in the finest-grained way—in perceiving the same color, there is no question of who is right and who is wrong, assuming that there is comparable sensitivity and comparable functional efficiency among the individuals. Right and wrong hinge on the referent, the objective color. Different individuals can have different modes of perceptual presentation in perceiving the referent. Some of these modes might be standard for a given species. Others might be statistically somewhat deviant. And there may be a great variety
Reply to Block
415
of modes within a given species. These are empirical matters. Block is certainly right that these matters cannot be reduced to questions about the meaning of words, or to questions about the representational functions of qualia. Notes 1. It is assumed that reflexive self-representation by the sensation does not count. 2. See Shoemaker (1994a), pp. 291–314; and (1994b), pp. 21–38. 3. I assume throughout that we are considering representation as something more than the sort of informationcarrying exhibited by tree rings. Even in cases where the noise is taken to be a sign of defects in the visual system, the qualia does not always have the representational content that indicates the defect. Sometimes the individual takes the qualia to be an object of reference from which he infers some defect in the system—as a natural sign of defectiveness rather than itself having intentional content that means the defect. These are delicate matters. The key point is that such qualitative noise does not always have intentional content, particularly when it is not used for any representational purpose. 4. Block notes this point and indicates that it has “more than a little force.” I believe that he can handle the objection to his overall strategy. I am sympathetic with his overall strategy, but want to discuss principles of individuation that are relevant to this particular example. 5. I leave open whether it is part of the intentional content. I am inclined not to regard it strictly as a part of conceptual content because of abstract uses of the content, removed from recognitional uses. 6. I agree with Block and others that there is also a sense of ‘looks’ according to which red will ‘look’ differently to males and females, given the above stated supposition. Whether we have a clearly established public use for ‘looks red’ that follows this usage is more doubtful, especially if Block is right that qualia differ widely among normal humans’ perception of red. 7. It is compatible with this point that differences in phenomenal quality determine differences in intentional content, at a certain fine-grained level of individuating intentional content for intrasubjective and even intersubjective purposes. It is even compatible with this point that in cases of more coarse-grained individuation, phenomenal character (finely individuated) plays some looser role in the individuation of representational content.
References Burge, Tyler. 1979. Individualism and the Mental. Midwest Studies in Philosophy 4: 73–121. Shoemaker, Sydney. 1994a. Self-Knowledge and Inner Sense; Lecture III: The Phenomenal Character of Experience. Philosophy and Phenomenological Research (54)2: 291–314. ———. 1994b. Phenomenal Character. Noûs 28: 21–38.
22
Mental Agency in Authoritative Self-Knowledge: Reply to Kobes
Bernard W. Kobes’s creative and closely reasoned paper connects the performative element in basic self-knowledge to knowledge of what one will do in intentional action. I think this an illuminating comparison. In both cases, the knowledge derives from intellectual control over what one is doing, and in both cases the warrant derives partly from one’s status as a rational agent. Some knowledge of what one will do is partly constitutive of being an intentional or rational agent. Some performative knowledge of one’s propositional attitudes is partly constitutive of being a critically rational agent.1 I There are differences between the cases. When one engages in intentional physical action, knowledge of what one will intentionally do depends partly, in a relatively immediate way, on matters outside one’s control. One can know that one will pick up a fork, when one intends to do so. But the knowledge depends on the veridicality of one’s perception of the fork and on one’s using the perception to guide one’s hands in the normal way to grasp the fork. Knowledge of one’s thoughts in relevant performative cases is not hostage to the brute contingencies on which perception and physical action rely. One can also know that one will raise one’s arm. Here the perception of an object is not necessary in quite the way it is in the case of picking up a fork. One’s perceptual relation to one’s arm may be through proprioception. Proprioception is subject to fewer brute contingencies than visual perception. Still, I think that the dependence on a sense renders one liable to brute errors. In knowledge of what one will intentionally do through physical action, one is generally subject to such errors. One’s intentions can be thwarted in ways that undermine true belief, hence knowledge, of what one will do. Some of these ways do not undermine one’s epistemic warrant, nor do they involve malfunction of one’s cognitive capacities. That is to say, brute error seems always to be possible. By contrast, authoritative self-knowledge of one’s own mental states and events, including self-knowledge that involves a performative element, is not subject to brute error. Self-knowledge that involves a performative element includes a broader range of cases than those that are logically self-verifying. The thought I am hereby entertaining the thought that writing requires concentration is logically self-verifying. I call such cases pure cogito cases. The logic and meaning of the intentional thought content requires that if the content is thought, it is true. But there are cases that normally involve selfverification and a performative element, even though the intentional content does not strictly entail that a thinking of the thought be self-verifying, or even true. These are impure cogito cases. For example, one can conceive of a case in which one commits oneself to the whole content of the thought I am hereby thinking (in the sense of committing myself
418
Tyler Burge
to the view) that writing requires concentration without committing oneself to the truth of the thought that writing requires concentration, thus without making the whole content (I am thinking . . .) true. One might believe that one is committing oneself to the component thought, but one simply absent-mindedly thinks it through. But such a case would be highly abnormal, even pathological. The intentional content is such that its normal use requires a performative, reflexive, self-verifying thought. It is an (impure) cogito case only where the thought is performatively self-verifying. Or on writing out a check for a charity organization, one might think I intend (i.e., hereby form the intention) to give to Oxfam, where making the judgment about one’s intention is the formation of the intention. The same act constitutes both. Thoughts of this form are fallible. Performative cases of this sort also count as impure cogito cases. One can not only think through this content and leave it false. One can even judge-true the content (I intend [i.e., hereby form the intention] to give to Oxfam) yet fail to form the intention, thus leaving the judgment false. But such a judgment would again be pathological. The normal use is to form the intention performatively in the formation of the judgment. So the act of making the judgment makes the judgment true. In all of these pure and impure cogito cases, intellectual control is coordinate with the act of thinking a thought—whether forming a belief or making a judgment—that both commits one to the truth of the intentional content and carries out the thought that the content is about, thereby making the thought true. I believe that most performative cases of authoritative self-knowledge involve a reflexive element. Although I think that there is something to Kobes’s description of performative cases as ‘telic’, I am not sure that there is full agreement on these matters. Kobes’s explanation of this notion is somewhat metaphorical. But it seems to me that in most paradigmatic cases of judgments with performative elements, including basic selfknowledge, the relation is reflexive or reciprocal between judgment and subject matter, rather than unidirectional. For example, when I judge, in the performative way, I hereby form the intention to give to Oxfam, as I make the decision and begin to write the check, the judgment is normally reflexive. It normally constitutes, rather than being caused by, the formation of the intention. In performative cases, a mental act of self-attribution makes itself true.2 Of course, the judgment about the intention can be caused by an antecedent intention. The intention can form just beforehand; or it can be a standing state. (Or perhaps the causation can be simultaneous, but between distinct events.) I believe that self-attributions of such intentions and beliefs can be authoritative. But in most cases they are not, in my view, strictly performative. Since cases of authoritative self-knowledge in which the attributed propositional attitudes are occurrent thoughts or intentions that cause the self-attribution are similar to cases of authoritative self-knowledge of standing propositional attitudes, I
Reply to Kobes
419
shall discuss the two types later under the rubric of authoritative self-knowledge of standing states. Kobes holds (p. 204) that we have self-knowledge of what we are thinking even when we are thinking ordinary conscious thoughts about the world, and that in many such cases we are not thinking “explicitly” reflexive thoughts. I agree that self-conscious self-knowledge is present in many ordinary first-level thoughts about the world. But I think that there is a reflexive element in more such self-knowledge than most people realize. I do not know what he means by ‘explicitly’. We do not often verbalize such thoughts with ‘hereby’ or ‘in this very thought’. But I think that when we have authoritative propositional selfknowledge of what we are presently thinking when we are thinking about the world, particularly when there is a performative element in the self-knowledge, there is normally—or at any rate, very frequently—a reflexive second-order element in the logical form of the first-order thoughts.3 Although these performative or reflexive cases form a larger class than one might first think, I believe that they do not constitute the whole of what I see as authoritative self-knowledge. Kobes places heavy emphasis on the activity of beliefformation. I agree that agency is at the heart of our understanding of first-person authority. But I doubt that it can bear weight in just the way that Kobes or Descartes require it to. Here Spinoza seems to me to provide a salutary qualification on the Cartesian view to which Kobes’s emphasis is congenial. Descartes maintained that belief is always the product of an act—in fact a “willed” act—to assent to a proposition. Descartes seems to me right in maintaining that some instances of belief-formation are instances of a type of agency, and some of these cases are subject to considerations of intellectual responsibility. But I think that this model should not be fully generalized. Opposing Descartes, Spinoza maintained that belief is the default position, not an activity on a proposition that one noncommittally understands. He maintained, as a thesis of philosophical psychology, that belief is concomitant with understanding. Spinoza held that doubt and suspension of belief are acts. One can undo the initial nonactive default position. But formation of belief is, on his view, not an act. It is automatic if not checked.4 Spinoza’s view, in its fully general even if vague form, seems exaggerated. But he seems to me right in maintaining that the formation of a great number of beliefs, particularly perceptual beliefs, is not strictly an activity by the whole mental agent. Some of our authoritative self-knowledge resides in our self-attribution of such beliefs. These considerations help motivate my view that the performative model cannot fully explain first-person authority.5 Let us lay aside modular and otherwise inaccessible beliefs. There are still beliefs and other propositional attitudes whose formations do not constitute acts of ours. They seem rather to form in us. They may be part of a functional organization that is essential to being
420
Tyler Burge
an agent. But not all such beliefs that are accessible to self-conscious self-attribution are themselves the products of acts of commitment. At least, I can think of no natural sense of agency that applies to their formation. Most perceptual beliefs are of this sort. Many beliefs that rest on interlocution, especially in childhood, are as well. Beliefs that derive from perceptual beliefs by way of hard-wired inductive mechanisms—as opposed to active, person-level inferences—are also not in any obvious sense products of agency. Kobes sometimes speaks of nonmodular beliefs as “up to us.” There may be something in this. They make up a point of view that we as doxastic agents can claim as ours. Once we become critical reasoners, we are epistemically responsible for them. But it does not follow that they are products of agency. Rather, we may be responsible for revising them if counter-considerations arise, and we may have a parallel responsibility for their maintenance. It would seem to me mistaken to hold that all nonmodular propositional attitudes—or even all attitudes that we are potentially authoritative about—are formed through intellectual agency. Thus I think that our authority about occurrent and standing attitudes is not captured fully by appealing to the model of “hot” or “smoldering” telic self-knowledge. Nor do all cases of authoritative self-knowledge exhibit an ability to bestow the content on the propositional attitudes that form the topic of the knowledge. Often those attitudes have a nature and existence that is prior to and independent of authoritative judgments about them. Authoritative knowledge of our standing or occurrent perceptual beliefs, and of many of our other standing propositional attitudes, needs, I think, a broader account. Of course, in authoritatively attributing propositional attitudes to ourselves, we normally commit ourselves to those attitudes. Claiming them as one’s own may seem in effect to endorse them, at least when they are not viewed as foreign “objects” inside one’s self. And in those “foreign object” cases, the self-attribution is not authoritative. This may suggest that agency is pervasive in the propositional attitudes thus self-attributed. I think that this is an important point. But the truth of some authoritative self-attributions does not depend on these endorsements. In making some authoritative self-attributions, one is making a judgment that is not intended to be made true by one’s endorsing the attitude on the spot. It is intended to capture a stable attitude that was present antecedent to the self-attribution, and whatever re-endorsement that might involve. As example, consider: I believe that my sister is younger than I am. Such self-attributions are fallible. The ways that they are corrected— for example, by reference to past statements or behavior—show that endorsements of the first-level propositional attitude that are implicit in the self-attributions themselves are not in general taken to be sufficient to guarantee the truth of the self-attributions. Thus a full account of the specialness and authority of some of our self-knowledge needs to go beyond both Kobes’s telic model and beyond my paradigm of self-verification and
Reply to Kobes
421
performative acts in cogito cases, pure or impure. When I introduced my self-verifying model as a paradigm, I called the sort of self-knowledge that it encompasses “basic selfknowledge.” But I was fully aware, as Kobes recognizes, that there is a range of other cases of authoritative self-knowledge that does not exhibit self-verification. There is firstperson knowledge of sensations, of occurrent perceptual beliefs, of nearly all standing states that predate the formation of judgments or even knowledge about them. There are certain cases of memory. There is knowledge of some of one’s feelings or emotions. And so on.6 II My view in “Individualism and Self-Knowledge” was that there are features of self-knowledge—other than self-verification—that are dramatically and paradigmatically realized in what I called the basic cases and that provide a key to understanding the whole range. I have yet to carry out this strategy fully. But I have taken some further steps.7 And I still believe that it is a viable and promising enterprise. I want to stress that “Individualism and Self-Knowledge” was not intended as a full account of authoritative self-knowledge. Not only does it indicate that the paradigmatic “basic” cases do not constitute the full range of authoritative self-knowledge. It is focused on the semantical or logical role of attribution of intentional content in self-knowledge and on the easiest cases of a self-verifying attitude relation. It does not discuss the notion of warrant, in any depth, even for the basic cases. The point of the article was to raise, and provide an initial response to, an apparent problem about the relation between authoritative self-knowledge and anti-individualism as an account of the nature of propositional attitudes. Although I made a number of comments about differences between authoritative self-knowledge and perceptual knowledge, I concentrated on pointing in a direction for an account of authoritative self-knowledge. Criticism of the paper for being “thin” as an account of authoritative self-knowledge is off the mark. The points, which Kobes cites others as making, about differences between the basic cases and other cases of authoritative self-knowledge were points that I anticipated and in several cases explicitly noted myself. I thought that those were matters to be taken up later. I still expect to deliver on the promissory note. I hope that this reply will advance matters. Kobes’s discussion of extensions from the paradigmatic cases is sympathetic and discerning. I think that he is right that some cases of authoritative memory and some cases of authoritative knowledge of standing states can be understood in terms of mechanisms preserving traces of earlier states known or knowable in the paradigmatic ways.8 I think that these mechanisms should be seen as part of the rational apparatus that is constitutive
422
Tyler Burge
of being a critical reasoner. As I have noted, however, I do not think that all of these cases of preservation can be assimilated to the act-paradigm that Kobes develops. III I turn now to the two objections Kobes discusses. His treatments of the Loar and slowswitching cases are well reasoned, and broadly plausible to me. My views differ in some significant ways, however. There are also some matters I would like to try to clarify. Kobes’s discussion of slow-switching cases makes some points in common with my discussion of such cases in my (1998b).9 In that recent paper I focused on unaware slow switching rather than knowledgeable slow switching. I emphasized the nonreflective role of preservative memory in knowledge of one’s past thoughts.10 Even in knowledgeable slow switching, I see no reason to think that the individual cannot remember the past thoughts, even if he cannot distinguish them from cohabiting twin thoughts. The individual is not forced to ask the question that Kobes and Boghossian focus on. He need not ask whether he was thinking about water or twater. In fact, where he asks this question in this way, I believe that he loses his authority in his application of memory (see Burge 1998b). The individual can instead rely on preservative memory to take up the content and attitude-type of the thought that he in fact thought at the earlier time. Relying on memory to individuate rather than to preserve would be a mistake. Such reliance in the switching cases could indeed undermine knowledge of the past. It would treat the remembered event as an object, rather than anaphorically—as part of a single point of view. In such a case, the memory would not be authoritative. If one uses information about having been switched between earth and Twin Earth to try, through memory, to discriminate earth- from Twin Earth-type thoughts, one loses one’s authority over one’s past thoughts. Errors deriving from such uses of memory are not naturally counted cases of forgetting. But if they are not so counted and are seen rather as unfortunate use of newly acquired information, then, as Kobes and I both point out, Boghossian’s “platitude” that if one forgets nothing one cannot lose knowledge is shown to be false. Indeed, the “platitude” is false for a wide range of cases, many of which have nothing to do with switching scenarios. New information, misleading information, can drive out old knowledge. So no pressure is generated on the anti-individualist by these means. I am uneasy about Kobes’s appeal to equivocal thoughts as the entire basis for his account of reasoning in the relevant switching cases. He is, I think, quite right to reject the idea that anti-individualism commits us to an unacceptable susceptibility to equivocation in deductive reasoning, even by the most rational reasoners. It may be that in some instances one thinks equivocal thoughts. Or as I would prefer to see it, one may think mul-
Reply to Kobes
423
tiple thoughts, on given occasions, without distinguishing them. It may be that Kobes’s idea is part of a necessary solution. But all of the reasoning cases that I know of can be accounted for by noting that even someone who has switched and has “twin” concepts in his repertoire is usually not using both concepts when he uses either of them. Rather, features about the cognitive context and cognitive point of a thought determine which of the concepts is employed. I refer the interested reader to a fuller account of the matter in my (1998b).11 Thus I do not accept Kobes’s assumption (p. 217) that because the switched individual is on Twin Earth and has the twin concept, he “cannot escape including an exercise of the Twin Earth concept” in any given thought in which he remembers something about the twin object back on Earth. This assumption suggests some magical effect that merely being on Twin Earth has on one’s thinking Twin Earth thoughts (once one has acquired them). I believe that, at least in the cases that have been discussed in the literature so far, one does not have to appeal to equivocal or multiple thoughts to block unacceptable results about reasoning in slow-switching cases. An analogue of the reasoning cases occurs for demonstratives, without appeal to anything as complex as Twin Earth. A person can be looking at a tomato and think that is healthy so that is healthy. In place of a healthy tomato, a perceptually indiscernible, rotten one could be substituted so quickly between the antecedent and the consequent of the thought that the reasoner would not notice. Then a rational agent could become committed to an invalid proposition through no irrationality—if he used the demonstrative twice and independently to indicate the object before him. He might even assume mistakenly that the thought is a logical truth. A correct account of the logical form of the invalid, false thought will not treat ‘that’ in its two occurrences as being syntactically the same. The token applications count, from the point of view of a syntax relevant to logical form, as formally different. If the thinker treats both occurrences as deictic, as independently applied demonstratives, then he is liable to error with regard to what could seem to be a very safe logical truth. But the thinker would not be committed to the syntactic or logical form of a logical truth. If the thinker does not treat both occurrences as deictic, then there need be no difficulty. If the thinker is to avoid any susceptibility to difficulty, the second occurrence of ‘that’ must be tied anaphorically to the first. Obviously, one could avoid switching the tomatoes, and still get the contrast between the logical truth and the non–logical truth. The difference does not lie in the objects or res. The difference lies in the uses or applications in thought that are counterparts of linguistic uses of the demonstrative ‘that’. These are under the potential cognitive control of the thinker. Although the twin concepts that I have discussed in slow-switching cases are not demonstrative or indexical, the twin concepts contribute differently to logical form or logical
424
Tyler Burge
syntax. One can engage in equivocation without realizing it, in a way analogous to the way one can engage in a demonstrative shift of reference without realizing it. To avoid susceptibility to equivocation, one must tie one’s concepts together “anaphorically” in one’s reasoning. But this is the normal way that we implicitly understand steps in a piece of deductive reasoning as fitting together anyway. We implicitly understand such steps in that way even apart from anti-individualist considerations. When I think “Every man is mortal; Socrates is a man; so Socrates is mortal,” I allow no equivocation on ‘man’, ‘mortal’, or ‘Socrates’ by implicitly relying on a sameness of conceptual and indexical use. If I understand him correctly, Kobes makes substantially this point (pp. 211–212). But I think that the links between steps in reasoning are formed anaphorically. Sometimes there is a telic line that is “forged,” but it seems to me that commonly the anaphora is correctly seen as simply preserving a content that was already unequivocally in place at the earlier step in the reasoning. Such preservation seems more “thetic” than “telic,” but I am quite happy to dispense with this terminology. I turn now to Kobes’s discussion of Loar’s doubt. There are elements in the very posing of the doubt that seem to me to be odd and in some respects significantly off the mark. The problem is supposed to be to explain how someone could assure himself of the apriority of an inference from the existence presuppositions that Socrates and hemlock exist, to the reflexive judgment I am now thinking that Socrates drank some hemlock. In the first place, the relevance of the existence assumptions seems to me quite unclear. Kobes (on Loar’s behalf) writes, “Now let us suppose that S knows that such reflexive thoughts [as “I am now thinking that Socrates drank some hemlock”] are always true, given the existence presuppositions [that Socrates and hemlock exist]” (p. 203). The authority of authoritative self-knowledge does not extend to the res in de re judgments. The relevant self-attributions in authoritative self-knowledge are to be seen as about the intentional content and the attitude-type of the attributed attitude. There are intentional elements in thought referring to Socrates and hemlock. These are trivially not identical with Socrates or hemlock. They have aboutness or representational properties and functions. Socrates and hemlock do not. The intentional or representational content does not even include the referents of the conceptual elements. It includes only the “senses” or modes of presentation, or conceptual and applicational elements of the thoughts. Thus it is epistemically possible that one think mistakenly that Socrates drank some hemlock even if Socrates and hemlock were to turn out not to exist. If we were to find out that Socrates did not exist, I would still have thought a thought properly expressed in such terms. And I would have authoritative self-knowledge of my thought (of the thought that I would have thought) even if Socrates and hemlock were to turn out not to exist. In that case, my thinking the relevant thought would not, of course, depend on my bearing causal
Reply to Kobes
425
relations to Socrates himself or to hemlock. It would depend on another causal complex connecting me to other relevant uses of the name ‘Socrates’ or the noun ‘hemlock’.12 Given that Socrates and hemlock do exist, of course, my thinking the intentional content of the thought—not merely my thinking things de re about Socrates—depends on my actually bearing those causal relations to those objects. But from an epistemic as opposed to individuative point of view—which is the point of view at issue—the existence or nonexistence of Socrates and hemlock is irrelevant to the self-knowledge. In the second place, and more centrally, it seems to me that the challenge to produce an apriori inference is misconceived. Kobes takes up this challenge to show how a thinker S “could demonstrate to himself a priori that he is thinking that Socrates drank some hemlock.” But I think that the challenge itself needs some scrutiny. Loar’s idea appears to be that, in light of externalist claims that thinking the reflexive thought depends on bearing causal relations to some external objects, one should worry that one could know one’s thoughts only empirically, because one can know these relations to the environment only empirically. So one needs to be able to reproduce some apriori inference from the existence presuppositions to the reflexive thought. That this way of understanding the problem that I set up in “Individualism and Self-Knowledge” is offtrack is suggested by the following fact. No inference to a conclusion constituted by one of the reflexive, self-verifying cogito-cases could possibly be needed to justify the relevant thought. Consider the thought I am in this very thought entertaining the thought that Socrates drank some hemlock. Because of the performative, self-verifying character of the thought, no inference to it could possibly be needed to provide support for it. It is clearly a starting point, a basic thought, whose justification lies in its own performance and content, not in the content of other thoughts from which it might be inferred. So no inference to it is needed for it to have epistemic support. This point seems to me to be completely independent of the truth or falsity of antiindividualism. So the demand that one be able to support it through some inference is misconceived. Of course, not all instances of authoritative self-knowledge are reflexive, performative, or self-verifying. In fact, the real issue is not one of justifying the truth of the whole selfattribution. It concerns the warrant for the attribution of the intentional content of the attitude that one attributes. This remark brings me to a third respect in which the problem is misconceived as posed. In my (1988), I asked why the fact that we have only empirical access to causal relations that fix the nature of our thoughts does not entail that we cannot know that we are thinking such and such unless we engage in empirical investigation that shows the conditions for thinking such and such are satisfied. I said that the answer “can be seen as a series of variations on the point that one must start somewhere” (Burge 1988, p. 654).
426
Tyler Burge
I think that Loar and Kobes overlook the starting point that I took to be basic. They suggest that the key to my “reconciliation strategy” is a schematic generalization: Kobes writes: What S knows is that all thoughts of a certain form—including “I am now thinking that Socrates drank some hemlock”—are true. But it would seem that in order to use that information to resolve his empiricist doubt, S would already have to know that he is thinking a thought of that reflexive form! And that knowledge may, for all we have said, depend on empirical knowledge of external causal or historical relations. (p. 204)
He later calls the schematic generalization my reconciliation strategy “unadorned.” There are two things wrong with this account of my view. One is that the fundamental relevant generalization is not that thoughts of a certain form are true. Only pure cogito cases—not even all performative or reflexive cases—of authoritative self-knowledge are true in virtue of their form, together with the fact that they are actually thought. For example, the thought I hereby intend to give to Oxfam is not true in virtue of its form, as I am hereby entertaining the thought that writing requires concentration is. Only the latter is a pure cogito case. I discussed both sorts of cases in my article. But the key generalization fixes on the intentional content attributed in the relevant thoughts, not their truth. Thus, the main issue concerns the contents to give to Oxfam and that writing requires concentration. The point, as I stated it, is that performative or reflexive cases are such that the intentional content that they attribute is thought and thought about at the same time. So the content of the attributed bottom-level attitude and the content attributed in the selfattributional thought are locked together. Thoughts of that reflexive form cannot mistake the intentional content of the attributed thought, though some of them—the performative cases that are not pure cogito cases—are fallible. Thus one could be mistaken in holding that one is intending in the relevant case. But one could not be mistaken because of some mismatch in content between self-attribution and an attributed intention. I think that one can know, by simply understanding one’s thought, when one’s thoughts have a reflexive form and require the relevant locking. The apparent threat in the switching cases is that the content of one’s self-attribution and the content of the attributed propositional attitude will come apart. The schematic generalization shows that this apparent possibility is illusory, at least in a large number of cases of authoritative knowledge. It is illusory in all cogito cases, whether pure or impure—all cases of reflexive performatives. Successful reflexive performatives, whether pure cogito cases or other sorts of reflexive performatives, are self-verifying. Some thoughts with the form of reflexive performatives—impure cogito cases—can be false. Yet in all such cases, the content attributed in the self-attribution cannot fail to be the content of the attitude that is attributed. One may mistake the attitude. But one cannot get the atti-
Reply to Kobes
427
tude right and mistake its content—at least not in the performative, reflexive cases. More will need to be said about authoritative instances of self-knowledge that are not performative or reflexive, as for example, self-attributions of standing states. I shall return to these cases in a short while. The second thing wrong with the account of my reconciliation strategy is more fundamental. The exposition Kobes gives of my account understates what my “reconciliation story” says. My account does not merely appeal to a schematic generalization. It emphasizes that people must understand the content well enough to think and attribute it. In particular cases of thinking the relevant performative or reflexive thoughts, the content is thought and thought about at the same time. The same content is deployed at the lower level and at the higher level of self-attribution at the same time. I wrote that my answer to the problem I raised would be a series of variations on the theme that one must start somewhere. The basic starting point that I alluded to is one’s understanding of the intentional content of one’s own thought. The starting point is not a generalization about the form of thoughts. It is the minimal understanding necessary to think the self-attribution, and to raise sceptical scenarios with respect to it, in the first place.13 Thus thinkers who self-attribute in the relevant way must be taken to understand their contents well enough to think them. In thinking them in the reflexive, performative cases that I centered on, they must think and think about their contents in the self-attribution itself in such a way that the intentional content at the different levels cannot come apart. It is the same, understood content at both levels. The problem as Loar poses it ignores the reflexive nature of the relevant thinking. By treating the thinker as if he must wonder what content he is thinking (in the conditional, “If I am now thinking that I am now thinking that Socrates drank some hemlock . . . ,” p. 204), Loar treats the thinker’s understanding of his own thought as if it were understanding from the third-person point of view. The thinker is not in a position to wonder about the content of his thought, in the relevant way, given that he makes the self-attribution. He must minimally understand the content in order to think the self-attribution in the first place. The problem as posed fails to acknowledge that the thinker must be able to understand, grasp, the particular intentional content in thinking it. In thinking reflexively, the thinker thinks the content and self-attributes it in the way that is expressed linguistically in that-clauses. Kobes is right (p. 222) to make the important point that my account of reconciliation is from the point of view of a theorist constructing an account of the thinker’s epistemic entitlement to the self-attribution. Or rather I am constructing an account of one element in the entitlement. An entitlement is an epistemic warrant that the individual has but need not have the concepts or abilities to explain or understand, even on reflection. So in my
428
Tyler Burge
view there is no need for the individual to be able to give an apriori, or any other kind of, account of why he is nonempirically warranted in his self-attributions. The individual need know nothing of anti-individualism or of any reconciliation strategy to have the relevant warrant. In thinking that I am now (in this very act of thought) thinking that Socrates drank hemlock, I understand the content of my thought well enough to satisfy the condition on understanding that is relevant to making it a reasonable question whether the thought that I am thinking constitutes knowledge. Given that the understanding is in place, and given that I actually think the relevant thought, I cannot be mistaken in my self-attribution of the content. No empirical knowledge is needed to establish an understanding of the intentional content of the thought that one is thinking. But suppose, as Loar and Kobes do, that the individual knows about anti-individualism and the reconciliation strategy. What are we to say about the challenge to produce an apriori inference to the self-attribution as conclusion? What I think we should say depends on the particular type of authoritative self-knowledge. In pure cogito cases, no justificatory inference is possible, and none is needed. The judgment I am hereby entertaining a thought that Socrates drank some hemlock, when made, is self-evidently self-verifying. It is obvious that empirical issues and issues about switching are irrelevant. For from the point of view of the person making the judgment, the thought is understood by him and understood to be self-verifying, hence obviously true. It is at least as much a piece of nonempirical knowledge as is a self-evident truth of logic. In cases of reflexive performatives that are not pure cogito cases—like I hereby judge that Socrates drank some hemlock or I hereby intend to give to Oxfam—the selfverification is not formally guaranteed merely by thinking the thought. But their truth is guaranteed, in normal circumstances, by the clearly understood performance of the act. As I indicated earlier, however, the mistake that is supposed to be threatened by the slow switching cases is not just any sort of mistake. It is supposed to be a mistake that derives specifically from some dislocation of content between that of the self-attribution and that of the attributed attitude. But again, no justificatory inference here is necessary, or, as far as I can see, possible. If one understands the content of one’s mental act of selfattribution, one understands that there is no room for a transition between the content of the self-attribution and that of the attributed attitude. For the content of the attributed attitude is fixed as that of the self-attribution. One can think the first thought and fail to make a judgment, and thus think something false. One can make the judgment articulated in the second thought and fail to have the relevant intention, and thus judge falsely. But one cannot indicate through the thought a judgment or intention that lacks the content that is attributed—the intentional content: that Socrates drank some hemlock. One can understand these points on reflection. That is what the reconciliation point emphasizes. It simply
Reply to Kobes
429
elicits something present in one’s ordinary understanding. Again, empirical considerations and issues about switching are obviously irrelevant, once one reflects on one’s own thoughts. So again no apriori inference is needed, or as far as I can see, possible. It seems to me that Kobes may himself fail to appreciate the implications of reflexivity when he writes: For all S has is the thought that he, S, thinks that p, and this higher-order thought, even if it is a belief, is not yet presented as something that S can think about. . . . From the thought I think that p, S is not in a position straightforwardly to infer that he thinks any sort of reflexive or higher-order thought, and that is what he needs. (pp. 222–223)
This seems wrong, or at least misleading. At any rate, it is wrong for the cases that I explicitly discussed, which are all reflexive, dual-level (‘hereby’, ‘in this very thought’) cases— performative analogues of the cogito. The relevant thought is I think with this very thought that p. Insofar as the performance is reflexive, as the cases I discussed explicitly were, at least the content is both thought and thought about in the same act. In the pure cogito cases (I am hereby entertaining the thought that . . .), the attitude relation and the content of the attitude attributed are both in the position of being both thought and thought about. This is true even in the nonpure reflexive case I hereby judge that. . . . In many of the other reflexive cases—for example, the intending case—at least one of the attitude relations lacks this dual-level role. Thus, the judgment about the intention is not being thought about as a judgment. But in all such cases, the intentional content that is attributed to a propositional attitude has the dual-level role. Only the intentional content is really at issue in the switching scenarios. As far as I can see, for reflexive cases no hierarchy is relevant to the justification. One needs no engagement with a hierarchy to justify to oneself that one has not made a mistake that results from a disengagement of content between the self-attribution and the attributed attitude. And no inference is appropriate. It is self-evident from the understanding present in one’s making the judgment itself that the problem cannot arise. In these cases, the very posing of the problem results from ignoring the implications of the sort of reflexive understanding necessary to thinking the relevant intentional contents. The puzzle misleads one into treating those contents as objects of identification or potential investigation, or as otherwise separable from the content of the self-attribution. But in fact they are already understood and “individualized” in the only way necessary for the relevant self-knowledge. The thoughts at both the self-attribution level and the attributed level are dependent for their content on environmental relations. But given that one is thinking the thoughts—and understands them in thinking them, as opposed to understanding them through empirical investigation or in some explicatory way—no switching and nothing about the
430
Tyler Burge
environmental relations could possibly lead one into error. This realization is not a product of an inference from premises. It is the result of reflection on the nature of one’s understanding. I have left open whether some self-knowledge that can reasonably be counted “performative” might not be reflexive. Kobes’s solution goes through an ingenious discussion of how a hierarchy is generated through the telic mechanism. Perhaps sometimes a hierarchy is generated in some nonreflexive performative cases. I would like to understand this better. If there are such cases, I think his solution is likely to be, in its main outline, correct. I do not reject Kobes’s account as wholly inapplicable to the problem. There might be cases that can be handled in Kobes’s way at any level up an infinite hierarchy. But I do not think that the account covers the most common cases, or the cases that I centered my discussion on. In cases where one thinks the intentional content reflexively as the content of the attribution and of the attributed state or act, no hierarchy arises in dealing with the switching cases. The explicit articulation of ‘hereby’ or ‘in this very thought’ is not necessary for reflexivity. Most self-attributions of occurrent attitudes are reflexive. Moreover, many cases of self-aware conscious thoughts that p—that is, self-aware occurrent propositional attitudes that do not explicitly formulate a self-attribution of the propositional attitude—are nevertheless reflexive self-attributions of the relevant attitude toward the content that p. The self-awareness often involves an unarticulated reflexive self-attribution. In judging consciously and explicitly that p, at whatever level, one commonly implicitly believes that one thereby judges that p, as a component of the judgment that p. Whether one also believes that one believes that one judges that p is a matter of the subtlety of one’s selfawareness; it is not required by the self-awareness of the bottom-level judgment. Language often suggests a hierarchical separation that is not present in the actual thinking. IV Everything that I have said so far about the problem Kobes poses centers on reflexive cases, or at least performative cases. But I have emphasized that some authoritative selfknowledge is not performative or reflexive at all. I have in mind knowledge of one’s standing states and of certain of one’s past standing states—for example, perceptual beliefs —through preservative memory. In these cases, there is nothing in merely understanding the self-attribution itself that prevents a disengagement between the content of the selfattribution and the content of the attributed state. Is an apriori inference needed in these cases? No, an entitlement to self-knowledge holds in these cases as well. So no inferential justification by the subject is needed.
Reply to Kobes
431
But suppose the subject were apprised of anti-individualism and of the reconciliation strategy. Is a nonempirical justificatory inference possible? It seems to me that an inference is not in place even in these cases. One needs to explain to oneself a relation of noninferential transition between lower-level attitude and selfattribution that is not subject to environmental vicissitudes or in need of empirical support. One starts by postulating some belief with a certain content—say, that Roberto drank some hemlock. This belief may not be the product of any mental act. It might be acquired perceptually, or through interlocution, in a nonactive way. It may or may not be warranted. Suppose that this belief has been residing in one. Something brings this belief to consciousness; or something causes one to remember the belief. One comes to employ the first-person concept and one’s concept of belief in judging: I believe (or remember) that Roberto drank some hemlock. Why, in light of the way that one’s concepts of Roberto, drinking, and hemlock depend on relations to an external environment, is one entitled nonempirically to one’s selfattribution? There can, in these cases, be dislocations between the content of one’s initial standing state and the content of one’s self-attribution. But insofar as the relation between standing state and self-attribution is not dependent on investigation of or other reliance on the environment, beyond the causal input that made the standing state with its particular content possible in the first place, no dislocation would be affected by switches. The selfattribution would simply inherit the content of the environmentally determined standing state. I have elsewhere explained rational noninferential relations that fit this description. In the case of self-attribution of a present standing belief, the relation is necessary for rational deliberation. Individuals come to be reliable in making self-attributions through such relations. In the case of authoritative self-attributions of past standing states, the relation is a combination of purely preservative memory and the bringing to consciousness of the standing state that I just described. Purely preservative memory is the more basic rational relation of the two. It is necessary for engaging in any kind of reasoning in time, not just self-conscious critical reasoning. It preserves the content between different attitudinal states over time. Both relations are or are supported by causal relations. Both can be broken in ways that would lead to error. But the breaks would be internal to the thinker’s cognitive system. They would not involve brute errors. They would not in any way depend, for their accuracy in preserving the content of the attributed stated, on relations to or input from the environment. They are rational relations internal to a cognitive point of view and practice. Only the initial standing state, at least in the case of the belief involving Roberto, would depend for its nature, content, veridicality on relations to the environment. That content would then be inherited and operated on in the rational relations on which self-attributions
432
Tyler Burge
are founded. Similar points could be made in moving from first-level, standing selfattributions to mental acts that self-attribute, at the second level, the standing first-level self-attributions. And so on. I have elsewhere characterized both such relations as essentially transitions within a point of view.14 In this respect, they are like inferences. The transitions do not require that the initial subject matter for the self-attribution be mental acts. No reference is made to performatives in explaining them. There is no element of self-verification. The selfattributions are acts, to be sure. But they conform to an antecedently established subject matter. They are not established reflexively in the self-attribution itself. They are fallible. They may be thrown offtrack by bias, top-down reasoning, internal malfunction. But in cases of authoritative self-attributions they are not subject to brute error. Their relation to the content of the standing states is not hostage to vicissitudes of the environment. Their warrants derive not from relations to the environment that could be known only empirically. Their warrants derive from the reliability of the causal connections and from the roles of the relevant relations, and the associated self-attributions, in various aspects of rational systems. The warrants are thus not perceptually based. They are nonempirical. Notes 1. See my (1996), pp. 91–116. The notion of brute error that is used in what follows was introduced in my (1988), pp. 649–663. Some of the points made regarding the relation between authoritative self-knowledge and brute error are also made in the same article. 2. I think that such performatives as I promise to give it to you are also self-verifying. I reject analyses that claim that they lack a truth-value. 3. See elaboration of this point in my discussion of examples (1) and (2) in my (1996). 4. Descartes, Meditation IV; Spinoza, Ethics II, 49. For psychological evidence bearing on the matter, see Gilbert (1991), pp. 107–119; and Wegner and Pennebaker (1993). 5. I believe that Kobes’s brief remarks about our being the author of our thoughts (p. 208) tend to overrate our authorship. Many of our ordinary beliefs are not self-conscious or the products of agency, but we can be authoritative in our reflective self-attributions of them. 6. Although all of these extended cases involve the possibility of fallible self-attributions, I do not believe that they are any less authoritative than the performative or self-verifying cases, in my sense of ‘authoritative’. There is at least the appearance of disagreement with Kobes on this point. See p. 215. I might also say that I find the talk of substantiality and insubstantiality, which derives from a very odd and misleading technical use of this term in Boghossian’s article, unfruitful in many ways. See Boghossian (1989), pp. 5–26, section III. I regard authoritative self-knowledge, even in the self-verifying cases, as substantial in all normal senses of the word. Kobes’s explanations of the term are different from Boghossian’s, and I have no criticism of what he says on this score—except insofar as use of the term, with yet more different special meanings, remains a source of possible confusion. 7. Burge (1988). For the further steps, see my (1996), (1998b), (1998c), and (2000). 8. See my (1993), pp. 457–488; (1998a), pp. 1–37; (1998b); (1999). 9. Neither of us had read the other’s account.
Reply to Kobes
433
10. Kobes’s “smoldering self-knowledge” is I think a special case of preservative memory—the case where the antecedent involves a performative. I also see the role of agency in preservative memory somewhat differently. But I believe that we are on to the same phenomenon. 11. The fundamental point that I make in this article is anticipated by Schiffer (1992), in his comment on the paper by Boghossian that Kobes discusses. Of course, reference to the contextual determination of which thought is thought is only part of the full account. 12. Such cases are discussed in my (1983), pp. 79–110. 13. In the last paragraph of his essay, Kobes seems to me to state very well an essential aspect of the starting point—that one starts with a mental act. My account emphasizes this as well, but also emphasizes the type of understanding commonly involved in performative cases, the reflexive understanding that is present in most performative-type self-attributions. The intentional content associated with the that-clause is thought and thought about at the same time. 14. See my (1996) and (1999).
References Boghossian, Paul Artin. 1989. Content and Self-Knowledge. Philosophical Topics 17: 5–26. Burge, Tyler. 1983. Russell’s Problem and Intentional Identity. In Agent, Language, and the Structure of the World, Tomberlin (ed.), pp. 79–110. Indianapolis: Hackett. ———. 1988. Individualism and Self-Knowledge. Journal of Philosophy 85(11): 649–663. ———. 1993. Content Preservation. Philosophical Review 102: 457–488. ———. 1996. Our Entitlement to Self-Knowledge. Proceedings of the Aristotelian Society 96: 91–116. ———. 1998a. Computer Proof, Apriori Knowledge, and Other Minds. Philosophical Perspectives 12: 1–37. ———. 1998b. Memory and Self-Knowledge. In Externalism and Self-Knowledge, Ludlow and Martin (eds.). Stanford: CSLI Publications. ———. 1998c. Reason and the First-Person. In Knowing Our Own Minds, Crispin Wright, Barry C. Smith, and Cynthia Macdonald (eds.). Oxford: Clarendon Press. ———. 1999. A Century of Deflation and a Moment about Self-Knowledge. Presidential Address at Pacific APA, April 1999. Proceedings of the APA 73: 25–46. Descartes, Rene. 1641. Meditations in The Philosophical Writings of Descartes. Cottingham, Stoothoff, and Murdoch (eds.). Cambridge: Cambridge University Press. Gilbert, Daniel T. 1991. How Mental Systems Believe. American Psychologist 46: 107–119. Schiffer, Stephen. 1992. Boghossian on Externalism and Inference. Philosophical Issues 2: 29–38. Spinoza, Baruch. 1985. The Collected Works of Spinoza, volume 1. Edward Curley, ed. and trans. Princeton: Princeton University Press. Wegner, Daniel M. and James W. Pennebaker (eds.). 1993. Handbook of Mental Control. Englewood Cliffs, NJ: Prentice-Hall.
23
Phenomenality and Reference: Reply to Loar
Brian Loar’s complex, interesting essay develops a view very different from mine. I will concentrate on disagreement, but later I will explore a substantial area of agreement as a way of isolating differences. I Loar’s initial statement of his motivation for an internalist view, in terms of what is in the head, seems to me to be flawed. If I were a materialist, I would maintain both that antiindividualism is true and that conceivings are in the head. I think that a conceiving’s intentional properties are essential to it. I think that holding this view in unqualified form helps engender problems for a materialist. But combining it with anti-individualism does not, without addition of further premises, defeat materialism. At most, the combination defeats type and token identity forms of materialism.1 Some materialists support only a dilute cousin of this essentialist claim. They hold that a conceiving’s intentional properties are intrinsic to it qua cononing: Nothing could be the same conceiving and either lack intentionality or have different intentional properties or content. I cannot tell exactly what Loar intends by his essentialist claim. But the view just formulated is prima facie compatible with his words, “the conceiving cannot be pulled apart from the intentional properties.” I intend the essentialist claim to mean: For any event that is a conceiving, the intentional properties of the conceiving are essential to that event. If Loar holds the unqualified essentialist position that applies not only to the event qua conceiving, but to the event that is the conceiving regardless of how it is designated, then all the better. The facts (a) that the intentional content of a mental event is essential to the event and (b) the intentional content of a mental event is necessarily and constitutively dependent on relations to an environment, simply do not by themselves logically entail that the event, or its intentional character, is not in the head, much less that it is in the environment. A heart can be individuated essentially in terms of its relations to other body parts outside it. It may be that it cannot be identified with the tissue that makes it up. But it is a material object. And it is where the tissue is.2 Conceivings and their intentional content are constitutively dependent on relations to an environment without themselves being relations, and without themselves being in the environment. I am doubly unmoved by Loar’s motivation because I believe that spatial location is not the central issue. This is one reason why I prefer the term ‘anti-individualism’ to the term ‘externalism’. The latter invites a conflation of the locus and properties of the mental states and events with the locus and character of the environmental relations on which they are constitutively dependent. I believe that the congenitally loose talk, which derives
436
Tyler Burge
from Putnam’s original paper, about what is and what is not in the head should be laid aside. Loar discusses the relation between oblique occurrences in that-clauses and mental content. This is an area in which I have been misinterpreted, although my own early overemphasis on linguistic considerations in “Individualism and the Mental” encouraged misinterpretation. Nothing in Loar’s present discussion constitutes misinterpretation of my position. But an earlier article of his does involve misinterpretation. In “Social Content and Psychological Content,” he wrote, “Behind the [anti-individualist] arguments . . . lies something like the following assumption: Sameness of de dicto or oblique ascription implies sameness of psychological content.” Loar also claimed that the following assumption “seems to be required” by my argument: “Differences in de dicto or oblique ascription imply differences in psychological content” (Loar 1988). In fact, one will look in vain for a statement of either principle in my work. I have never believed either principle. Nor has any argument I have given presupposed or relied on either principle. I think that the first principle, taken in the way Loar takes it, is obviously false. And except for an extremely idealized language whose purpose in describing psychological states is very strictly circumscribed, it is beyond help. I think that the second principle is of interest, but would need heavy qualification to approximate a truth. My arguments are example-driven, not principle-dependent. So Loar’s arguments against these principles do not touch the arguments I actually gave. I have always considered the defeasible, open-ended principles that my arguments have suggested as subject to sharpening (or to countercases) through reflection on further cases, or through further theoretical considerations.3 I believe that certain deep but complex principles do underlie the thought experiments. But they are to be found by reflecting on our intuitive judgments in a variety of such thought experiments. The conclusions of the thought experiments are not, and are not presented as, derived from any such principles. They derive from our judgments about the cases. It is a further matter to try to find the principles that underlie and generalize the judgments. Although I believe that oblique positions in that-clauses of true propositional attitude ascriptions almost always indicate something about a person’s mental content, and often characterize it accurately, I have never thought that oblique positions in that-clauses of true propositional attitude ascriptions always “capture” mental content or “define precisely the individuating conditions of psychological states,” or even co-vary exactly with the individual’s intentional content.4 I agree with Loar that in most cases of perceptual content, and of conceptual content that is “perceptually nuanced,” that-clauses of ordinary speech give only a crude indication of the nature of the content. I also agree that there is scope for individual variation in mental content associated with a single publicly used word. I do not, however, agree
Reply to Loar
437
with Loar that oblique occurrences always fail to specify mental content. And I do not accept some of his specific arguments for differences between linguistic “oblique” content and the content of mental states. The Paderewski example is a case in point. I believe that Loar’s arguments regarding the Paderewski example depend on an insufficiently refined conception of oblique linguistic content. I think it clear that the expression ‘Paderewski’ is understood differently in the relevant two that-clauses. That is, suppose Stanislaus knows Paderewski under that name as the famous pianist from newspaper clippings and knows the same Paderewski, under the same name-form, as a drinking buddy, not realizing that the drinking buddy is the pianist— thus not realizing that there is only one relevant Paderewski. When we say both “Stanislaus believes that Paderewski is a pianist” and “Stanislaus believes that Paderewski is not a pianist,” we understand a shift in context in the understanding of ‘Paderewski’ in the two oblique occurrences. We understand the name to be associated with different construals on the part of Stanislaus in the two attributions. Whether or not something “semantic” (other than negation) distinguishes the ordinary meanings of those that-clauses, it is clear that the language, contextually used, is properly interpreted as requiring a shift in the understanding of the oblique occurrence. That shift is marked or expressed in the use of the language. Interpreting the language, in context, as expressing nothing more than a pair of mutually contradictory ascriptions would be a mistake. The logical form of the language, as used in the context, must be marked as indicating a distinction in the two occurrences of ‘Paderewski’. The difference is in something other than reference; and the appropriate marking of the shift would, of course, involve an expression that would not allow of substitution of coreferentials. This is a shift in linguistic oblique content. One can argue over whether the linguistic difference “captures” Stanislaus’s mental content. I think that a proper understanding of the contextual difference, in the case of names, does adequately characterize the mental content specific to his belief, although of course he will inevitably associate more with the name than its content.5 But what I want to emphasize is that the mental content is often tracked, and indeed expressed, more closely in language than Loar’s arguments indicate. No argument that I have given for anti-individualism rests on a general view about the exact relation between mental content and the linguistic content of oblique occurrences in that-clauses. The arguments center on examples. They suggest open-ended, somewhat schematic, defeasible principles that need filling in through reflection on a variety of cases. It is enough for the relevant argument that in the cited case it be possible that differences in oblique occurrences in that-clauses signal differences in mental content. The argument that makes reference to oblique occurrences does not have a premise like “a person’s mental content is in general captured by oblique occurrences in true propositional attitude attributions”; or even a premise like “differences in oblique occurrences in
438
Tyler Burge
true propositional attitude attributions to a person always correlate with differences in the person’s intentional mental content.” I did assume—and still believe—that differences in oblique occurrences in true propositional attitude attributions prima facie signal differences in mental content. One argument I gave for anti-individualism relies on this assumption. In specific Twin Earth cases I argued that differences in the mental content of the twins is signaled or indicated by differences in oblique occurrences in normal, true propositional attitude attributions. This argument was, and was intended to be, case-based and prima facie. In fact, I spent the bulk of “Individualism and the Mental,” and some subsequent articles, arguing against putative defeaters of such prima facie arguments. The force of my arguments lies in their tending to show that there is no compulsory, or even plausible, contextual interpretation of the differences in ordinary linguistic ascription that indicates the twins’ mental content—in natural elaborations of the particular cases discussed—to be the same. I believe that these linguistic arguments, though example-driven and resolutely openended, are strong and undefeated. But I want to emphasize that the main case for antiindividualism does not, and never did, go through considerations of linguistic ascription. The main case invites one simply to consider mental states and intentional mental content directly. The statement of the examples focuses on the attitudes themselves, not on how we attribute them. The emphasis on oblique occurrences in linguistic ascriptions was primarily intended to prevent misinterpretations of the point of the examples, and to provide supplementary support. The main case is a set of thought experiments that show that a given person can, under certain circumstances, have a given thought or attitude; but if certain environmental conditions were different or lacking, a counterpart person could not, as a matter of metaphysical necessity, have that same thought or attitude. The point can be seen in terms of concept possession: Given certain background conditions, the individual on earth can have a concept aluminum or arthritis (or one of a number of concepts of aluminum or arthritis), and the relevant individual on Twin Earth cannot. Even the Twin Earth methodology is not essential to the main case. That methodology helps show specifically how the environment can matter. The primary form of argument appeals directly to particular types of thoughts and argues that given certain normal background conditions, the absence of a certain range of relations between the individual and the individual’s environment, whether physical or social, renders having those types of thoughts impossible. II I am uncomfortable with Loar’s discussion of “the externalist reasoning” in section 1. There are various types of “externalism” and various arguments for them. But the argu-
Reply to Loar
439
ment that he concentrates on is not mine. The argument he discusses is as follows: Thoughts can be intentional or “directed,” can “purport to refer,” only by presupposing actual references. Referring to external properties consists in externally determined relations between concepts and properties, at least for concepts that purport to be outwardly directed. So outwardly directed thoughts depend on externally determined relations between concepts and properties. Loar accepts the second premise, but rejects the first. Although I am not confident how to interpret Loar’s intended reading of the first premise he states, I think that there is an interpretation of it that would make the argument sound. I will discuss Loar’s attempts to defeat it. But I reemphasize that the main case for anti-individualism does not rest on general theoretical principles, but on reflections on possibilities elicited by particular cases. Relevant principles get their plausibility from reflection on cases. Such principles are almost always messier and more qualified than one initially thinks. The point in “the externalist” reasoning that Loar rejects is the connection between having particular intentional thought contents and those contents’ having specific success conditions—broadly, conditions for intentional or referential success. In his arguments he also appears to reject, perhaps in an ad hominem mode, the idea that any such specific success conditions constitutively and necessarily depend for their having those conditions on a pattern of successful reference. I think that both of these rejections are mistaken. I will start with the first. Propositional intentional thought content bears essential relations to specific truth conditions. I do not claim that all such content is constituted by truth conditions. Having a thought entails being situated in a system whose essential core components are evaluable for both rationality and for truth or other sorts of intentional success. Evaluation for truth or intentional success requires that intentional contents be referentially committed to specific referents. “Outwardly directed” intentional contents must be committed to there being specific “outer” referents. This is the point of contention in Loar’s paper. I think that disputing it leaves Loar with either an incoherent or an empty conception of outwardly directed intentionality. Before examining Loar’s central case for his conception (in his section 9), I will discuss some points in his preliminary case. I am in agreement with much of what he says about perception. Clarifying the agreement will sharpen what is in dispute. I agree with Loar that there are demonstrative, “singularly directed” concepts. A complex concept associated by some person with ‘that lemon’ on a particular occasion and applied in conjunction with a perception as of a lemon would usually be such a concept. (I see such complexes as having a nonconceptual element in applications on particular occasions—an element corresponding to the token event that is the application of the demonstrative-like element.) I agree that such concepts are essentially associated with phenomenal aspects of
440
Tyler Burge
perceptual experience. The intentional application of the concepts depends partly on the phenomenal quality of a perceptual or imagined image. The perception also has intentional, including demonstrative-like, aspects. The singular demonstrative application is not reducible to the associated intentional concepts or perceptual types.6 Given the right associated conceptual equipment, we can identify the directedness in experience. We can be aware of the singularity of the purported reference to a lemon— whether or not a lemon is actually present. The directedness is phenomenal in that limited and, I think, somewhat artificial sense.7 I agree that visual qualia, at least those functioning visually as opposed to being noise or blur in the visual field, do have representational or intentional characteristics. Normally functioning qualitative aspects of the visual field are all co-opted for representational purposes. This is not to say that all the characteristics or properties of the visual field are intentional properties. It is just that all normal elements of a visual field function intentionally—function visually!—and have intentional properties. I also agree that there is such a thing as “how” one’s perceptions and concepts phenomenally represent things. This is at least in some respects distinct from the intentional content of the actual representation, as well as whether they have a referent. I applaud Loar’s criticisms, in sections 5 and 6, of any representationism that tries to appeal only to the referents of intentional contents or that invokes intentional object theory. I believe that his discussion of the how and the what of pictures in section 7 is imaginative, insightful, and illuminating. But I find much of what Loar writes, especially in sections 4 and 5, hard to understand. I am not always sure what he thinks he has supported at given stages of his discussion. This may be my fault. But I want to enter some caveats about passages in these sections. I do not think that these sections—or any points argued prior to section 9—are even relevant to supporting the idea that singular directedness is an “internally constituted property” in a sense that supports “internalism” against my sort of “externalism,” or antiindividualism. All purported singular reference must be guided by purported reference—whether perceptual or conceptual—to a relation, property, or kind type. Purported reference to a type may occur through predication of concepts or through perceptual type-presentations. Singular purported “outer” reference to individual objects can, of course, sometimes fail to secure an actual reference. Even concepts and percepts with purported “outer” reference to properties or relations can fail to secure a reference to actual property- or relation-types (or, on a more platonic view of kinds or types, can at least fail to have actual instances of the types)—although this failure is less common. That is, particular instances of intentional states can be “nonrelationally” intentional in that they lack actual referents, types or individuals. But that point is granted by anti-individualism from the beginning.
Reply to Loar
441
What I think has not been made persuasive is that nonrelational “outer” intentionality in this weak sense can be present in an individual’s mental states without being connected in various ways with some successful “outer” reference—by other states—whose intentionality is essentially connected to actual outer objects and properties. Loar never explicitly explains what he means by ‘outer’. I think he associates it with some sort of objectivity, and perhaps even with purported reference to spatially located entities outside the individual’s mind. I will assume that he has some sort of objective reference in mind. We can identify singular and nonsingular “outer” intentionality—purported reference—in experience only if it is there to be identified. And it can be there only if the relevant successful references and supporting nonintentional relations between the individual and the entities referred to are in place. I want also to enter a caveat about the term ‘phenomenal intentionality’ as applied to the irreducibly singular elements in intentionality. If the term applies strictly, as Loar says it does, to an intentionality that one can be aware of as being present in experience, then as I have said I have no objections to it. I think though that the relevant awareness is really a range of awarenesses, varying in richness and sophistication. They require a variety of background abilities. For example, being aware of—and even more, being able to identify—the singularity and outerness of singular intentionality as such requires tracking abilities and certain relatively sophisticated concepts of objectivity, as well as certain perceptual and conceptual abilities that connect one to purported types. These abilities in turn presuppose relations to an objective subject matter. If the concepts and perceptions are spatial objects and properties, then the relations must be to some of those objects and properties. As far as I can see, nothing that Loar says in this section supports any doubts about this antiindividualist conception of phenomenal intentionality—or even comes to grips with it. If the term ‘phenomenal intentionality’ is meant to suggest a phenomenon of singular intentionality that is itself qualitative (has a specific “what it is like” quality—something like what Loar calls ‘mental paint’), I am doubtful about how much weight it can bear. (See note 7.) I think that singular intentionality in thought resides in intellectual agency— in the application of concepts. Such intentionality essentially involves mental activity— use. It cannot reside simply in phenomenality, which in the ordinary sense is passive. Awareness of such intentionality is an intellectual reflexive awareness, not merely a matter of phenomenal awareness. At a lower level, singular intentionality is also involved in the functional commitments of a perceptual system. Perhaps there is an even lower level singular intentionality in the mere feeling of sensations. No outer singular intentionality—singular intentionality directed at spatial items—can be derived from the phenomenality of sensation alone. Regardless of the outcome of these differences, I see nothing in Loar’s discussion in these passages that casts doubt on anti-individualism about “outer-directed” intentional
442
Tyler Burge
states or abilities. I see nothing in these sections that even tends to suggest that singular intentionality in perception is individualistically constituted, or resides purely in the phenomenal quality of the perception. I do not think that we have any coherent understanding of such intentionality or any coherent account of how it would be possible. Whether the singularity of phenomenal intentionality can be separated from purported outerness, or from objectivity, is a difficult and complex question that I will not discuss here. What I am firm about is that any purported “outer” reference—whether singular or otherwise—necessarily presupposes relations to “outer” entities. One further group of caveats concerns the discussion of the what/how distinction in section 5. I agree that what an intentional content refers to is different from how it refers. I agree that the directness of visual demonstrative concepts goes beyond predication. I even believe—although I doubt that Loar agrees—that the mode of presentation involved in some intentional contents involves phenomenal properties that are not intrinsically intentional. But all visual demonstrative concepts are applied by way of phenomenal characteristics that involve predication—presenting something as F in a way that has a propositional form. The singularity of the application depends on the function of the visual system and attendant tracking abilities with respect to definite particulars, and in the agency of singular demonstrative application of concepts of properties of those particulars. I see no reason to believe that singular outer purported reference is intrinsic to the quality of the phenomenal presentations. I do not know how to even understand such a notion. Although I find Loar’s discussion of singular reference as a matter of style in section 5 uncomfortably vague, most of the argumentation in the section seems to be successful against crude conceptions of intentionality that would explain each instance of intentionality as a matter of actual reference to some object, including an intentional object. Those conceptions are not mine. I think that they should have been abandoned after one reading of Frege. III Loar’s main case for his position is laid out in his sections 9–11. His position is that “one can hold constant phenomenologically accessible intentional visual qualia while varying all the properties that they represent things as having.” This formulation is not incompatible with my anti-individualism, for various reasons. One could maintain that our actual perceptual and propositional states are constitutively dependent on relations to the environment. Yet one could simultaneously allow that our visual qualia, which are in fact intentional, could be held constant while varying their spe-
Reply to Loar
443
cific intentional characters by varying the properties they represent things as having. One could also accept that some of the intentional aspects of the visual qualia could remain the same as the qualia remained the same, while maintaining that the intentional content of many of our perceptual states—what the states represent as being visible properties— would vary with the systematic variation of the objects and properties that the visual system is related to. Evidently Loar does not intend either of these interpretations. Loar apparently intends by the quoted sentence that all the intentional aspects of the visual qualitative representations can remain constant while all the properties that those representations represent things as having vary. Loar thinks that the concept that kind, where a visual qualia provides a recognitional basis for identifying the kind, is a paradigm of singular perceptual reference. Which kind the concept refers to depends on what environment the individual is in. Loar rests his case on intuitions about brains in vats. I think that his discussion is hampered by underdescription of the example he uses. I think that it is also hampered by overlooking fundamental elements in the anti-individualist position—at least in my position. He writes: I could have a mental twin whose brain is a molecule for molecule duplicate of me; and I can conceive that twin as having the same visual experiences that I have, even though its brain is isolated from all the normal causal relations to the world that give my visual experiences their actual references. The point is that when I imagine how the brain’s visual experiences represent their (merely intentional) objects, I apparently imagine those experiences as in some sense intentional, despite its difference from me in all its references.
‘Same visual experience’ here must be understood to mean ‘visual experience with exactly the same intentional content’. If the sameness of visual experience consisted only in the twin’s having the same phenomenal qualia, even with some of the same intentional properties, that would not suffice to provide opposition to anti-individualism. Moreover, a twin’s varying in all its references while having experiences that are “in some sense intentional” (or even “intentional and outer-directed”) is not in itself individualist or internalist. The view must be that the whole of the twin’s visual intentional content is the same as mine, but there is no commonality of actual reference. Loar’s assumption that the twin has a brain creates a simple incoherence from the beginning. Most of our most basic visual categories are innate. The references of these categories are set through the evolution of the brain and its visual system. Given any surface stimulation that produces relevant patterns in the visual system, the visual system of a creature with a brain will refer in vision to types of surfaces, edges, textures, spatial relations, probably colors. Although there may be no successful reference to particulars by my twin, we will share a large range of references to property types. The references are
444
Tyler Burge
innate in the sense that given the appropriate stimulation—regardless of the individual’s particular learning history—the visual system will make reference to these properties. The properties, much less objects with them, need not be the causes of the individual’s stimulation. The individual need not interact with the relevant properties in a patterned, systematic way at all. The relevant interactions with the environment occurred in the evolution of the visual system that I and my twin share. The twin inherits these interactions with instances of the represented types, even if he does not add any new ones. Thus insofar as an envatted brain has any visual intentionality, its intentionality will involve reference to many of the same properties that structures associated with our brains do. The brain and its visual system evolved to represent and respond to certain stimuli because those stimuli provided systematic access to visible aspects of the environment. And representing and responding to basic visible features of the environment had survival value. Suppose that we waive the appeal to brains and imagine a system that is molecule for molecule homologous to our brains, but came together as a cosmic accident. I have little confidence about how to imagine such a being from the inside or outside. At least in its first moments, it would seem to lack most of the cognitive and perceptual systems that I have. I am inclined to think that it would have similar qualitative, phenomenal “feels,” since I conjecture that certain qualitative aspects of the mind depend purely on the underlying chemistry. But at least until it has interacted with its world, I do not think that it has any “outer-directed” intentionality. I think that it does not even have a visual system until it has interacted with its world. One can be easily confused in phenomenal exercises. One can imagine that things would “look” just the same to the homologous accident. Such an imagining would be corrupt. The notion of “look” already depends on presumptions of perceptual and conceptual content that I believe are illegitimately imported into the envatted accident. The processes in this thing at first lack meaning and function. At most the individual would have similar phenomenal features. Suppose that we waive the appeal to a cosmic accident and imagine some natural entity that is chemically homologous to our brains. Suppose that it is not a human brain. Suppose it is in the vat. Then we need to know how it evolved, or better, what its functions are and what its relevant relations are to whatever it has conceptions of. When we fill out the story, we might conclude that it has some sort of intentionality. If it is a sort of thing that is always developed in vats, that is one story. If it evolved or was created somewhere else and got plunked into the vat, those are other stories. But if it purports to refer to mindindependent objects with empirical concepts, what its intentional content is will depend on more than whatever is supervenient on its chemical structure. If the content has a function in purporting to refer to things beyond the system, it will have to be supported by systematic interaction between the system and things beyond the system. This interaction
Reply to Loar
445
could occur in the evolution of the system or in the learning history of the individual. I do not believe that one can coherently imagine that such a being would have the same intentional content as ours unless it had some of the same references, at least to perceptual and conceptual kinds. And such references will depend on some sort of interaction with appropriate referents. So I see no intuitive support for Loar’s internalist view in the brain-in-vat scenarios. Much of Loar’s strategy is to build on cases of nonreference. We can imagine purporting to refer to a lemon that is not in fact there. We can imagine all its purported “recognitional” applications to have been hallucinatory. This may be true of the concept witch. We can imagine it to be true of some sort of nonexistent fruit all of whose “recognitions” have been caused by induced hallucination. Loar takes such cases to illustrate “objectindependent intentionality.” But these cases in themselves present no difficulty to antiindividualism. They are not cases in which intentionality is independent of all reference to objects and properties, but cases where the intentionality of a specific concept or perception, or a particular demonstrative singular application of a concept or perception, lacks a reference. Intentionality, since Frege, has been distinguished from actual successful reference. We can imagine most concepts in other possible worlds failing to apply to the properties they actually apply to—because the properties are lacking in that world. Contrary to some enthusiastic externalists, I believe that we can imagine any given singular application of perceptions and perceptual concepts as lacking singular reference: An individual object might be absent while the perception and conception has the same intentional type-content. And of course, we actually fail to secure a reference with some singular applications, in perception or otherwise, and a few concepts and perceptions of purported types. But none of this is new. Anti-individualism claims that our actual concepts, perceptions, and singular applications are dependent for their intentionality on some of them succeeding, and on that success being supported by nonintentional individual-object or individual-property relations. In my view, all of Loar’s examples depend on these facts. We can imagine failing with a purported singular reference to a lemon. But our purported reference commonly depends on our—or other relevant individuals’—having interacted with lemons on other occasions. Or we can imagine not interacting with lemons, and not interacting with anyone who has interacted with them, if we have a correct theory of lemon structure and perhaps an ability to imagine what lemons would look like. Or we can imagine a recognitional concept of a fruit failing to refer to an actual fruit-kind, but only because we have concepts and/or perceptions of fruit, color, surface texture, shape, and so on that hold the nonreferring concept in place. Some of these other concepts must succeed in applying to types. Similarly, with our imaginative capacities.
446
Tyler Burge
Loar acknowledges this reply. But he says nothing in response to it except that it appears to him “quite coherent to ascribe object-independent intentional directedness to recognitional concepts all at once, including basic spatial concepts.” I think that he has done literally nothing to show this view coherent beyond the flawed brain-in-vat scenario. Simply asserting its coherence begs the question. We have anti-individualistic individuative explications of empirical intentionality. These explications accord with, and explain, common intuition and the practice of the sciences. I think we have no idea what intentionality would look like if it were deprived of specific intentional contents (edge, rectangular, spotted, fruit, physical object, and so on) that have definite referents. It would certainly not remain the same as our actual intentional contents. As I mentioned earlier, I think that “outer,” objective, purported singular reference is necessarily dependent on being guided by the purported type-references of perceptions and concepts. And for outer attempted singular reference to be possible, it must rest on concepts that have been applied successfully to outer objects—if not applied successfully by the individual, then by his fellows or by ancestors who shared some of his cognitive systems. Demonstratives like ‘that’ are directed by the specific abilities of perception and tracking, and the particular functions that these abilities have for the individual. I do not find at all intuitive the idea that purported singular “outer” reference is intrinsic in having some phenomenal presentation, even supplemented with the tendency to react similarly to phenomenal presentations of the same type. Singular directedness in “outer” perception is a phenomenal notion in the sense that we are aware of it in reflecting on experience. But the awareness of such directedness, as well as the directedness itself, depends on having conceptions or at least perceptions that have a function ultimately grounded in successful attributions (and in the case of perceptions, also successful references) to particular objects. The successful attributions can be by other individuals that bear appropriate relations, perhaps evolutionary or communicational relations, to the given individual. I see no reason and no intutiton favoring the idea that “outer directedness is a phenomenal notion” in the stronger sense that it is constituted purely internally, for example, by qualia (understood individualistically) or by the chemical processes of the brain. (See Loar’s section 12.) Singularity when applied empirically beyond the phenomenal properties themselves must be associated with abilities to track an individual and discriminate it from other individuals of the same type. I do not see how such tracking can yield intentionality without yielding intentionality that is individuated ultimately through particular individuals that are tracked and discriminated, by means of the particular types that concepts and perceptions use to discriminate them. The most obvious sorts of individuals that could be individuated and tracked in order to yield outer singular representation in intentional content are spatially located particulars. I believe that for outer reference, reference to such particulars is necessary.
Reply to Loar
447
Perhaps mental items—images or sensations, for example—can be tracked as well, so as to provide a basis for singular reference. And perhaps successful reference to the numbers can yield a different kind of ground for singular reference. What seems to me impossible is for a mind to have singular intentionality without having intentionality that is individuated ultimately in terms of use that interacts with and successfully refers—both singularly and in the attributive way—to specific particulars or specific types. Singular directedness is not a notion that derives from the what-it-is-likeness of phenomenal characteristics of experiences. Outer singular directedness must be individuatively grounded in relations to outer particulars and types. To think otherwise is, I think, to imagine untenably that we have a concept of a transcendental object = x that is not an abstraction from more specific concepts and perceptions of objects and properties and from applications of those specific concepts and properties. Loar’s view bears some comparison to the metaphysics that Kant criticized for imagining it could obtain objective reference without being schematized to the ability to perceive particular object- and property-kinds. IV I come back to the central force behind the anti-individualist view, as I conceive it. That force does not lie in general principles about intentionality and reference. It lies in reflection on particular thoughts that we know we have. The central implausibility of Loar’s particular form of internalism seems to me to derive from our knowledge of the fact that our thoughts are not inspecific in their intentional type-content in the way that his theory requires. We do not merely think inspecific thoughts like that kind of individual is that way, where we carry phenomenal icons along to give color to the intentional content that kind and that way. On Loar’s view, the intentional type-content of a recognition judgment is inspecific in that it is open to an enormous variety of possible property referents—which bear no natural relation to one another. Any number of properties in different possible worlds—not all of which properties need be spatially located—could, given Loar’s picture, be signaled by the phenomenal icon, which we think of as normally applied to the objects in our environment. Since Loar does not say what it is that secures reference to actual “outer” individuals, I can only speculate. But if it is what causes the phenomenal icon, the point would be that the variety of things that, under varying environmental conditions, metaphysically could cause the relevant icons to occur would bear no further natural relation to one another. In fact, the variety is nearly limitless. Even in our given world, the representation that kind, backed merely by phenomenal presentation, will not have a definite referent unless the icon is related to a specific kind
448
Tyler Burge
through functionally significant interactions with the relevant kinds. But there is no way to prevent abilities associated with such interactions from being codified in the intentional content of mental states. That is, once one engages in a pattern of interaction with kinds in one’s actual environment, one’s intentional type-content will take on a content specific to those kinds. Loar’s conception of our intentional content—even if it could get off the ground—is too indefinite to apply to the intentional content that we actually have. We think thoughts whose subjects and predicates have nonindexical contents that purport to apply to very specific objects and properties in our world. The intentional content provides a typeor kind-marking of cognitive abilities specific to representing such objects and properties. My concept of water applies to water and could not apply to anything else. Loar’s view does not, I think, indicate any definite concept for us to be thinking. The transworld indexicality and inspecificity that Loar postulates for all our concepts is, I think, a radicalization of the mistake that Putnam made in describing his original thought experiment. It is the mistake of conflating kind concepts with indexicals. I believe that I have criticized that mistake decisively.8 But the mistake remains a recurrently tempting one for internalists, or individualists. I believe that we can know on conceptual grounds that most, indeed almost all, of our intentional content—our concepts and perceptions—is not like that. When one further reflects on the conditions that allow us to have the sort of intentional content that we have, one realizes that the specific references that these concepts actually have—especially references in the physical world—are necessarily dependent on complex relations between the individual and the environment. I would like to conclude with a loose end. For all my opposition to his theoretical position, I think that Loar is on to something that needs better understanding. It seems to me that phenomenality may be an essential element in intentionality. I think that it is not sufficient for conceptual or perceptual intentionality. But it may be necessary. There is a primitive analogue to conceptual and perceptual intentionality in the very feeling of a sensation. That involves a directedness between individual and sensation, or the phenomenal character of the sensation. Is there intentionality here? What role, if any, does such feeling play in making possible genuine perceptual and propositional intentionality about objects— entities actually and purportedly independent of a particular subject’s phenomenal experience? I do not know the answers to these questions, but I think that they should not be dismissed. Loar’s taking phenomenal experience seriously may be fruitful for understanding intentionality—even that which is constitutively dependent on our relations to an environment.
Reply to Loar
449
Notes 1. See my (1979), pp. 109–113; (1993). 2. Robert Stalnaker (1989); also my (1979), which very carefully confines my criticism of materialisms to token and type identity theories. I was quite aware that my arguments did not defeat materialist views that took the relation between mental state/event and the neural substrate to be one of constitution rather than identity. 3. I do not deny that I am guilty of occasional misleading formulations. For example, in my (1982), p. 107, I wrote: “Propositional attitude attributions which put the terms in oblique occurrence will thus affect the content of the propositional attitudes.” The term ‘affect’ suggests a causal or constitutive relation between the attribution and the actual content. I never intended this suggestion, as a full reading of my early papers will indicate. ‘Signal’ or ‘bears on understanding’ would have been better phrases. I think that the intense focus on the philosophy of language as a key to understanding all matters, including the nature of the mind, led many of us in those days to be sometimes inattentive to keeping track of the distinction. The passages from “Individualism and the Mental” that Loar concentrates on (for example in note 3 of his “Social Content and Psychological Content”) are either part of passages that are attempting to insist on the relevance of oblique occurrences in belief attributions to understanding the nature of the beliefs, or they are part of a single, secondary argument that claims that differences in attributions in oblique content signal differences in belief states, in a sense of “belief state” that makes the argument more interesting than a mere appeal to environmental relations in de re beliefs would. In fact, on pp. 87–88 of “Individualism and the Mental”, I explicitly say that the relation between ordinary discourse and the nature of the mental states that are referred to is a complex one and that there is a bias in favor of taking ordinary discourse literally. I clearly indicate that there is no entailment or constitutive relation between the nature of the attributions in ordinary discourse and the nature of the states attributed. 4. Loar (1988). The quoted phrases come from the last page of the article. 5. See my (1973), (1977), and (1983). These papers give a partial account of the role and content of proper names in language and thought. 6. The perceptual image, the perceptual representation, may or may not be considered part of the perceptual concept. I am inclined to regard the relation as necessary to the concept, but not a part–whole relation. The singular direction involved in demonstrative application is an instance of the element of application that I appealed to in my (1977) and (1983). I also discuss the matter in (1997) note 12, (indirectly) in the last section of this essay, and in my reply to Normore, sections IV and V. 7. Perhaps there is another sense of phenomenal awareness of the singularity that would not require conceptual identification of that feature. Perhaps insofar as a singular usage involves consciousness, one might allow a phenomenal awareness of the singularity, even though one is not conceptually and identificationally aware of it. These matters are, of course, delicate, and invite more investigation. 8. See my (1982). Hilary Putnam (1996) has accepted the criticism in his introduction to The Twin Earth Chronicles.
References Burge, Tyler. 1973. Reference and Proper Names. Journal of Philosophy 70: 425–439. ———. 1977. Belief De Re. Journal of Philosophy 74: 338–362. ———. 1979. Individualism and the Mental. Midwest Studies in Philosophy 4: 73–121. ———. 1982. Other Bodies. In Thought and Object, A. Woodfield (ed.). Oxford: Oxford University Press. ———. 1983. Russell’s Problem and Intentional Identity. In Agent, Language, and the Structure of the World, Tomberlin (ed.), pp. 79–110. Indianapolis: Hackett. ———. 1993. Mind-Body Causation and Explanatory Practice. In Mental Causation, John Heil and Alfred Mele (eds.), pp. 97–120. Oxford: Clarendon Press.
450
Tyler Burge
———. 1997. Interlocution, Perception, and Memory. Philosophical Studies 86: 21– 47. Loar, Brian. 1988. Social Content and Psychological Content. In Contents of Thought, Grimm and Merril (eds.). Tucson: University of Arizona Press. Putnam, Hilary. 1996. Introduction. In The Twin Earth Chronicles: Twenty Years of Reflection on Hilary Putnam’s “The Meaning of ‘Meaning’,” Andrew Pessin and Sanford Goldberg (eds.). New York: M. E. Sharpe. Stalnaker, Robert. 1989. On What’s in the Head. In Philosophical Perspectives 3: Philosophy of Mind and Action Theory, J. E. Tomberlin (ed.), pp. 287–316. Atascadero: Ridgeview.
24
Psychology and the Environment: Reply to Chomsky
Noam Chomsky’s contributions to the study of language have included several important contributions to philosophy. Most of these contributions—notably ones that are firmly based in his work in linguistics—have been objected to by prominent philosophers. I believe that most of Chomsky’s philosophical positions have aged better than the objections. In particular, his opposition to behaviorism and his defense of the methods of studying syntax as a part of psychology turned philosophy, psychology, and linguistics in more fruitful directions. His accounts of the relevance of syntax and phonology to unconscious psychological states, of the large role of innateness in language development, of the universality of many specific linguistic structures, and of the modular character of linguistic abilities—all these have helped provide a specific shape to our understanding of mind. Chomsky’s contribution to this volume is presented in an exploratory spirit that leaves open (see note 36) whether and where we disagree on a number of issues. I would like to make progress on clarifying the extent and nature of disagreement, in hopes that this will lead to progress on the issues themselves. Chomsky’s essay covers a lot of ground. To sharpen my focus I would like to begin by indicating a number of further points on which we agree. I We agree that eliminativism about mental/psychological kinds and structures is not a serious possibility even for science, much less common sense.1 The progress of cognitive psychology and modern linguistics—among other things—indicates that this idea has gotten more play than it deserves. We agree that the primary topic of our discussions is meaning and belief, not ordinary beliefs about meaning and belief. Intuitive beliefs about meaning and belief are relevant to understanding these topics. But they are fallible. There is room for discovery of unfamiliar kinds and structures that enter into the explanation or illumination of meaning and belief. Science must make its own way, even where it begins by testing and developing commonsense assumptions. We agree that many actual linguistic and other psychological structures are inaccessible to conscious introspection, intuitive judgment, and philosophical reflection. The details of psychology—especially the unconscious structures and the psychological and computational mechanisms—are largely beyond the reach of immediate intuitive judgments. They must be discovered through theorizing that appeals to a variety of types of evidence. There is a difference in emphasis between us here. I believe that whereas nearly all of the structures of syntax are unfamiliar to untutored preconceptions, a larger—but still
452
Tyler Burge
minority—proportion of psychological kinds that are relevant to understanding belief and meaning are familiar. This is because a relatively larger proportion of the structure and psychology of meaning and belief is accessible to conscious applicational and introspective understanding. Part of being a critically rational being is having some immediate cognitive control over one’s beliefs and meaning. Commonly such control extends to an ability to understand the meanings and beliefs of others. How much of these subject matters is inaccessible is open to investigation. I think, however, that there is no reason whatever to believe that our basic framework for describing and explaining human belief and meaning will be substantially altered or overturned, as opposed to supplemented and enriched. We agree that there is a large and psychologically specific innate component in our linguistic and other psychological abilities. Many of the basic structures of reasoning, perception, language, and social imitation are probably innate and universal to the species. We agree that belief-desire-intention psychological explanation of most human action is unsystematic, highly contextual, and at present more the province of common sense than systematic science. We may differ about the cognitive value of common sense, or about how substantial a contribution ordinary belief-desire-intention psychology will make to scientific psychology. I think that human beings have quite a lot of genuine knowledge about the mind through ordinary, nonsystematic judgments. And I believe that some recognizable though vast refinement and elaboration of the belief-desire-intention model will find a place in systematic psychology. But we are in accord about where scientific progress has been made. Most scientific progress in psychology has centered on explaining abilities involved in modular or relatively simple psychological subsystems. We are much further along in explaining the structure of various particular competences than in providing detailed, systematic, law-based explanation of performance—particularly intentional action—except in narrowly confined circumstances. We agree that there is no decisive philosophical objection to the systematic study of meaning on grounds that “meaning is holistic.” We may differ in our reasons. Chomsky concedes that the slogan applies to the sciences, but denies that it is relevant to natural language. I think that empirical confirmation, both in science and in common sense, is with some qualifications relatively holistic, but that meaning is not to be understood purely or fundamentally in terms of confirmation. I believe that the work on languageworld relations (“reference”) and the relation between such relations and meaning count strongly against the positivist account of meaning in terms of confirmation. Perhaps ordinary methodology in lexical semantics does as well. Holism has been enormously overplayed in philosophical accounts of meaning and reference—in accounts both of science and of common sense.
Reply to Chomsky
453
II In exploring areas of difference, I would like to begin by making some remarks about the terms ‘internalism’ and ‘individualism’. I will sometimes use the former term in this response, though I prefer the latter. It is hard to tell how much disagreement there is here. There is no explicit statement of what internalism is in Chomsky’s essay. Some of his remarks or arguments for “internalism” do not directly connect with my objections to the view I designate with the term, or with the term “individualism.” So some apparent disagreement may not be real. Chomsky’s I-languages are idiolects—languages of individual speakers. Antiindividualism, even social anti-individualism, does not presuppose the existence of public languages and is perfectly free to focus on individuals’ idiolects.2 I think that there is much more to the idea of public, shared linguistic meanings and norms than Chomsky does. But I do not think that anything fundamental to my views hinges on this point. It may well be that the study of syntactic and even semantic structures can, for now, reasonably concentrate on the study of I-languages—though I think that how speakers allow their idiolects to depend on one another is a matter of deep importance. I see no reason why such interdependence cannot be studied systematically and successfully. I think that there is no question that reference and even meaning in individual languages often depend on reliance on others. Sometimes it appears that Chomsky means by an ‘internalist’ theory one that studies the internal states of individuals. On this understanding of the term ‘internalist’, an externalist theory would include only the study of relations between individuals and the individuals’ environment, including perhaps other individuals of the same kind.3 I think that there are worthwhile systematic studies of relations between individuals and their environment—studies of semantic reference, of social relation, and of other matters. I shall return to this point. But anti-individualism, as I understand it, does not depend on these claims. Internalism or individualism, as I understand it, is not simply a claim that psychology studies the internal states of individuals. Even if I were to agree for the sake of argument that all psychological and linguistic theory focused on the internal states of individuals—thus accepting internalism in the sense I have found in the passages from Chomsky cited above—I would hold that internalism, as I understand it, is mistaken. Anti-individualism or externalism, in my sense, need not affect the way psychology studies the structures and mechanisms of internal psychological states. I am happy to agree that all psychological states, properly so-called, are “in” the individual’s mind. Antiindividualism is about the nature of “internal” psychological states. Anti-individualism is the view that an individual’s being in a significant range of particular intentional psychological states (beliefs, understandings, and so on) necessitates
454
Tyler Burge
that the individual bear certain causal, functional, or historical relations to an environment beyond the individual. Further, the natures and identities of those states are constitutively dependent on certain relations between individual and environment. The intentional psychological states that require the individual’s being in such relations may themselves be internal to the individual—genuine states of the individual. These states are not themselves relations to the environment. They are causally local to the individual. Many of them are internally accessible to the individual through reflection. Studying those states and their relations to one another can, for many of the purposes of systematic psychology, ignore the environmental relations. It might even turn out that systematic study of the relations between individuals and their environment is scientifically fruitless. I think that this is not the case; and I shall return to the point. But its being the case would not establish individualism (internalism), or undermine anti-individualism, as I understand those positions. It would not even show that the psychological states studied by scientific psychology were not anti-individualistically (“externally,” in my sense) individuated. Internalism, in my sense, concerns not the locus of the psychological states, or the best ways to study them, but whether being in them presupposes individual-environmental relations. It concerns whether the existence and nature of certain psychological kinds depends necessarily on the existence and nature of certain relations to specific kinds or situations in the environment. To recur to a well-worn analogy: What it is to be a heart depends essentially on what the heart does in the context of the body. It pumps blood to other parts of the body. A chemically identical object that did not pump blood would not be a heart. Imagine that it had evolved to carry out an entirely different function in an organism entirely different from any animals with hearts. Similarly, parts’ being certain valves and ventricles in the heart depend on their being parts of a heart, which in turn depends on the heart’s functional and causal relations to the wider “environment” of the body. In this sense what it is to be a heart depends on relations between the heart and things outside the heart. But one could still study the “internal” physiology of the heart—states and structures that are purely “inside” the heart. Their being the structures they are in some cases depends on their functions and on the heart’s relations to the rest of the body. But their mechanics and internal structural features need make no explicit reference to this wider “environment.” So it seems that on one of Chomsky’s understandings of ‘internalism’, internalism is the view that the study of belief and meaning concerns internal states of an individual. On my understanding of ‘internalism’, however, internalism is the view that all psychological states, including beliefs and understanding of meanings, are completely independent— in a metaphysical rather than causal sense—of any individual-environment relations for being what they are. This difference in construal of ‘internalism’ seems to me to explain
Reply to Chomsky
455
why Chomsky makes a number of points that are intended to support “internalism,” or to raise difficulties for “externalism,” but with which I agree. These points are completely compatible with anti-individualism (externalism) in my sense. For example, Chomsky sometimes writes as if Putnam’s initial view that meaning is not “in the head” defines opposition to internalism (Chomsky 1995, p. 44). Meaning is abstract and hence not anywhere. But the psychological state of understanding a meaning is naturally seen as “in” the mind or brain. Nothing in anti-individualism requires rejecting this natural view. I accept it (Burge 1982). Similarly, Putnam’s criticism of the approach to the traditional study of meaning and reference does not define anti-individualism or externalism, in my sense. With his original thought experiment, Putnam attempted to show that it was not both true (a) that understanding the meaning of a term involves being in a certain psychological state (or associating a concept with the term), and (b) that the meaning (or the concept) determines the reference of the word (Putnam 1975).4 Contrary to Putnam’s arguments, I believe that both of these principles are true for a wide range of terms and concepts. In particular, (b) is true for terms or concepts that are nonindexical—terms or concepts whose range of application is fixed (up to vagueness) in a way that is definite and stable across different occasions of use and does not become definite only on occasions of use. For example, I can say apriori that ‘aluminum’ applies, if to anything, to aluminum; and I can do so, understanding aluminum as a specific kind. My understanding is constant across different occasions of use and does not become definite only on occasions of use.5 Chomsky notes that Putnam’s critique of the conjunction of (a) and (b) does not bear on I-meaning. But the critique is not representative of anti-individualism. In fact, it is incompatible with, or at least uncongenial with, anti-individualism in my sense.6 Chomsky criticizes views that suppose that lexical items have no internal meaning, but do nothing other than “denote a semantic value external to the person” (p. 271). On such views, semantic value just is external denotation. I agree with Chomsky that such denotational views of meaning and belief, if seen as giving a complete account, are without psychological plausibility or interest. One can reasonably insist that there must be more to meaning than “external” denotation, and more to propositional attitudes than relations to directly referred-to denotations. This “something more” must be somehow represented—at least as a structured state of understanding it, or of having an attitude typed by it—in the mind or psychology of the individual. But whether the relevant “internal” psychological and semantic kinds are dependent for their meaning, content, or nature on relations to an environment beyond the individual is a question left completely open by such insistence. A more subtle example is provided by Chomsky’s discussion of Albert and Bill’s looking at indistinguishable apples A (Albert’s apple) and B (Bill’s). He writes, “Shall we say that
456
Tyler Burge
the thoughts, visual images . . . are the same for Albert and Bill but ‘directed’ to different things? Or different for Albert and Bill, the external objects A, B being ‘part of’ the thoughts?” (p. 282). He goes on to identify the second view with “externalist” usage, and claims that for the science of human nature an “internalist” picture seems appropriate. This passage is understandable given Chomsky’s construal of internalism as a claim that relevant psychological states are internal to the mind/brain. But an anti-individualist (or externalist in my sense) could accept either view. In fact, I do accept the first view with minor qualifications, the one Chomsky favors.7 Differences between Albert and Bill’s minds go no further than different token applications of their common perceptual and conceptual content. But their having beliefs about apples at all—as well as about the particular apples that they are related to—is individuatively dependent on their bearing certain relations to an environment. III None of this is to say that there is no real disagreement. It may well be that Chomsky accepts internalism and rejects anti-individualism in my senses. It may well be that he believes that all mental/psychological states (as distinct from relations between psychological states and things people refer to) are what they are in complete independence of the nature of the environment, beyond the individual’s body. But I know of no forceful, specific arguments for this position, or against mine. I want to discuss some places where Chomsky’s “internalism” seems to be straining toward the position I reject. One such place is Chomsky’s apparent scepticism about the Twin Earth thought experiments. As I have said, some of his scepticism seems to be directed against the idea that the thought experiments show that meaning or belief is not “in the head.” But sometimes he appears to oppose understanding natural kind words like ‘water’ in ways supportive of Putnam’s original cases. And he appears to want to utilize such opposition to oppose the results of the thought experiments. Sometimes it appears that he rejects the Twin Earth methodology altogether. He suggests that it depends on cases that are too strange, or too subject to variation with slight changes in context, to provide stable intuitive judgments. And he implies that the Twin Earth methodology provides no more than data about people’s commonsense beliefs about belief and meaning, with no significant implications for belief and meaning themselves. It seems to me that Putnam discovered something important about how a significant range of “kind” words actually work, in many people’s idiolects. I think that it is true that water has turned out through chemical analysis to be constituted of H2O molecules, and that anything that is not so composed is not water. Most of us who know the chemical facts and who use the word ‘water’, or counterparts in other languages or idiolects, would
Reply to Chomsky
457
speak as I have just spoken. And even those who do not know the chemical facts are open to finding that water has a “nature” that is not necessarily dictated by its superficial characteristics.8 This point can be hedged to allow for vagueness or borderline cases. But heavy water, often cited as a difficulty, is strictly a form of H2O; and we seem to count it a special form of water as well. The point is not impugned by Chomsky’s tea examples, or indeed Putnam’s examples about the dirtiness of Lake Erie or about coffee (see Putnam 1990). We all recognize that tea, coffee, and Lake Erie are each mostly water. A body of liquid can be tea while also being water with small amounts of tea in it. Whether and in what contexts we are inclined to point to a cup and call it ‘tea’ as opposed to ‘water’ does have to do with contextual conversational factors, but has little to do with our beliefs about what is water and what is not water. Such cases are not decisive, or even clearly relevant, in my view, for determining the semantical application of the term ‘water’ in individual idiolects. Theories of the reference-fixing of the term or concept through examples must make allowances for impurities. The anti-individualist account of these terms and concepts can easily accommodate these matters. What is more important is that the anti-individualistic force of the thought experiments does not depend on the assumption that water is H2O. The water-type thought experiments can be understood this way: All of us recognize that for a large number of terms, including ‘water’, it is in principle possible for something to look or seem to fall under the term, but fail to do so. We can make mistakes, by our own lights. We recognize that any one of us could be fooled into thinking that something is water—even on moderate normal experience with the stuff—and later be convinced that it is not water. One might deny this with respect to some terms. But most people would agree that there are many terms in their idiolects for which this is true. Agreeing is fundamental to recognizing the independence of the world from our beliefs. The relevant terms include not just natural-kind, constitutiondriven terms, like ‘water’ and ‘aluminum’, but most other terms for empirically experienced objects and properties. By hypothesis, we imagine something that does not fall under the term that would systematically fool one into thinking it does so. One would be fooled short of deeper investigation than one has actually undertaken. The possibility of such a thing has already been granted by most of us, with respect to a wide range of terms. This is the status of the hypothetical stuff XYZ. The point that water is not XYZ does not depend on a prior assumption that water is H2O. XYZ is introduced as something that is not water. As long as one takes it to be a possibility that something could fool one in some relatively systematic way—because what water is depends on more than its superficially experienced properties—one is committed to XYZ’s not being water. To argue that XYZ is water would be to misconstrue the thought experiment.
458
Tyler Burge
The rest of the thought experiment depends merely on the point that one could not think about, or refer to, water as water if one were in the same position with respect to XYZ as the protagonist is in with respect to water. That is, if one had no account of the nature of water that went beyond one’s experiences, and if one had experiences only with XYZ and contact only with others who had experiences only with XYZ, then one could not specify or think of anything as water. I think that this point is apriori and, on sufficient reflection, self-evident. This is the key reasoning underlying the natural-kind thought experiments. It does not depend on science fiction, Twin Earth, or anything particularly esoteric or hard to evaluate. The reasoning is, I think, decisive. The key step in the water-type thought experiment depends only on one’s acknowledging, with respect to a wide group of terms or concepts in one’s own idiolect or system of thought, that the correct range of application is not completely fixed by the way the things look or the descriptive properties one knows to ascribe to the objects that one calls by those terms. One’s knowledge and ordinary common experiences with the objects need not be sufficiently specific to fix the range of things that the terms or concepts in fact succeed in specifying. Or if one likes, one’s present abilities to describe do not suffice to fix the range of application of one’s reference with the term. Virtually anyone will recognize such a group of terms in his idiolect—a very large and varied group—given sufficient explanation of what is at issue. Actual usage will commonly be in accord with such acknowledgment. No individual’s experience and knowledge, or behavioral or physical responses, can be expected to be sufficient to fix, by analytical or metaphysical necessity, the nature of many of the empirical objects and properties that he thinks and talks about. There is slack between what one knows and experiences, or more broadly how the world impinges on one and the nature of the things that one specifies. This fact, together with facts about particular conditions under which particular sorts of reference or intentionality are possible, drives the thought experiments. One need not rely on some general thesis about the constitution of water, or the way that a certain class of natural-kind terms works.9 Similar points apply to the social anti-individualist thought experiments. The arguments do not depend on general claims about a public language or about shared meaning. They depend on recognition of our ability to think and talk nonindexically about certain specific items, despite an understanding of the terms or concepts that is insufficient to specify the items in other terms, except insofar as the specification goes through a reliance on others.10 The use of thought experiments like the various Twin Earth thought experiments that I have proposed requires judgment as well as insight. Most such thought experiments are inevitably underdescribed. They do not function as proofs from self-evident principles. One must have exercised judgment about whether or not a detail, or an omission, or an
Reply to Chomsky
459
oddity, is crucial to the main point of the thought experiment. Some of the cases Chomsky discusses are simply hard cases. Hard cases do indeed sometimes make bad law. It is especially noteworthy that there is much that we do not know about language and concept development that makes judgment in developmental cases difficult in principle. In some of the cases, however—Chomsky’s robin case, or his whale-fish case—the underdescription is crucial. Fuller description, properly targeted, will produce a much “easier” case. One needs to know more about the speaker’s usage and dispositions in order to settle the case. An individual’s simply belonging to a community in which many people know that whales are not fish does not itself ensure that the word-sound ‘fish’ in the individual’s idiolect does not apply to whales. But the case can be settled if more relevant features of the individual’s disposition and usage are included in the example. (See also my reply to Owens.) The sofa case that I produced and that Chomsky discusses does not depend on an attempt to provide a general characterization of radical disagreement about the nature of empirically identifiable objects.11 It depends on the possibility, in particular cases, of shared reference between the disputants; on a possible gap between what we know about the referent—or denotation, or nominatum—and what the nature of the referent in fact is; and on the impossibility of reference to certain kinds of objects (sofas) under certain conditions. I agree with Chomsky that we need deeper clarification of the fundamental notions. I think, however, that not all of the relevant issues are empirical. I believe that the claim that the cases I sketched are possible, as described, is very solid and stable. IV I would like now to discuss methodological issues. The thought experiments that I have proposed depend on “intuition”—reflective judgment—about cases. Chomsky expresses discomfort over the use of intuition. I do not think that his point that the cases use semantical terms, like ‘refers’ or ‘denotes’, that are partly technical carries any weight against the cases. The cases can be explained so as to connect with widely understood concepts about language–world relations. Insofar as Chomsky’s discomfort is grounded in his experience in theoretical linguistics, however, I find the discomfort understandable. Intuitive judgments about the grammar of a language or about what mentalistic structures underlie certain types of competence have often turned out to be badly mistaken. In many instances Chomsky is right in saying, “intuitive judgments are data, nothing more.” But I think that if this slogan is generalized, it greatly overstates the case. Sometimes intuitive judgments about belief, meaning, or reference constitute knowledge—certainly warranted belief. The point is embedded even in Chomsky’s own methodology for
460
Tyler Burge
studying lexical items. He holds that in having a language, an individual can know that ‘chase’ in his idiolect entails ‘follow’, or that chasing entails following. The reflective judgment on which this knowledge rests need not be buttressed by some empirical explanatory theory. It is apriori for the individual who judges it rightly. The judgment constitutes knowledge, both about words and about a relation between chasing and following, that does not rely on empirical investigation. This is not to deny that it is desirable to fit the knowledge into a broader explanatory theory. It is not to say that such judgments are immune to correction. But it is to say that intuitive judgments can provide knowledge about the topic of the judgments, not just data for knowledge about the judge. The most secure home for such judgments is mathematics and logic. It is clear that meaning and belief are normatively constrained by broadly logical and mathematical structures. I think that in complex ways they are also constitutively constrained by such structures. Some of these structures are accessible to reflective “intuitive” judgment. Often the reflection goes by way of complicated, dialectical, self-correcting processes. And it would be absurd to claim infallibility or even reliability for any and every off-the-cuff judgment. Finding the scope and limits of reflective judgments, as sources of objective knowledge, is itself a complex matter. In some cases, the dialectic leads to something like a systematic theory, and such theory may include empirical elements. But the structural aspects of belief and meaning that are accessible to reflection go well beyond what is codified in traditional logic and mathematics. The thought experiments that I have developed are ways of exploring and clarifying, through intuitive judgments, structural aspects of reference, meaning, and belief that are accessible to intuitive reflection. I believe that intuitive judgments in this domain are more likely to be secure when they fix on particular cases. The cases instantiate more general, structural principles. But exactly what principles they instantiate requires reflection, comparison with other cases, invocation of background knowledge—sometimes including empirical knowledge—and theory development. Here too intuitive judgments are more than data for empirical theory about the individual doing the judging. They provide warranted belief, and sometimes knowledge about reference, meaning, and belief. Although such judgments often make use of broad, well-established empirical background knowledge, they characteristically have elements that are apriori warranted. Some apriori warrant is prima facie, or pro tanto. Then the warranted judgments can be overthrown. Some apriori warrant is not prima facie, but depends on having achieved a fully secure understanding of the principles underlying the cases. Such understanding is hard to come by. Reflection is a complex and delicate enterprise. In both prima facie and non-prima-facie cases, intuitive judgments are vulnerable to correction. The judgments that purport to be non-prima-facie apriori warranted can be corrected because they rest on misunderstanding and are not warranted at all, though they seemed to be. Sometimes the
Reply to Chomsky
461
correction can come only from further apriori reflection. Sometimes the correction is empirical. But vulnerability to correction, even empirical correction, does not itself prevent a warrant for an intuitive judgment from being apriori. Kripke and Donnellan offered compelling cases for the possibility—in particular cases—of a name’s referring to a specific object, or a person’s referring with a name to a specific object, not perceptually present, even though the person could not fix the object with a description (Kripke 1972; Donnellan 1970, pp. 335–358). Their judgments provide knowledge. The judgments do not need support from further empirical investigation. Empirical investigation can help show where such possibilities are realized. But I believe that it cannot undermine the possibility that they illustrate, or even show that the possibility is never realized in our actual use of names. Intuitive judgments of the sort offered in these thought experiments are not infallible. But these have a certainty that is not seriously in danger of being overthrown. Empirical theory is not irrelevant to understanding them. But the examples capture how reference works in a way that is accessible to reflective judgment. Empirical theory, insofar as it does not change the subject, will in some measure have to conform to the thought experiments. The same can be said of some of the anti-individualist thought experiments. Not all warranted reflective judgments in the thought experiments that I have discussed are apriori warranted. Many involve a mixture of empirical and apriori elements. Disentangling the two is an important philosophical project. But the empirical elements are broad, well-established points that are relatively uncontroversial. It seems to me that we should be trying not only to find universal, innate components in our linguistic and cognitive structures and abilities. We should also investigate what components can be known apriori. And we should investigate what can be known or warranted about other matters through reliance on these cognitive structures. The apriori component does not coincide with the innate component. There are innate components that can be known only empirically. There are innate structures whose use in acquiring knowledge about other matters provides only empirical warrant and knowledge. For example, many of the categories and transformational mechanisms in our perceptual systems are innate; the warrant for their application is always empirical. And some of the apriori knowable and apriori warranted elements in our cognitive make-up are not innate. The apriori component need not coincide with what is necessarily present in any thinker or language speaker. Thus, it does not follow from the fact that a Martian speaker, or even some human speakers, need not have a certain linguistic or cognitive structure that we cannot know it through apriori reflection. It also does not follow that we cannot obtain apriori knowledge of other matters by relying on it. Even if certain mathematical concepts are not universal to all speakers or thinkers, those concepts can be used to provide apriori warranted knowledge of mathematical structures.
462
Tyler Burge
There have been many psychological studies in recent years to show that our intuitions about our grammar, motivation, forms of reasoning, and so on, are in certain cases very unreliable. This has led many to hold that reflection is not a source of knowledge except insofar as it is made redundant by empirical theory. I think that this is an overreaction, born of a simplistic conception of apriori intuitive reflection. In particular, one hears the recurrent claim—as if it were an exciting, revolutionary insight—that intuitions and intuitive reflection are fallible, and that it is easy to fall into groundless, entrenched prejudice while purporting to rely merely on intuitively obvious truths. These points are correct. But they are no surprise to traditional proponents of reflection as a source of apriori knowledge. No serious rationalist has ever regarded attempts at intuitive reflection as infallible. No serious rationalist has seen knowledge by reflection as easy to come by, except in a few cases. Such knowledge commonly requires skill, reflection on a variety of examples, dialectical or other inferential checking, and an openness to correction. Examples of misuse of reflection abound. Descartes himself emphasized this point repeatedly. For all that, I believe that more of our psychological and semantical structures are apriori accessible, through careful reflection, than is commonly thought. I believe that we are also sometimes apriori warranted in acquiring, sometimes unreflectively, knowledge of nonpsychological, nonsemantical matters, by making use of structures embedded in our minds and languages. Not all of what we can know through reflective judgment need be a part of systematic science. For example, I think that some of what we know about reference may never fit into a systematic scientific study of (say) the sorts of causal relations that enter into reference. What we know about reference through reflection is either very particular and casebound, or very general. The fine structure of the cases may not lead to a fruitful science. I think that, in itself, such a result would not make our knowledge even slightly less solid or less genuine. So I do not accept systematic empirical science as the only arbiter for all knowledge of language or of psychology. But we should, of course, develop systematic empirical science where we can. V I have been discussing ordinary reflective judgments that may produce knowledge without being elaborated in an empirical science. What of the status of cognitive psychology and empirical linguistics? Are these “internalist” in my sense? Elsewhere I have given two types of arguments that they are not. One is that they use psychological and semantical terms or concepts (believes, perceives, means, refers to) that are relevantly similar to those used in the thought experiments. So the terms and concepts of the relevant sciences and the psychological kinds that they pick out are subject to the same considerations. Since I
Reply to Chomsky
463
believe that these sciences must use relevantly similar concepts in order to solve the problems that they set themselves, I believe that the anti-individualism of these sciences is necessary and fundamental to their basic objectives. The second type of argument consists in reflection on particular theories in these sciences. Numerous theories in psychology and semantics that use standard scientific methodology, and that produce results that are as solid and potentially as rich as those of any purely syntactic or phonological theory, are manifestly anti-individualist or noninternalist in my sense.12 Chomsky’s discussion of individual theories is inevitably brief and general, and hence hard to evaluate. But at least some of the discussion seems to depend on the difference in his understanding of ‘internalist’ that I noted earlier. He points out that some studies of vision use, instead of real-world objects, tachistoscopic presentations that cause a subject to have a visual experience as of a rotating cube. He points out that the same investigations could have proceeded by stimulating the retina directly, or the optic nerve. Similarly, studies of hearing could go by way of direct stimulation of auditory receptors, rather than by way of loudspeakers.13 These points certainly suggest that these investigations of particular aspects of perception concern the processing of internal representations. But if one thought that they presented any difficulty at all to anti-individualism in my sense, one would be misinformed. Anti-individualism does not require that one study only cases of veridical perception. It does not stipulate that one study veridical perception only by fixing on cases of veridical perception: One can learn a lot about perception by simulating veridical cases. It does not even stipulate that one study veridical perception, although psychologists of perception normally do. One can study perceptual representation. But the representation of a rotating cube in cases where there is no actual rotating cube to be perceived, requires some ability to represent cubes and motion. Anti-individualism holds that this ability is not possible unless there are relations between the perceptual system (not necessarily the individual’s system, but the system the individual shares with other members of the species, or perhaps other species) and objects in space. The meaning or nature of the representational states depends on these relations. Science need not study these relations, but they have to be in place for science to study representations of those kinds. Similarly, in the discussion of Marr’s theory of vision, Chomsky claims that only Marr’s informal patter fails to be ‘internalist’. The serious theory is, according to Chomsky, ‘internalist’. One of his arguments for this view consists in claiming that ‘representation’ in Marr’s account is “not to be understood relationally, as ‘representation of’ ” (Chomsky 1995, pp. 52–53). ‘Representation’ in Marr’s theory is indeed not in general the relation of successful, extensional representation of. But this provides no support at all for internalism as I understand it. In the passages in Marr that Chomsky cites, ‘representation’
464
Tyler Burge
means ‘representation as of’. On this usage, a representation would be, approximately, a percept. ‘Represent’ as a verb is followed by a grammatical object in an intensional context. Representations function to represent certain properties, kinds, or tropes as real. Representations are intentional items, items with aboutness properties. The anti-individualist argument applies to them, or at least to a subject’s having them. So a lot of Chomsky’s discussion of Marr’s theory defends a view that I accept. It does not engage with my position. There are also, however, central aspects of Marr’s theory— and indeed virtually all serious theories of vision—that Chomsky fails to call attention to. The theory’s main objective is explicitly stated to be that of explaining how we visually determine the properties of actual objects in physical space that we in fact visually represent as being as they are. I shall quote Marr at some length: The purpose of these representations [in early visual processing] is to provide useful descriptions of aspects of the real world. The structure of the real world therefore plays an important role in determining both the nature of the representations that are used and the nature of the processes that derive and maintain them. An important part of the theoretical analysis is to make explicit the physical constraints and assumptions that have been used in the design of the representations and processes, and I shall be quite careful to do this. From an information-procession point of view, our primary purpose now is to define a representation of the image of reflectance changes on a surface that is suitable for detecting changes in the image’s geometrical organization that are due to changes in the reflectance of the surface itself or to changes in the surface’s orientation or distance from the viewer. . . . Hence we can see in a general way what our representation should contain. It should include some type of “tokens” that can be derived reliably and repeatedly from images and to which can be assigned values of attributes like orientation, brightness, size . . . and position. . . . It is of critical importance that the tokens one obtains correspond to real physical changes on the viewed surface; the blobs, lines, edges, groups, and so forth that we shall use must not be artifacts of the imaging process, or else inferences made from their structure backwards to the structure of the surface will be meaningless. (Marr 1982, pp. 43–44)14 . . . the true heart of visual perception is the inference from the structure of an image about the structure of the real world outside. The theory of vision is exactly the theory of how to do this, and its central concern is with the physical constraints and assumptions that make this inference possible. (Ibid., p. 68)
These points cannot be passed off as informal patter, as Chomsky does. They are the most serious, explicit statement of the objective and method of the theory. The method is to individuate representations and constrain their processing in such a way that one explains how they are reliably caused by and how they reliably represent the physical properties in the environment beyond the perceptual system that they in fact reliably represent in normal circumstances. Moreover, Marr’s theoretical constructions consistently follow his stated methodology. A central formal assumption in the construction of the account of the primal sketch—a
Reply to Chomsky
465
representation of simple edge patterns in the physical world—is the spatial coincidence assumption. This assumption holds that the coincidence of zero-crossings in the outputs of different-sized filters indicates the presence of a feature of the visual image that is due to a single physical phenomenon, such as a change in reflectance, illumination, depth, or surface orientation (Marr 1982, p. 70). The justification of this assumption makes further use of empirical assumptions about the nature of the physical world, which Marr sums up as “the constraint of spatial localization” (p. 68). Moreover, earlier, the choice of filters is motivated partly through considerations of what filter best simulates our detection of physically real edges (pp. 54–61). This is just one of many possible examples. Detailed acquaintance with the theory will not allow one to dismiss Marr’s assumptions about the physical world, and about our visual detection of features of it, as informal patter. They are explicitly constitutive of the heart of the theory. Chomsky has criticized my talk of a perceptual system’s solving a problem or having a purpose (Chomsky 1995, p. 55). He is surely right in warning against anthropomorphizing the perceptual system, or interpreting this talk in a way that would ascribe intent or deliberation to it. But some reference to a system’s provision of physical information about the environment through states with intentional (aboutness) properties is inevitable in any scientific account of perception. In this respect, psychology differs from chemistry or physics. Explanation of a system’s specific abilities in representing or registering15 or perceptually presenting features of the world is the primary traditional task of the psychological theory of perception. One can imagine psychological theories that totally prescind from this task. But I cannot see that such theories would be of any interest, at least as theories of perception. They would have changed the subject.16 Chomsky’s view that misperception by an animal or person is merely a matter of what people, observing the perceptual system, “decide to” call a “misperception” (Chomsky 1995, p. 53) does not accord with the actual explanatory practice of the psychology of perception. That practice takes the explanation of illusions and misperceptions as one of its tasks (Wandell 1995, chapter 11). Illusions are facts that constitute part of the subject matter of vision. They indicate real tendencies within the visual system, or in its relation to the world, that make it deviate from the norm of presenting veridical visual information about the physical world. Another area where explicitly relational elements enter into empirical theorizing is the formal semantics of natural language. In my view, this area is less advanced than the psychology of perception. But I see no reason why it should not be an area of fruitful systematic scientific investigation. I see no reason why reference, or a technical analogue, as a relation between linguistic representations and real aspects of the world, should not be partly systematizable in a formal account of the truth-conditions of sentences and utterances.
466
Tyler Burge
I agree with Chomsky’s criticism of referential semantical theories that purport to do away with, or ignore, I-meaning—or with the way individuals understand their terms—in favor of a purely denotational or referential semantics. But I cannot sympathize with his approving quotation of Hobbes’s claim that it is “manifest” that names are signs not of things but of our cogitations. I think that the negative part of Hobbes’s view is manifestly mistaken. It is as mistaken as applied to our ordinary commonsense use of language as it is to scientific language. The idea that we name and talk only about our cogitations seems to me beyond serious discussion. As I noted earlier, I think that we can know things about referential or denotational relations even if we cannot systematize them scientifically. Scientific systematization is dizzyingly difficult and complex. I do not, however, think that the sorts of considerations that Chomsky raises provide ground for believing that the enterprise is fruitless. The points about the differences in ways nominal expressions relate to the world (‘Joe Sixpack’ as opposed to ‘Julius Caesar’) are part of our intuitive self-understanding and have to be accounted for in a formal semantics that purports to systematize language–world relations. The Joe Sixpack example is not hard to provide at least an approximate gloss on. It may well be that for some purposes, semantic values can be taken to be shorthand constructs with no simple real-world correlates. But insofar as it does this, semantic theory should distinguish between the semantics of this sort of name and the semantics of ordinary names of people. For such a distinction is part of our ordinary linguistic understanding and use. It should explore the relations between such constructs and those things in the world that make talk about the constructs capable, often as a kind of shorthand, of contributing to true statements. I see no reason to believe that semantic theory should collapse into the idealist and unsupported philosophical ideology that naming is in general nothing more than a kind of “worldmaking.” What is true is that a good linguistic theory must say something about the understanding or cognitive value of names as well as something about their reference. A successful theory of reference cannot reasonably confine itself to the austere ontology of physics. It will allow objects like colors, tables, and symphonies. These are objects that we believe to be real even if no science studies them per se. Insofar as Joe Sixpack is real, “he” is not a person, or a single thing. Language–world relations are varied and complex, but variety and complexity are obstacles faced by all sciences. Taking account of language–world relations is part of the way semantics is actually practiced. I see no reason to think that there is anything scientifically wrong or fruitless in studying language–world relations, or with taking them to be part of the formal structures elaborated in semantical theory. Like aspects of the theory of perception, this aspect of semantics is not internalist, even in Chomsky’s broad sense of internalist. No serious
Reply to Chomsky
467
scientific objection has been entered against this practice. It may well be, of course, that systematic science will be more successful in studying formal structures into which referential items enter than in studying the ways such items connect to the world (i.e., in studying the reference relation itself). Quite apart from the fate of systematic formal theories of reference or denotation, theories that make use of semantic features of lexical items are still subject to antiindividualist strictures. Even if these semantic properties of lexical items are “determined by the ways they are constituted, with a rich innate contribution,” the constituent elements depend for their meaning on relations between the individual and an environment.17 The marking of the gender of a noun, the distinction between persons or animals and nonliving things, the indication that a term ‘chase’ entails ‘follow’ (where ‘follow’ is understood in the normal way in terms of moving in the same direction as, but behind)—all presuppose mind–world relations between the cognitive/linguistic system and the world beyond the system. This aspect of semantics is anti-individualist in my sense even if it is ‘internalist’ in Chomsky’s. I believe that the study of lexical meaning is no more advanced and no more clearly scientific than the study of truth-conditions and reference. But anything that we recognize as a study of the meanings of a wide range of ordinary words or lexical items will have substantial anti-individualist presuppositions. There is no need to defend generalized internalism in either sense in order to maintain Chomsky’s brilliant insights and methods in the study of the language faculty. That faculty involves a rich set of specific unconscious structures and rules, many of which are innate. The psychological states typed by phonological and syntactic structures are clearly internalist in Chomsky’s sense.18 But scientific enterprises that study language and mind and that are not internalist in Chomsky’s sense are already well launched. Many of these enterprises feed on Chomsky’s insights in syntactic theory. Reinterpreting these enterprises to fit an internalist paradigm, or counting them, on general or anecdotal grounds, unscientific or bound to fail, will not carry conviction—any more than philosophically motivated criticisms of the manifestly fruitful methods and results of Chomskian syntax have carried conviction. Formal semantic theory, which includes formal theories of truth-conditions, seems to have some promise as a scientific enterprise. I have no doubt that some aspects of what we know about reference will not be systematized into a science. Some knowledge is not systematic or scientific. How much of a theory of reference stabilizes as a science remains to be seen. The upshot will not be determined by philosophical views, but by the rough and tumble of linguistic scientific investigation. My primary point, however, has not been to defend scientific studies of the relation between individuals and their environment. It is to claim that even actual scientific enterprises within cognitive psychology or semantics that do not study individual–world relations, and are internalist in Chomsky’s sense, are not internalist in mine. Many of their
468
Tyler Burge
basic psychological kinds are dependent for their natures on relations to an environment. The relations are presupposed, rather than made a topic for scientific study. Some knowledge derives not from empirical science, but from reflection. Perhaps some of this knowledge will never be systematized in a science. But some of it is presupposed by and even frames scientific reasoning. Where scientific reasoning comes to grips with human nature, it will have to deal with systems that have intentional or aboutness properties. For human nature is partly constituted by our rational, cognitive capacities; and these capacities are fundamentally intentional or representational. Our rational and cognitive capacities represent a world that we can make mistakes about. These facts, together with apriori-knowable conditions on the possibility of such intentional representation, make it inevitable that many of our intentional representations are constitutively dependent on certain nonintentional relations between our cognitive systems and a world beyond them. Notes 1. In “Language and Nature” (1995), p. 31, Chomsky quotes my description of the eliminativist view in a way that might suggest to someone who does not know my work that I take the view seriously. In fact, I think that the existence of mental states and events is as epistemically solid as the existence of rocks and trees. 2. See my (1989). 3. This understanding of ‘internalism’ seems to motivate much of Chomsky’s (1995). For example, he writes in introducing ‘internalism’: “Internalist naturalistic inquiry seeks to understand the internal states of an organism” (27; see also p. 46). He gives as examples of “non-internalist studies of humans” the study of phases in the oxygen-to-carbon-dioxide cycle or the study of gene transmission, or of individuals as participants in associations and communities (28). 4. See my (1982) for my criticism of Putnam on these matters. Putnam has accepted my central criticism of the claims of his original, brilliant paper. See the introduction to Putnam (1996). My criticism about indexicality undermines his rejection of the conjunction of the two central principles that he says in “The Meaning of ‘Meaning’ ” cannot both be held. These points are laid out in “Other Bodies.” For later anti-individualist work by Putnam that is more congenial to my view, see “Computational Psychology and Interpretation Theory” in Putnam (1983). 5. Chomsky may doubt that there are any such terms when he holds that “reference” or language–world relations is confined to something a person establishes only on particular occasions of use. I see no ground for this view. It seems even less plausible as applied to thought than as applied to terms. 6. See note 4 and Putnam (1975). Putnam now rejects his own earlier account of the thought experiments. 7. I discuss this very case, down to the example of apples, at the beginning of my (1982). 8. A systematic empirical study that supports this view about natural kinds has been done by Susan Gelman. See Gelman, Coley, and Gottfried (1994). 9. Some have suggested that the mental content might leave an open parameter for the referent of the mental representation. This method seems to me to have little relevance to the way we actually think. Our thoughts commonly have specific commitments to specific referents (kinds, properties, relations, especially). See further discussion of this point in my replies to Owens, Peacocke, Loar, and Donnellan. 10. I have discussed these points in greater detail in my (1989), as well as in my (1979). 11. See my (1986b), pp. 697–720.
Reply to Chomsky
469
12. For an example of such a theory, beyond the visual perceptual cases I have already discussed in “Individualism and Psychology” (1986), pp. 3–45, see Hutchins (1995). For discussion of a range of other such theories, primarily in psychology, see McClamrock (1995); Silverberg (1998); and Wilson (1995). 13. See Chomsky (1995), p. 52. Chomsky claims that these theories apply exactly as well to a brain in a vat as to an individual perceiving actual objects. I think that this is true, but only because such a brain has a perceptual system that has representational abilities typed by relations that other brains have borne to their environments. See my reply to Loar. 14. I indicated the anti-individualist caste of Marr’s theory at some length in my “Individualism and Psychology” (1986a). My “official” argument for the anti-individualist caste of the theory was unnecessarily complicated and blurred the distinction between failure of local supervenience and anti-individualism. But the discussion of the way Marr’s theory works made the basic point clear. The science associates the content of perceptual representations with their responding successfully to features in the environment according to certain formation principles, which also have their content by virtue of normally corresponding to facts in the environment. (See “Descartes and Anti-Individualism: Reply to Normore,” this vol., note 13.) A number of discussions have specifically doubted the basics of my account of the theory. I believe that all of these, including some that Chomsky cites approvingly, either manifestly fail to understand the theory or fail to understand the philosophical issues that my account turns on. There is also a large literature which accords with my account of the theory and its successors. For a recent example, see Georges Rey, “Chomsky, Intentionality, and a CRTT” (forthcoming), section V(ii). The response to Marr’s work in the psychological literature has not missed the fundamental role of features of the actual physical world in providing constraints on the theory’s attribution of the nature of perceptual content and processing procedures. See Grimson (1981), e.g., chapters 1, 2, and 9.4.2; Bruce and Green (1985, 2000), preface and passim; and Ullman (1996), passim. Moreover, the anti-individualist character of the empirical psychology of vision is quite general and in no way is confined to Marr’s theory. For other work that shows, very explicitly, anti-individualism in the explanatory methods and actual theory of vision, see Shepard (1984), Ullman (1979), Spelke (1990), and Palmer (1999). 15. This is Chomsky’s word. See Chomsky (1995), p. 52. 16. These points are illustrated in Wandell (1995), and in Ullman (1996). For example, in chapter 2 Wandell discusses in detail measurable responses of the retinal image to measurable properties of light. In chapter 3, he tests forms of representation against the viewing angle of the actual light and proposes a theory of how the cone mosaic encodes the high spatial frequency patterns created by visual interferometers, and how the spatial arrangement of the cones provides information about the spatial distribution of light. In chapter 4, he discusses what information about spectral power distribution in actual light is encoded when rods initiate vision. When the discussion turns to visual representation (as opposed to neural registration) in chapter 5, Wandell uses the discussion of the relation between actual light and neural registration in preceding chapters as a constraint on the nature and mechanism of visual representation. Chapter 10 includes a detailed account of the construction of representations of depth from representations of motion that—like the earlier chapters that I have glossed—is not purely internalist even in Chomsky’s sense. 17. It is understood throughout that such relations can be indirect. The relations relevant to the individual’s innate structures might go through the individual’s progenitors and their relations to the environment—relations that played a role in giving the innate structures their intentional content and function. 18. I believe that some elements of phonology and syntax are not internalist in my sense, but I shall not discuss these points here.
References Bruce, Vicki and Patrick Green. 1985, 2000. Visual Perception: Physiology, Psychology, and Ecology. Hillsdale, NJ: Lawrence Erlbaum. Burge, Tyler. 1979. Individualism and the Mental. Midwest Studies in Philosophy 4: 73–121. ———. 1982. Other Bodies. In Thought and Object, A. Woodfield (ed.). Oxford: Oxford University Press. ———. 1986a. Individualism and Psychology. Philosophical Review 95: 3–45.
470
Tyler Burge
———. 1986b. Intellectual Norms and Foundations of Mind. Journal of Philosophy 83(12): 697–720. ———. 1989. Wherein Is Language Social? In Reflections on Chomsky, A. George (ed.). Oxford: Blackwell. Chomsky, Noam. 1995. Language and Nature. Mind 104(413): 1–61. Donnellan, Keith. 1970. Proper Names and Identifying Descriptions. Synthese 21: 335–358. Gelman, S. A., J. D. Coley, and G. M. Gottfired. 1994. Essentialist Beliefs in Children: The Acquisition of Concepts and Theories. In Mapping the Mind: Domain Specificity in Cognition and Culture, Hirschfeld and Gelman (eds.). Cambridge: Cambridge University Press. Grimson, W. E. L. 1981. From Image to Surfaces. Cambridge, MA: MIT Press. Hutchins, Edwin. 1995. Cognition in the Wild. Cambridge, MA: MIT Press. Kripke, Saul. 1972. Naming and Necessity. In Semantics of Natural Language, Davidson and Harman (eds.), pp. 253–355. Dordrecht: Reidel. Marr, David. 1982. Vision. San Francisco: W. H. Freeman. McClamrock, Ron. 1995. Existential Cognition: Computational Minds in the World. Chicago: University of Chicago Press. Palmer, Stephen. 1999. Vision Science. Cambridge, MA: MIT Press. Putnam, Hilary. 1975. The Meaning of “Meaning”. In Mind, Language, and Reality: Philosophical Papers, volume II. Cambridge: Cambridge University Press. ———. 1983. Computational Psychology and Interpretation Theory. In Collected Papers, volume III. Cambridge: Cambridge University Press. ———. 1990. Is Water Necessarily H2O? In Realism with a Human Face, James Conant (ed.). Cambridge, MA: Harvard University Press. ———. 1996. Introduction. In The Twin Earth Chronicles: Twenty Years of Reflection on Hilary Putnam’s “The Meaning of ‘Meaning’,” Andrew Pessin and Sanford Goldberg (eds.). New York: M. E. Sharpe. Rey, Georges. Forthcoming. Chomsky, Intentionality and a CRTT. In Chomsky and His Critics, Antony and Hornstein (eds.). Oxford: Blackwell. Shepard, Roger N. 1984. Ecological Constraints on Internal Representation: Resonant Kinematics of Perceiving, Imagining, Thinking, and Dreaming. Psychological Review 91: 417–447. Silverberg, Arnold. 1998. Semantic Externalism: A Response to Chomsky. Protosociology 11: 216–244. Spelke, Elizabeth. 1990. Principles of Object Perception. Cognitive Science 14: 29–56. Ullman, Shimon. 1979. The Interpretation of Visual Motion. Cambridge, MA: MIT Press. ———. 1996. High-Level Vision. Cambridge, MA: MIT Press. Wandell, Brian A. 1995. Foundations of Vision. Sunderland: Sinauer. Wilson, Robert. 1995. Cartesian Psychology and Physical Minds: Individualism and the Sciences of the Mind. Cambridge: Cambridge University Press.
Tyler Burge Bibliography
1972. Truth and Mass Terms. Journal of Philosophy 69(10): 263–282. 1973. Reference and Proper Names. Journal of Philosophy 70(14): 425–439. Abstracted in (1971). Journal of Philosophy 6(19): 627–627. Reprinted in (1975). The Logic of Grammar. D. Davidson and G. Harman (eds.). Encino, CA: Dickenson Publishing. Translated into German and reprinted in (1985). Eigennamen. Ursula Wolf, (ed.). Frankfurt am Mair: Suhrkamp. Reprinted in (1998). Readings in the Philosophy of Language. Ludlow (ed.). Cambridge, MA: MIT Press. Reprinted in (1999). Semantics: A Reader. Davis and Gillon (eds). Oxford: Oxford University Press. 1973b. Review of Anthony Quinton, The Nature of Things. International Philosophical Quarterly 14: 363–367. 1974. Truth and Singular Terms. Noûs 8(4): 309–325. Reprinted in (1980). Reference, Truth and Reality. M. Platts (ed). London: Routledge and Kegan Paul. Reprinted in (1991). Philosophical Applications of Free Logic. Lambert (ed.). Oxford: Oxford University Press. Translated into German in (1994). Wahrheit und singulare Terme. In Der Wahrheitsbegriff, Puntel (ed.). Darmstadt: Wissenschaftliche Buchgesellschaft. 1974. Demonstrative Constructions, Reference, and Truth. Journal of Philosophy 71(7): 205–223. 1975. Mass Terms, Count Nouns and Change. Synthese 23: 186–205. Reprinted in (1979). Mass Terms: Some Philosophical Problems. F. J. Pelletier (ed.). Dordrecht: Kluwer. 1975. On Knowledge and Convention. Philosophical Review 84(2): 249–255. 1977. Kaplan, Quine, and Suspended Belief. Philosophical Studies 31: 197–203. Reprinted in (2000). The Philosophy of Quine. Dagfinn Follesdal (ed.). Hamden, CT: Garland Publishing. 1977. Review of Meaning, Reference, and Necessity, Simon Blackburn (ed.). Journal of Philosophy 74(4): 241–245. 1977. A Theory of Aggregates. Noûs 11(2): 97–117. Abstracted in (1973) Journal of Philosophy 70(17): 575–575. 1977. Belief De Re. Journal of Philosophy 74(6):338–362. 1978. Self-Reference and Translation. In Translation and Meaning. Guenther-Reutte and Geunther (eds.). London: Duckworth. 1978. Belief and Synonymy. Journal of Philosophy 75(3): 119–138. Reprinted in (1993). Language and Cognition. Higginbotham (ed.). Oxford: Blackwell. 1978. Buridan and Epistemic Paradox. Philosophical Studies 34(1): 21–35. 1978. Concept of Mind in Primates. Behavioral and Brain Sciences 1(4): 560–562. 1979. Individualism and the Mental. Midwest Studies in Philosophy 4: 73–121. Reprinted in (1991). The Nature of Mind. David Rosenthal (ed.). Oxford: Oxford University Press. Reprinted in (1993). Basic Topics in the Philosophy of Language. Harnish, R. (ed.). Hemel Hempstead: Harvester Wheatsheaf. Reprinted in (1993). Language and Cognition. Higginbotham (ed.). Oxford: Blackwell. Reprinted in (1994). Basic Topics in Philosophy of Language. Harnish (ed.). Englewood Cliffs, NJ: Paramount Publishing. Reprinted partially in (1996). The Twin Earth Chronicles. Pessin and Goldberg (eds.). New York: M. E. Sharpe. Translated into Spanish in (1997). El Individualismo y Lo Mental. Valdez (ed.). Sabb (trans). In Propositional Attitudes. Mexico City: Instituto de Investigaciones Filosoficas. Reprinted in (1998). Externalism and Self-Knowledge. Ludlow and Martin (eds.). Stanford: CSLI Publications. Reprinted in (2003). Contemporary Analytic Philosophy, second ed. James Baille (ed.). Englewood, N.J.: Prentice-Mall. Reprinted in (2003). Philosophy of Mind: Classical and Contemporary Readings. Chalmers (ed.). Oxford: Oxford University Press. Translated into French in (2003). Cahiers Philosophiques de Strasbourg. 1979. Frege and the Hierarchy. Synthese 40(2): 265–281. 1979. Semantical Paradox. Journal of Philosophy 76(4): 169–198.
472
Tyler Burge Bibliography
Reprinted in (1985). Recent Essays on Truth and the Liar Paradox. Robert L. Martin (ed.). Oxford: Oxford University Press. 1979. Sinning against Frege. Philosophical Review 88(3): 398–432. Reprinted in (1993). The International Research Library of Philosophy. Skorupski (ed.). Aldershot: Dartmouth. 1979. Reasoning about Reasoning. Philosophia 8(4): 651–656. 1979. Critical Notice: Jaakko Hintikka, The Intentions of Intentionality and Other New Models for Modalities. Synthese 42(2): 315–334. 1980. The Content of Propositional Attitudes. Noûs 14(1): 53–58. 1981. Review of Syntax and Semantics, Choon-Kyu Oh and David A. Dinneen (eds). Volume 11, Presupposition. Journal of Symbolic Logic 46(2): 412–415. 1982. Other Bodies. In Thought and Object, Andrew Woodfield (ed.). Oxford: Oxford University Press. Reprinted in (1996). The Twin Earth Chronicles. Pessin (ed.). New York: M. E. Sharpe. Reprinted in (2000). Problems in Mind: Readings in Contemporary Philosophy of Mind. Crumley (ed.). Mountain View, CA: Mayfield Publishing. 1982. The Liar Paradox: Tangles and Chains. Philosophical Studies 41(3): 353–366. 1982. Two Thought Experiments Reviewed. Notre Dame Journal of Formal Logic 23(3): 284–293. 1982. Review of Naming, Necessity, and Natural Kinds, Stephen P. Schwartz (ed.). Journal of Symbolic Logic 47(4): 911–915. 1983. Russell’s Problem and Intentional Identity. In Agent, Language, and the Structure of the World. James E. Tomberlin (ed.). Indianapolis: Hackett. 1983. Review of Essays on Actions and Events, Donald Davidson. Ethics 93(3): 608–611. 1983. Review of Grammar in Philosophy, Bede Rundle. Philosophical Review 92(4): 639–642. 1984. The Concept of Truth in Frege’s Program. Philosophia Naturalis 21: 507–512. 1984. Frege on Extensions of Concepts, from 1884 to 1903. Philosophical Review 93(1): 3–34. 1984. Epistemic Paradox. Journal of Philosophy 81(1): 5–29. 1984. Review of The Interpretation of Frege’s Philosophy, Michael Dummett. Philosophical Review 93(3): 454–458. 1985. Postscript to “Semantical Paradox,” 1982. In The Liar Paradox. Robert L. Martin (ed.). Oxford: Oxford University Press. 1986. Frege on Truth. In Frege Synthesized. L. Haaparanta and J. Hintikka (eds.). Dordrecht: D. Reidel. Reprinted in (1993). The Philosophy of Frege. Sluga (ed.). New York: Garland. 1986. Cartesian Error and the Objectivity of Perception. In Subject, Thought, and Context. John McDowell and Philip Pettit (eds). Oxford: Oxford University Press. Reprinted in (1987). Contents of Thought. Grimm and Merrill (eds.). Tucson, AZ: University of Arizona Press. 1986. On Davidson’s “Saying That.” In Truth and Interpretation. Ernest LePore (ed.). Oxford: Blackwell. 1986. Individualism and Psychology. Philosophical Review 95(1): 3–45. Reprinted in (1989). Representation. Stuart Silvers (ed.). Dordrecht: D. Reidel. Reprinted in (1991). The Philosophy of Science. Boyd, Gasper, Trout (eds.). Cambridge, MA: MIT Press. Translated into Spanish in (1993). Philosophia y Ciencia Cognitiva. Rabossi (ed.). Buenos Aires-Barcelona: Editorial Paides. Reprinted in (1995). Philosophy of Psychology: Debates on Psychological Explanation. MacDonald and MacDonald (eds.). Oxford: Blackwell. Reprinted in (1994). Readings in Philosophy and Cognitive Science. Goldman (ed.). Cambridge, MA: MIT Press. Reprinted in (1995). Language and Mind: Contemporary Readings in Philosophy and Cognitive Science. Losonsky and Geirsson (eds.). Oxford: Blackwell. 1986. Intellectual Norms and Foundations of Mind. Journal of Philosophy 83(12): 697–720.
Tyler Burge Bibliography
473
1987. Marr’s Theory of Vision. In Modularity in Knowledge Representation and Natural-Language Understanding. Jay L. Garfield (ed.). Cambridge, MA: MIT Press. 1987. Reply: Authoritative Self-Knowledge and Perceptual Individualism. In Contents of Thought. Grimm and Merrill (eds.). Tucson, AZ: University of Arizona Press. 1988. Individualism and Self-Knowledge. Journal of Philosophy 85(11): 649–663. Translated into German in (1993). Analytische Theorie des Selbstbewusstseins. Frank (trans.). Frankfurt am Main: Suhrkamp. Reprinted in (1994). Self-Knowledge. Cassam (ed.). Oxford: Oxford University Press. Reprinted in (1996). The Twin Earth Chronicles. Pessin and Goldberg (eds.). New York: M. E. Sharpe. Reprinted in (1998). Externalism and Self-Knowledge. Ludlow and Martin (eds.). Stanford: CSLI Publications. Reprinted in (2000). Readings in Contemporary Epistemology. Dretske and Bernecker (eds.). Oxford: Oxford University Press. Translated into Czech in (2002). Filosoficky Casopis. 1989. Wherein Is Language Social? In Reflections on Chomsky. Alexander George (ed.). Oxford: Blackwell. Reprinted in (1990). The Role of Content in Logic, Language, and Mind. Anderson and Owens (eds.). Stanford: CSLI, Stanford University Press. Reprinted in (1999). Mente e Linguaggio. Alfredo Paternoster (ed.). Milan: Guerini e Associati. Reprinted in (1998). Modern Philosophy of Language. Baghramian (ed.). London: J. M. Dent. Translated into Estonian in (2001). Akadeemia 13(10): 2146–2169. 1989. Individuation and Causation in Psychology. Pacific Philosophical Quarterly 70: 303–322. 1990. Frege on Sense and Linguistic Meaning. In The Analytic Tradition. David Bell and Neil Cooper (eds.). Oxford: Blackwell. 1991. Frege. In Handbook of Ontology and Metaphysics. Smith and Burkhardt (eds.). Munich: Philosophia Verlag. Philosophy of Language and Mind: 1950–1990. Philosophical Review 101: 3–51. Translated into Polish in (1995). Filozofia jezyka i umyslu (1950–1990). Filozofowac dzis. Zbadan nad filozofia najnowszq. Andrej Bronk (ed.). Reprinted in (1995). Language and Mind: Contemporary Readings in Philosophy and Cognitive Science. Losonsky and Geirsson (eds.). Oxford: Blackwell. Spanish translation of approximately half in (1999). Cuadernogri. Alberto Lopez (ed.). 1992. Frege on Knowing the Third Realm. Mind 101: 633–650. Reprinted in (1996). Frege: Importance and Legacy. Schirn (ed.). Berlin: de Gruyter. Reprinted in (1996). Early Analytic Philosophy: Frege, Russell, Wittgenstein. Tait (ed.). Chicago: Open Court. 1992. Mind-Body Causation and Explanatory Practice. In Mental Causation. Heil and Mele (eds.). Oxford: Oxford University Press. Review of Studies in the Way of Words, Paul Grice. Philosophical Review 101: 619–621. 1991. Vision and Intentional Content. In John Searle and His Critics. Lepore and Van Gulick (eds.) Oxford: Blackwell. 1993. The Unreduced Mind. Review of Peacocke, A Study of Consciousness. Times Literary Supplement. 4707: 14–15. 1993. Concepts, Definitions, and Meaning. Metaphilosophy 24: 309–325. 1993. Content Preservation. Philosophical Review 103: 457– 488. Reprinted in (1995). Content. Villanueva (ed.). Atascadero, CA: Ridgeview Press. Reprinted in (1999). Apriori Knowledge. Casullo (ed.). Hampshire: Ashgate Publishers. 1995. Intentional Properties and Causation: A Reply to Fodor. In Philosophy of Psychology: Debates on Psychological Explanation. MacDonald and MacDonald (eds.). Oxford: Blackwell. 1996. Our Entitlement to Self-Knowledge. Proceedings of the Aristotelian Society 96: 91–116.
474
Tyler Burge Bibliography
Reprinted in (1998). Externalism and Self-Knowledge. Ludlow and Martin (eds.). Stanford: CSLI Publications. German translation in Essays in honor of Davidson, Kohler and Roska-Hard (eds.). Forthcoming. 1997. Two Kinds of Consciousness. In The Nature of Consciousness. Block, Flanagan, Güzeldere (eds.). Cambridge, MA: MIT Press. First published in translation as Zwei Arten von Bewusstsein. In Bewusstsein. Metzinger (ed.). Raderborn: Schoningh. 1998. Reason and the First Person. In Knowing Our Own Minds: Essays on Self-Knowledge. Smith, Wright, and MacDonald (eds.). Oxford: Clarendon Press. Reprinted in (1999). Philosophers Annual. Grimm and Mar (eds.). Translated into German in (2000). Philosophisches Jahrbuch 107. Esser (trans.). 1997. Interlocution, Perception, and Memory. Philosophical Studies 86: 21–47. 1998. Computer Proof, Apriori Knowledge, and Other Minds. Philosophical Perspectives 12: 1–37. 1998. Frege on Knowing the Foundation. Mind 107: 305–347. 1998. Memory and Self-Knowledge. In Externalism and Self-Knowledge. Ludlow and Martin (eds.) Stanford: CSLI Publications. 1998. Question to Dretske. The Dualist 5: 86–87. 1999. Some Personal Remarks about Greg Kavka. In Rational Commitment and Social Justice, a commemorative volume of essays for Gregory Kavka. Coleman and Morris (eds.). Cambridge: Cambridge University Press. 1999. Comprehension and Interpretation. In The Philosophy of Donald Davidson, Lewis Hahn (ed.). Chicago: Open Court. 2000. Frege on Apriority. In New Essays on the A Priori. Peacocke and Boghossian (eds.) Oxford: Oxford University Press. Reprinted in (2001). Building on Frege: New Essays about Sense, Content, and Concept. Newen, Nortmann, Stuhlmann-Laeisz (eds.). Stanford: CSLI Publications. Reprinted in (2001). The Philosopher’s Annual, vol. 22. Grim et al. (eds.). 2000. A Century of Deflation and a Moment about Self-Knowledge. Presidential Address, Pacific APA, Proceedings of the APA. 73: 25–46. 2003. Perception. International Journal of Psychoanalysis 84: 157–167. Forthcoming. Memoir of Noam Chomsky. In M. Seabrook, authorized biography of Chomsky, America’s Conscience. London: Pluto Press. Forthcoming. Perceptual Entitlement. Philosophy and Phenomenological Research.
Secondary Sources on Tyler Burge
The following bibliography makes no claim to being complete, although considerable effort went into gathering all we could find. That said, we made much less of an effort to cover those areas of Burge’s work—on the philosophy of language, on Frege, and on the paradoxes—that do not figure as promimently as the philosophy of mind in the present volume. For the reader’s convenience (and because we concluded that philosophers such as Descartes and Frege were not, after all, responding to Burge), we have included separate bibliographies at the end of each selection in the volume. Adams, Fred. 1993. Fodor’s Modal Argument. Philosophical Psychology 6: 41–56. Adams, Fred, D. Drebushenko, Gary Fuller, and Robert Stecker. 1990. Narrow Content: Fodor’s Folly. Mind and Language 5: 213–229. Adams, Fred and Gary Fuller. 1992. Names, Contents, and Causes. Mind and Language 7: 205–221. Adams, Fred and Robert Stecker. 1994. Vacuous Singular Terms. Mind and Language 9(4): 387–401. Akins, Kathleen. Of Sensory Systems and the “Aboutness” of Mental States. Journal of Philosophy 93: 337–372. Allen, C. 1995. It Isn’t What You Think: A New Idea about Intentional Causation. Noûs 29: 115–126. Almog, J. 1981. Dthis and Dthat: Indexicality Goes Beyond That. Philosophical Studies 34: 347–381. Anderson, C. Anthony (ed.). 1990. Propositional Attitudes: The Role of Content in Logic, Language, and Mind. Standford: CSLI Publications. Andler, D. 1995. Can We Knock off the Shackles of Syntax? In Philosophical Issues 6: Content, Enrique Villanueva (ed.). Atascadero: Ridgeview. Antony, L. 1989. Semantic Anorexia: On the Notion of Content in Cognitive Science. In Meaning and Method: Essays in Honor of Hilary Putnam G. Boolos (ed.). Cambridge: Cambridge University Press. ———. 1996. Equal Rights for Swamp Persons. Mind and Language (11): 70–75. Antony, Michael V. 1993. Social Relations and the Individuation of Thought. Mind 102(406): 247–261. Arjo, D. 1996. Sticking up for Oedipus: Fodor on Intentional Generalizations and Broad Content. Mind and Language. 11: 231–245. Aydede, Murat. 1997. Has Fodor Really Changed His Mind on Narrow Content? Mind and Language 12: 422–458. Bach, Kent. 1988. Burge’s New Thought Experiment: Back to the Drawing Room. Journal of Philosophy 85: 88–97. Baker, Lynne Rudder. 1985a. A Farewell to Functionalism. Philosophical Studies 48: 1–14. ———. 1985b. Just What Do We Have in Mind? Midwest Studies in Philosophy 10: 25–48. ———. 1987a. Content by Courtesy. Journal of Philosophy 84: 197–213. ———. 1987b. Saving Belief. Princeton: Princeton University Press. ———. 1990. Seeming to See Red. Philosophical Studies 58: 121–128. ———. 1994. Content and Context. Philosophical Perspectives 8: 17–32. Baker, Lynne-Rudder and Jan David Wald. 1979. Indexical Reference and “De Re” Belief. Philosophical Studies 36: 317–327. Baldwin, Thomas. 1988. Phenomenology, Solipsism, and Egocentric Thought. Aristotelian Society (Supplementary Volume) 62: 27–43. Barrett, Jonathan. 1997. Individualism and the Cross Contexts Test. Pacific Philosophical Quarterly 78: 242– 260. Beckermann, Ansgar. 1996. Is There a Problem about Intentionality? Erkenntnis 45(1): 1–23. Berg, Johnathan. 1998. First Person Authority, Externalism, and Wh-Knowledge. Dialectica 52(1): 41–44. Bernecker, Sven. 1996a. Externalism and the Attitudinal Component of Self-Knowledge. Noûs 30(2): 262–275. ———. 1996b. Davidson on First-Person Authority and Externalism. Inquiry 39: 121–139.
476
Secondary Sources on Tyler Burge
———. 1997a. Die Grenzen des Selbstwissens. Zeitschrift fur philosophische Forschung 51(2): 216–231. ———. 1997b. On Knowing One’s Own Mind. In Analyomen 2, Volume III: Philosophy of Mind, Practical Philosophy, Miscellanea, Georg Meggle (ed.). Berlin: de Gruyter. ———. 1998. Self-Knowledge and Closure. In Externalism and Self-Knowledge, P. Ludlow an N. Martin (eds.). Stanford: CSLI Publications. ———. 2000. Knowing the World by Knowing One’s Mind. Synthese. Bernier, Paul. 1993. Narrow Content, Context of Thought, and Asymmetric Dependency. Mind and Language 8: 327–342. Bezuidenhout, Anne. 1998. Is Verbal Communication a Purely Preservative Process? Philosophical Review 107(2): 261–288. Bilgrami, Akeel. 1987. An Externalist Account of Psychological Content. Philosophical Topics 15: 191–226. ———. 1989. Realism without Internalism: A Critique of Searle on Intentionality. Journal of Philosophy 86: 57–72. ———. 1992a. Belief and Meaning: The Unity and Locality of Mental Content. Oxford: Blackwell. ———. 1992b. Can Externalism Be Reconciled with Self-Knowledge? Philosophical Topics 20: 233–268. ———. 1998. Replies. Philosophy and Phenomenological Research 58: 635–647. Biro, John. 1992. In Defence of Social Content. Philosophical Studies 67: 277–293. ———. 1995. Testimony and “A Priori” Knowledge. In Philosophical Issues 6: Content, Enrique Villanueva (ed.). Atascadero: Ridgeview. Blackburn, Simon. 1993. Essays in Quasi-Realism. New York: Oxford University Press. Block, Ned. 1981. Psychologism and Behaviorism. Philosophical Review 90 (1): 5–43. ———. 1990. Inverted Earth. Philosophical Perspectives 4: 51–79. Reprinted in W. Lycan (ed.). 1999. Mind and Cognition (second edition). Blackwell: Oxford. Also reprinted in Block, Flanagan and Güzeldere (1997). ———. 1991. What Narrow Content Is Not. In Meaning in Mind: Fodor and His Critics, B. Loewer and G. Rey (eds.). Oxford: Blackwell. ———. 1995c. Mental Paint and Mental Latex. In Philosophical Issues 6: Content, Enrique Villanueva (ed.). Atascadero: Ridgeview. ———. 1995d. Ruritania Revisited. In Philosophical Issues 6: Content, Enrique Villanueva (ed.). Atascadero: Ridgeview. ———. 1998. Is Experiencing Just Representing? Philosophy and Phenomenological Research 58: 663–670. Boghossian, Paul Artin. 1989. Content and Self-Knowledge. Philosophical Topics 17: 5–26. ———. 1992. Externalism and Inference. Philosophical Issues 2: Rationality in Epistemology, Enrique Villaneuva (ed.). Atascadero: Ridgeview. ———. 1994. The Transparency of Mental Content. Philosophical Perspectives 8: 33–50. ———. 1997. What the Externalist Can Know A Priori. Proceedings of the Aristotelian Society 97: 161–175. Reprinted in Concepts (Philosophical Issues 9). 1998. Enrique Villanueva (ed.). Ridgeview: Atascadero. Bonjour, Laurence. 1998. In Defense of Pure Reason. Cambridge: Cambridge University Press. Bontly, Thomas. 1998. Individualism and the Nature of Syntactic States. British Journal for the Philosophy of Science 49(4): 557–574. Braisby, Nick, Bradley Franks, and James Hampton. 1996. Essentialism, Word Use, and Concepts. Cognition 59(3): 247–274. Braun, D. 1991. Content, Causation, and Cognitive Science. Australasian Journal of Philosophy 69: 375– 389. Brewer, B. 1995. Compulsion by Reason. Aristotelian Society (suppl. vol) 69: 237–253.
Secondary Sources on Tyler Burge
477
Bromberger, Sylvain. 1997. Natural Kinds and Questions. In Essays on Jaakko Hintikka’s Epistemology and Philosophy of Science. Matti Sintonen (ed.). Poznan Studies in the Philosophy of Science and the Humanities. Amsterdam: Rodopi. Bromberger, Sylvain and Morris Halle. 1996. The Content of Phonological Signs. Unpublished ms. Brook, Andrew. 1994. Kant and the Mind. Cambridge: Cambridge University Press. Brown, C. 1993. Belief States and Narrow Content. Mind and Language 8: 343–367. Brown, Deborah. 1992. “Swampman of La Mancha,” and Other Tales about Representation. Doctoral Dissertation, University of Toronto. ———. A Furry Tile about Mental Representation. Philosophical Quarterly 185: 448–466. Brown, Jessica. 1995. The Incompatibility of Anti-Individualism and Privileged Access. Analysis 55(3): 149–156. ———. 1999. Boghossian on Externalism and Privileged Access. Analysis 59(1): 52–59. ———. 2000. Critical Reasoning, Understanding, and Self-Knowledge. Philosophy and Phenomenological Research 61: 659–676. Brueckner, Anthony. 1986 Brains in a Vat. Journal of Philosophy 83: 148–167. ———. 1990. Scepticism about Knowledge of Content. Mind 99: 447– 451. ———. 1992a. Semantic Answers to Skepticism. Pacific Philosophical Quarterly 73: 200–219. ———. 1992b. If I Am a Brain in a Vat, Then I Am Not a Brain in a Vat. Mind 101: 123–128. ———. 1992c. What an Anti-Individualist Knows A Priori. Analysis 52: 111–118. ———. 1993. Skepticism and Externalism. Philosophia 22: 169–171. ———. 1994. Knowledge of Content and Knowledge of the World. Philosophical Review 103: 327–343. ———. 1994. Ebbs on Skepticism, Objectivity, and Brains in Vats. Pacific Philosophical Quarterly 75: 77–87. ———. 1995a. The Characteristic Thesis of Anti-Individualism. Analysis 55(3): 146–148. ———. 1995b. Trying to Get Outside Your Own Skin. Philosophical Topics 23: 79–111. ———. 1996. Modest Transcendental Arguments. Philosophical Perspectives 10: 265–280. ———. 1997a. Externalism and Memory. Pacific Philosophical Quarterly 78(1): 1–12. ———. 1997b. Is Scepticism about Self-Knowledge Incoherent? Analysis 4: 287–290. ———. 1998. Content Externalism and A Priori Knowledge. Protosociology 11: 149–159. ———. 1999a. Two Recent Approaches to Self-Knowledge. In Philosophical Perspectives 13, Epistemology, James E. Tomberlin (ed.). Oxford: Blackwell. ———. 1999b. Transcendental Arguments from Content Externalism. In Transcendental Arguments: Problems and Prospects, Robert Stern (ed.). Oxford: Clarendon Oxford Press. ———. 1999c. Difficulties in Generating Skepticism about Knowledge of Content. Analysis 59: 59–63. ———. 2000a. Externalism and the A Prioricity of Self-Knowledge. Analysis 60: 132–136. ———. 2000b. Ambiguity and Knowledge of Content. Analysis 60: 257–260. ———. 2001a. A Priori Knowledge of the World Not Easily Available. Philosophical Studies 104: 109–114. ———. 2001b. Defending Burge’s Thought Experiment. Erkenntnis 55: 387–391. ———. Forthcoming. Problems for a Recent Account of Introspective Knowledge. Facta Philosophica. Bruns, Manfred and Gianfranco Soldati. 1994. Object-Dependent and Property-Dependent Contents. Dialectica. 48: 185–208. Buekens, F. 1994. Externalism, Content, and Causal Histories. Dialectica 48: 267–286. Buller, David J. 1992. “Narrow”-Mindedness Breeds Inaction. Behavior and Philosophy 20: 59–70. ———. 1997. Individualism and Evolutionary Psychology (or: In Defense of “Narrow” Functions). Philosophy of Science 64: 74–95.
478
Secondary Sources on Tyler Burge
Butler, Keith. 1996a. Individualism and Marr’s Computational Theory of Vision. Mind and Language 11(4): 313–337. ———. 1996b. Content, Causal Powers, and Context. Philosophy of Science 63: 105–114. ———. 1996c. Content, Computation, and Individualism in Vision Theory. Analysis 56: 146–154. ———. 1997. Externalism, Internalism, and Knowledge of Content. Philosophy and Phenomenological Research 57(4): 773–800. ———. 1998a. Externalism and Scepticism. Dialogue 37(1): 13–34. ———. 1998b. Content, Computation, and Individuation. Synthese 114: 277–292. ———. 2000. Problems for Semantic Externalism and “A Priori” Refutations of Skeptical Arguments. Dialectica 54: 29–49. Campbell, J. 1982. Extension and Psychic State: Twin Earth Revisited. Philosophical Studies 42: 67–89. Cassam, Quassim (ed.). 1994. Self-Knowledge. Oxford: Oxford University Press. Chapuis, Andre. 1998. Recent Theories of Truth. Journal of Indian Council of Philosophical Research 15(2): 89–123. Christensen, David. 1992. Causal Powers and Conceptual Connections. Analysis 52: 163–168. Christensen, David and Hilary Kornblith. 1997. Testimony, Memory, and the Limits of the “A Priori.” Philosophical Studies 86(1): 1–20. Churchland, Paul. 1981. Eliminative Materialism and the Propositional Attitudes. Journal of Philosophy 78: 67–90. Reprinted in Folk Psychology and the Philosophy of Mind, Scott Christensen and Dale Turner (eds.), Erlbaum, 1993. Clark, Andy and David Chalmers. 1998. The Extended Mind. Analysis 58: 7–19. Conee, Earl. 1987. Evident, But Rationally Unacceptable. Australasian Journal of Philosophy 65: 316–326. Corbi, Josep E. 1998. A Challenge to Boghossian’s Incompatibilist Argument. In Concepts (Philosophical Issues 9), Enrique Villanueva (ed.). Ridgeview: Atascadero. Cowart, Monica R. 1997. Externalism and First Person Authority: Do We Have Knowledge of Our Own Thoughts? Contemporary Philosophy 19(6): 2–10. Crane, Tim. 1991. All the Difference in the World. Philosophical Quarterly 41(162): 1–25. ———. (ed.). 1992. The Contents of Experience: Essays on Perception. Cambridge: Cambridge University Press. Crimmins, M. 1992a. Talk About Beliefs. Cambridge, MA: MIT Press. ———. 1992b. Tacitness and Virtual Beliefs. Mind and Language 7: 240–263. Cummins, R. 1991. Methodological Reflections on Belief. In Mind and Common Sense, R. Bogdan (ed.). Cambridge: Cambridge University Press. Davidson, Donald. 1979. Moods and Performances. In Meaning and Use, A. Margalit (ed.). Dordrecht: D. Reidel. Reprinted in Davidson (1984). ———. 1980. The Individuation of Events. In his Essays on Actions and Events. Oxford: Clarendon Press. ———. 1983. A Coherence Theory of Truth and Knowledge. In Kant oder Hegel? D. Henrich (ed.). Stuttgart: Klett-Cotta. Reprinted in (1986), Truth and Interpretation, Lepore (ed.), Oxford: Blackwell. ———. 1984. Inquiries Into Truth and Interpretation. Oxford: Oxford University Press. ———. 1987. Knowing One’s Own Mind. Proceedings and Addresses of the American Philosophical Association 60: 441–458. ———. 1988. Reply to Burge. Journal of Philosophy 85: 664–666. ———. 1991. Epistemology Externalized. Dialectica 45: 191–202. ———. 1999. Reply to Tyler Burge. In The Philosophy of Donald Davidson: The Library of Living Philosophers, vol. 27, Lewis Hahn (ed.). La Salle: Open Court.
Secondary Sources on Tyler Burge
479
Davies, Martin. 1986a. Individualism and Supervenience: Externality, Psychological Explanation, and Narrow Content. Aristotelian Society 60: 263–283. ———. 1986b. Externality, Psychological Explanation, and Narrow Content. Proceedings of the Aristotelian Society 60: 263–283. ———. 1989. Connectionism, Modularity and Tacit Knowledge. British Journal for the Philosophy of Science 40: 541–555. ———. 1991. Individualism and Perceptual Content. Mind 100(4): 461–484. ———. 1992. Perceptual Content and Local Supervenience. Proceedings of the Aristotelian Society 92: 21–45. ———. 1993. Aims and Claims of Externalist Arguments. In Philosophical Issues 4: Naturalism and Normativity, Enrique Villanueva (ed.). Atascadero: Ridgeview. ———. 1996. Externalism and Experience. Philosophy and Cognitive Science: Categories, Consciousness and Reasoning, A. Clark, J. Ezquerro, and J. M. Larrazabal (eds.). Dordrecht: Kluwer. ———. 1998. Externalism, Architecturalism, and Epistemic Warrant. In Knowing Our Own Minds, C. Wright, B. Smith, and C. Macdonald (eds.). Oxford: Oxford University Press. ———. 2000. Externalism and Armchair Knowledge. In New Essays on the A Priori, P. Boghossian and C. Peacocke (eds.). Oxford: Oxford University Press. ———. Forthcoming. Externalism, Self-knowledge and Transmission of Warrant. In Meaning, Basic SelfKnowledge, and Mind: Essays on Tyler Burge, M. J. Frápolli and E. Romero (eds.). Stanford: CSLI Publications. Davis, Steven. 1992. Intentionalité et Sceptisme. In Langage et Intentionalité, François Lepage and Daniel Laurier (eds.). Paris: Vrin-Bellemarin. ———. 1994. Individualism and Speech Acts. In Foundations of Speech Act Theory, S. L. Tsohatzidis (ed.). London: Routledge. ———. 1994. Anti-Individualism, Responsibility, Deference, and Dissembling. In Pretending to Communicate, Herman Parett (ed.). Berlin: de Gruyter. ———. 1999. Externalism, De Dicto Belief, Proper Names and Reference Determination. In Intentionality and Consciousness, D. Fisette (ed.). Dordrecht: Kluwer. ———. 2000. Deference and Conversation. In Metarepresentations, D. Sperber (ed.). Oxford: Oxford University Press. ———. Forthcoming a. Arguments for Externalism. In Meaning, Basic Self-Knowledge, and Mind: Essays on Tyler Burge, M. J. Frápolli and E. Romero (eds.). Stanford: CSLI Publications. ———. Forthcoming b. Conversation, Epistemology, and Norms. Mind and Language. Dennett, Daniel C. 1983. Beyond Belief. In Thought and Object, A. Woodfield (ed.). Oxford: Oxford University Press. Reprinted in his (1987). ———. 1987. Evolution, Error, and Intentionality. In his The Intentional Stance. Cambridge, MA: MIT Press. ———. 1991. Consciousness Explained. Boston: Little Brown. ———. 1996. Cow-Sharks, Magnets, and Swampmen. Mind and Language 11: 76–77. Devitt, Michael. 1990a. Meanings Just Ain’t in the Head. In Meaning and Method, George Boolos (ed.). Cambridge: Cambridge University Press. ———. 1990b. The Narrow Representational Theory of Mind. In Mind and Cognition, W. Lycan (ed.). Oxford: Blackwell. de Vries, Willem. 1990. Burgeoning Skepticism. Erkennnis 34: 141–164. ———. 1995. Coming to Our Senses: A Naturalistic Program for Semantic Localism. Cambridge: Cambridge University Press. ———. 1996. Experience and the Swamp Creature. Philosophical Studies 82: 55–80.
480
Secondary Sources on Tyler Burge
———. 1989. A Narrow Representation Theory of the Mind. In ReRepresentation, Silvers (ed.). Dordrecht: Kluwer. Donnellan, Keith. 1983. Putnam and Kripke on Natural Kind Terms. In Knowledge and Mind, Genet and Shoemaker (eds.). Oxford: Oxford University Press. ———. 1993. There Is a Word for That Kind of Thing: An Investigation of Two Thought Experiments. Philosophical Perspectives 7: 155–171. Dretske, Fred. 1981. Knowledge and the Flow of Information. Cambridge, MA: MIT Press/A Bradford Book. ———. 1988. Explaining Behavior. Cambridge, MA: MIT Press/A Bradford Book. ———. 1992. What Isn’t Wrong with Folk Psychology. Metaphilosophy 23: 1–13. ———. 1993. The Nature of Thought. Philosophical Studies 70: 185–199. ———. 1995. Naturalizing the Mind. Cambridge, MA: MIT Press/A Bradford Book. ———. 1996. Phenomenal Externalism, or If Meanings Ain’t in the Head, Where Are Qualia? In Philosophical Issues 7: Perception, E. Villanueva (ed.). Atascadero: Ridgeview. ———. 1996. Absent Qualia. Mind and Language 11: 78–85. ———. 2000. Entitlement: Epistemic Rights without Epistemic Duties? Philosophy and Phenomenological Research 60: 591–606. Ebbs, G. 1992. Skepticism, Objectivity, and Brains in Vats. Pacific Philosophical Quarterly 73: 239–266. ———. 1992. Realism and Rational Inquiry. Philosophical Topics 20: 1–33. ———. 1996. Can We Take Our Words at Face Value? Philosophy and Phenomenological Research 56: 499–530. ———. 1997. Rule-Following and Realism. Cambridge, MA: Harvard University Press. ———. 1998. Bilgrami’s Theory of Belief Content. Philosophy and Phenomenological Research 58: 613–620. ———. 2000. The Very Idea of Sameness of Extension Across Time. American Philosophical Quarterly 37: 245–268. ———. 2001. Is Skepticism about Self-Knowledge Coherent? Philosophical Studies 105: 43–58. ———. 2002. Truth and Trans-Theoretical Terms. In Hilary Putnam: On Pragmatism and Realism. Urszula Zeglen and James Conant (eds.). London: Routledge. Edwards, Jim. 1998. The Simple Theory of Colour and the Transparency of Sense Experience. In Knowing our Own Minds, C. Wright, B. Smith, and C. Macdonald (eds.). Oxford: Oxford University Press. ———. 2000. Burge on Testimony and Memory. Analysis 60(1): 124–131. Edwards, Steven D. 1994. Externalism in the Philosophy of Mind. Brookfield: Avebury. Egan, Frances. 1991. Must Psychology Be Individualistic? Philosophical Review 100(2): 179–203. ———. 1992. Individualism, Computation, and Perceptual Content. Mind 101: 443–459. ———. 1995. Computation and Content. Philosophical Review 104(2): 181–203. ———. 1996. Intentionality and the Theory of Vision. In Perception, Kathleen Akins (ed.). Oxford: Oxford University Press. ———. 1999. In Defense of Narrow Mindedness. Mind and Language 14(2): 177–194. Elugardo, Reinaldo. 1993. Burge on Content. Philosophy and Phenomenological Research 53(2): 367–384. Engel, P. 1984. Functionalism, Belief, and Content. In The Mind and the Machine, Steve Torrance (ed.). Chichester: Horwood. Englebretsen, George. 1981. Predicates, Predicables, and Names. Critica 13: 105–108. Evans, Gareth. 1982. The Varieties of Reference. New York: Oxford University Press. ———. 1985. Collected Papers. Oxford: Oxford University Press. Factor, R. Lance. 1975. A Note on the Analysis of Mass Terms. Southern Journal of Philosophy 13: 247–249. Falvey, Kevin. 2000. Knowledge in Intention. Philosophical Studies 99: 21–44.
Secondary Sources on Tyler Burge
481
———. 2000. The Compatibility of Anti-individualism and Privileged Access. Analysis 60(1): 137–142. ———. Forthcoming. The Basis of First Person Authority. Philosophical Topics. Falvey, Kevin and Joseph Owens. 1994. Externalism, Self-Knowledge and Skepticism. Philosophical Review 103(1): 107–137. Faulkner, Peter. 2000. The Social Character of Testimonial Knowledge. Journal of Philosophy 97: 581–601. Ferraiolo, William. 1996. Individualism and Descartes. Teorema. Field, Hartry. 1989. “Narrow” Aspects of Intentionality and the Information-Theoretic Approach to Content. In Information, Semantics, and Epistemology, E. Villanueva (ed.). Oxford: Blackwell. ———. 1994. Deflationist Views of Meaning and Content. Mind 103(411): 249–285. Fodor, Jerry A. 1980. Methodological Solipsism as a Research Strategy in Cognitive Psychology. Behavioral and Brain Sciences 3: 63–109. ———. 1982. Cognitive Science and the Twin-Earth Problem. Notre Dame Journal of Formal Logic 23: 98–118. ———. 1983. Modularity of Mind. Cambridge, MA: MIT Press. ———. 1986. Individualism and Supervenience. Aristotelian Society (suppl. vol.) 60: 235–262. ———. 1987. Psychosemantics: The Problem of Meaning in the Philosophy of Mind. Cambridge, MA: MIT Press. ———. 1990a. A Theory of Content. Cambridge, MA: MIT Press. ———. 1991. A Modal Argument for Narrow Content. Journal of Philosophy 88: 5–26. ———. 1994. The Elm and the Expert. Cambridge, MA: MIT Press. Forbes, Graeme. 1987. A Dichotomy Sustained. Philosophical Studies 51: 187–211. ———. 1989. Cognitive Architecture and the Semantics of Belief. Midwest Studies in Philosophy 14: 84–100. ———. 1995. Realism and Skepticism: Brains in a Vat Revisited. Journal of Philosophy 92: 205–222. Frances, Bryan. 1999. On the Explanatory Deficiencies of Linguistic Content. Philosophical Studies 93(1): 45–75. Francescotti, Robert M. 1991. Externalism and Marr’s Theory of Vision. British Journal for the Philosophy of Science 42(2): 227–238. M. J. Frápolli and E. Romero (eds.). Forthcoming. Meaning, Basic Self-Knowledge, and Mind: Essays on Tyler Burge. Stanford: CSLI Publications. Gaifman, Haim. 1996. Is the “Bottom-Up” Approach from the Theory of Meaning to Metaphysics Possible? Journal of Philosophy 93(8): 373–407. Gajdusek, Carleton, Guy McKhann, and Laiana Bolis (eds.). 1994. Evolution and Neurology of Language: Discussions in Neuroscience 10(1–2), April. Gallistel, Randy. 1990. The Organization of Learning. Cambridge, MA: MIT Press. Gallois, Andre. 1994. Deflationary Self-Knowledge. In Philosophy in Mind: The Place of Philosophy in the Study of Mind, M. Michael and J. O’Leary-Hawthorne (eds.). Dordrecht: Kluwer. ———. 1996. The World without the Mind within: An Essay in First Person Authority. Cambridge: Cambridge University Press. Gallois, Andre and John O’Leary-Hawthrone. 1996. Externalism and Scepticism. Philosophical Studies 81(1): 1–26. Garcia-Carpintero, M. 1994. The Supervenience of Mental Content. Proceedings of the Aristotelian Society 68: 117–135. Garfield, Jay L. 1988. Belief in Psychology: A Study in the Ontology of Mind. Cambridge, MA: MIT Press. Gauker, Christopher. 1987. Mind and Chance. Canadian Journal of Philosophy 17: 533–552. ———. 1987. Language as Tool. American Philosophical Quarterly 24: 47–58.
482
Secondary Sources on Tyler Burge
———. 1991. Mental Content and the Division of Epistemic Labour. Australasian Journal of Philosophy 69(3): 302–318. ———. 1994. Thinking Outloud. Princeton: Princeton University Press. Gelman, S. A., J. D. Coley, and G. M. Gottfried. 1994. Essentialist Beliefs in Children: The Acquisition of Concepts and Theories. In Mapping the Mind: Domain Specificity in Cognition and Culture, Hirschfeld and Gelman (eds.). Cambridge: Cambridge University Press. Georgalis, Nicholas. 1990. No Access for the Externalist: Discussion of Heil’s “Privileged Access.” Mind 100: 101–108. ———. 1994. Asymmetry of Access to Intentional States. Erkenntnis 40: 185–211. ———. 1999. Rethinking Burge’s Thought Experiment. Synthese 118(2): 145–164. Gibbons, John. 1993. Identity without Supervenience. Philosophical Studies 70(1): 59–79. ———. 1996. Externalism and Knowledge of Content. Philosophical Review 105(3): 287–310. Gilbert, Daniel T. 1991. How Mental Systems Believe. American Psychologist 46: 107–119. Globus, G. 1984. Can Methodological Solipsism be Confined to Psychology? Cognition and Brain Theory 7: 233–246. Glock, Hanjo and John Preston. 1995. Externalism and First Person Authority. Monist 78(4): 515–533. Glymour, Clark. 1999. A Mind Is a Terrible Thing to Waste—Critical Notice: Jaegwon Kim, “Mind in a Physical World.” Philosophy of Science 66(3): 455–471. Godfrey-Smith, Peter. 1996. Complexity and the Function of Mind in Nature. Cambridge: Cambridge University Press. Goldberg, Sanford C. 1997. Self-Ascription, Self-Knowledge, and the Memory Argument. Analysis 57(3): 211–219. ———. 1999a. The Relevance of Discriminatory Knowledge of Content. Pacific Philosophical Quarterly 80(2): 136–156. ———. 1999b. The Psychology and Epistemology of Self-Knowledge. Synthese 118: 165–201. ———. 2000. Externalism and Authoritative Knowledge of Content: A New Incompatibilist Strategy. Philosophical Studies 100: 51–79. Goldstick, Dan. 1986. What Are “Purely Qualitative” Terms? American Philosophical Quarterly 23: 71–81. Graham, Peter J. 1997. What Is Testimony? Philosophical Quarterly 47: 227–232. ———. The Reliability of Testimony. Philosophy and Phenomenological Research 61: 695–709. ———. 2000. Conveying Information. Synthese 123: 365–392. ———. 2000. Transferring Knowledge. Noûs 34: 131–152. Grandy, Richard E. 1977. Predication and Singular Terms. Noûs 11: 163–167. Green, Mitchell. 2001. The Status of Supposition. Noûs 34: 376–399. Grice, P. 1989. Studies in the Ways of Words. Cambridge, MA: Harvard University Press. Grimm, Robert H. and Daniel D. Merrill (eds.). 1988. Contents of Thought. Tucson: University of Arizona Press. Hahn, Martin. 1994. How Not to Draw the De Re De Dicto Distinction. In The Locial Foundations of Cognition, Macnamara and Gonzales (eds.). Oxford: Oxford University Press. Hall, Lisa L. 1998. The Self-Knowledge That Externalists Leave Out. Southwest Philosophy Review 14: 115–123. Halle, Morris and Alec Marantz. 1993. Distributed Morphology and the Pieces of Inflection. In The View from Building 20, K. Hale and S. J. Keyser (eds.). Cambridge, MA: MIT Press. Hardcastle, V. G. 1997 [Explanation] is Explanation Better. Philosophy of Science. 64: 154–60. Harman, Gillbert. 1982. Conceptual Role Semantics. Notre Dame Journal of Formal Logic 23: 242–256.
Secondary Sources on Tyler Burge
483
———. 1990. The Intrinsic Quality of Experience. In Philosophical Perspectives 4, Action Theory and Philosophy of Mind, James Tomberlin (ed.), pp. 31–52. Atascadero: Ridgeview. ———. 1996. Explaining Objective Color in Terms of Subjective Reactions. In Philosophical Issues 7: Perception, Enrique Villanueva (ed.), pp. 1–17. Atascadero: Ridgeview. Hatfield, Gary. 1994. Philosophy of Psychology as Philosophy of Science. Proceedings of the Biennial Meetings of the Philosophy of Science Association. 2: 19–23. Heal, J. 1998. Externalism and Memory. Aristotelian Society Supplement 72: 77–94. Heil, John. 1987. Are We Brains in a Vat: Top Philosophers Say “No.” Canadian Journal of Philosophy 17: 427–436. ———. 1988. Privileged Access. Mind 97: 238–251. ———. 1989. Intentionality Speaks for Itself. In ReRepresentation, S. Silvers (ed.). Dordrecht: Kluwer. Heil, John and Alfred Mele. 1991. Mental Causes. American Philosophical Quarterly 28: 61–71. Hickey, Lance P. 1999. The Chomskian Challenge to Externalism. International Studies in Philosophy 31: 39– 51. Higginbotham, James. 1986. Linguistic Theory and Davidson’s Program in Semantics. In Truth and Interpretation: Perspectives on the Philosophy of Donald Davidson, E. LePore (ed.), pp. 29–48. Oxford: Blackwell. ———. 1988. Contexts, Models, and Meanings: A Note on the Data of Semantics. In Mental Representations: The Interface Between Language and Reality, R. Kempson (ed.), pp. 29–48. Cambridge: Cambridge University Press. ———. 1989. Elucidations of Meaning. Linguistics and Philosophy 12: 465–517. ———. 1990. Frege, Concepts, and the Design of Language. In Information, Semantics, and Epistemology. Villanueva (ed.), Oxford: Blackwell. ———. 1991a. Remarks on the Metaphysics of Linguistics. Linguistics and Philosophy 14(5): 555–566. ———. 1991b. Belief and Logical Form. Mind and Language 6: 344–369. ———. 1994. Priorities in the Philosophy of Thought. Proceedings of the Aristotelian Society (suppl.) 68: 85–106. ———. 1998a. Conceptual Competence. In Philosophical Issues 9: Concepts, Enrique Villanueva (ed.). Ridgeview: Atascadero. ———. 1998b. Response to Commentators. In Philosophical Issues 9: Concepts, Enrique Villanueva (ed.). Ridgeview: Atascadero. Hill, Christopher. 1991. Sensations: A Defense of Type Materialism. Cambridge: Cambridge University Press. Ho, Jih Ching. 1995. Individualism, Reductionism, and the Concept of Supervenience. In Mind and Cognition, Yu-Houng Houng (ed.). Taipei: Institute for European and American Studies. Holton, Gerald. 1996. On the Art of Scientific Imagination. Daedalus 183–208. Honderich, Ted. 1993. The Union Theory and Anti-Individualism. In Mental Causation, John Heil (ed.). New York: Oxford University Press. Hornsby, Jennifer. 1976. Proper Names: A Defence of Burge. Philosophical Studies 30: 227–234. Horowitz, Amir. 1995. Putnam, Searle, and Externalism. Philosophical Studies 81: 27–69. Houghton, D. 1997. Mental Content and External Representations: Internalism, Anti-Internalism. Philosophical Quarterly 47: 159–177. Humberstone, Lloyd. 1992. Direction of Fit. Mind 101: 59–83. Humphrey, N. 1992. A History of the Mind. New York: Simon and Schuster. Hurley, S. L. 1998. Vehicles, Contents, Conceptual Structure, and Externalism. Analysis 58: 1–6. Hutchins, Edwin. 1995. Cognition in the Wild. Cambridge, MA: MIT Press.
484
Secondary Sources on Tyler Burge
Hymers, Michael. 1997. Realism and Self-Knowledge: A Problem for Burge. Philosophical Studies 86(3): 303– 325. Jackendoff, Ray. 1994. Patterns in the Mind. New york: Basic Books. Jackman, Henry. 1997. Individualism and Interpretation. Southwest Philosophy Review 14(1): 31–38. ———. 1999. We Live Forwards but Understand Backwards: Linguistic Practices and Future Behavior. Pacific Philosophical Quarterly 80(2): 157–177. Jackson, Frank. 1982. Epiphenomenal Qualia. Philosophical Studies 32: 127–136. ———. 1996. Mental Causation. Mind 105(419): 377–413. ———. 1997. Perception. Cambridge: Cambridge University Press. Jackson, Frank and Philip Pettit. 1988. Functionalism and Broad Content. Mind 97: 318–400. ———. 1993. Some Content Is Narrow. In Mental Causation, J. Heil and A. Mele (eds.). Oxford: Oxford University Press. Jacob, Margaret. 1991. Living the Enlightenment. New York: Oxford University Press. Jacob, Pierre. 1987a. Thoughts and Belief Ascriptions. Mind and Language 2: 301–325. ———. 1987b. Comment Rapporter Une Pensée? Studia Philosophica 46: 77–97. ———. 1990. Externalism Revisited. Is There Such a Thing as Narrow Content? Philosophical Studies 60(3): 143–176. ———. 1992. Externalism and Mental Causation. Proceedings of the Aristotelian Society 66: 203–219. ———. 1993. Externalism and the Explanatory Relevance of Broad Content. Mind and Language 8(1): 131–156. Johnston, Mark. 1988. Self-Deception and the Nature of Mind. In Perspectives on Self-Deception, Brian P. McLaughlin and Amelie Oksenberg Rorty (eds.). Berkeley: University of California Press. Katz, J. 1990. The Domino Theory. Philosophical Studies 58: 3–39. Kazez, J. R. 1994. Computationalism and the Causal Role of Content. Philosophical Studies 75: 231–260. Kemp, Gary. 1996. Frege’s Sharpness Requirement. Philosophical Quarterly 46(183): 168–184. Kim, Jaegwon. 1993. Supervenience and Mind: Selected Philosophy Essays. Cambridge: Cambridge University Press. ———. 1995. Mental Causation: What? Me Worry? In Philosophical Issues 6: Content, Enrique Villanueva (ed.). Atascadero: Ridgeview. ———. 1998. Mind in a Physical World. Cambridge, MA: MIT Press. Kimbrough, Scott. 1998. Anti-Individualism and Fregeanism. Philosophical Quarterly 48(193): 470–482. Kitcher, P. S. 1984. Narrow Taxonomy and Wide Functionalism. Philosophy of Science 52: 78–97. ———. 1988. Marr’s Computational Theory of Vision. Philosophy of Science 55: 1–24. Klein, Martha. 1996. Externalism, Content, and Causation. Proceedings of the Aristotelian Society 96: 159–176. Kobes, Bernard W. 1989. Semantics and Psychological Prototypes. Pacific Philosophical Quarterly 70: 1–18. ———. 1990. Individualism and Artificial Intelligence. In Philosophical Perspectives, 4: Action Theory and Philosophy of Mind, Tomberlin (ed.). pp. 429–456. Atascadero: Ridgeview. ———. 1995. Telic Higher-Order Thoughts and Moore’s Paradox. Philosophical Perspectives 9: 291–312. ———. 1996. Mental Content and Hot Self-Knowledge. Philosophical Topics 24: 71–99. Koethe, J. 1992. And They Ain’t Outside the Head Either. Synthese 90: 27–53. Koons, Robert Charles. 1992. Paradoxes of Belief and Strategic Rationality. Cambridge: Cambridge University Press. Koslicki, Kathrin. 1999. The Semantics of Mass-Predicates. Noûs 33(1): 46–91. Kripke, Saul. 1972. Naming and Necessity. In Semantics of Natural Language, Davidson and Harman (eds.), pp. 253–355. Dordrecht: Reidel.
Secondary Sources on Tyler Burge
485
———. 1980. Naming and Necessity. Cambridge, MA: Harvard University Press. Kroon, Frederick. 1993. Rationality and Epistemic Paradox. Synthese 94(3): 377–408. Lalor, B. J. 1997. It Is What You Think: Intentional Potency and Anti-Individualism. Philosophical Psychology 10: 165–178. Lantin, Robert. 1994. Individualism, Physicalism, and Spinoza on Minds and Bodies. Manuscrito: Revista Internacional de Filosofia 17: 35–64. Larkin, William S. 1999. Brute Error with Respect to Content. Philosophical Studies 94(1–2): 159–171. Larson, Richard and Gabriel Segal. 1995. Knowledge of Meaning. Cambridge, MA: MIT Press. Lee, Byeong D. 1998a. Burge on Epistemic Paradox. Canadian Journal of Philosophy 28(3): 337–348. ———. 1998b. Kroon on Rationality and Epistemic Paradox. Southwest Philosophy Review 14(2): 169–174. Lenzen, Wolfgang, 1981. Doxastic Logic and the Burge–Buridan Paradox. Philosophical Studies 39: 43–49. LePore, E. 1990. Subjectivism and Environmentalism. Inquiry 33: 197–214. LePore, E. and Loewer, B. 1986. Solipsistic Semantics. Midwest Studies in Philosophy 10: 595–614. ———. 1989. Dual Aspect Semantics. In ReRepresentation, S. Silvers (ed.). Dordrecht: Kluwer. Lewis, David. 1990. What Experience Teaches. In Mind and Cognition: A Reader, Lycan (ed.). Oxford: Blackwell. Liu, JeeLoo. Forthcoming. Physical Externalism and Social Externalism: Are They Really Compatible? Journal of Philosophical Research. Loar, Brian. 1987. Subjective Intentionality. Philosophical Topics 15: 89–124. ———. 1988a. Social Content and Psychological Content. In Contents of Thought, Grimm and Merril (eds.). Tucson: University of Arizona Press. ———. 1988b. A New Kind of Content. In Contents of Thought, Grimm and Merril (eds.). Tucson: University of Arizona Press. ———. 1994. Self-Interpretation and the Constitution of Reference. Philosophical Perspectives 8: 51–74. ———. 1995. Reference from the First-Person Perspective. In Contents, Villanueva (ed.). Atascadero: Ridgeview. ———. 1998. Is There a Good Epistemological Argument against Concept-Externalism? In Philosophical Issues 9: Concepts, Enrique Villanueva (ed.). Ridgeview: Atascadero. ———. 1990. Phenomenal States. In Philosophical Perspectives 4, Action Theory and Philosophy of Mind, James Tomberlin (ed.). Atascadero: Ridgeview, 81–108. A much revised version of this paper is to be found in Block, Flanagan, and Güzeldere (1997), The Nature of Consciousness: Philosophical Debates, 597–616. Cambridge: MIT Press. Lockwood, Michael. 1975. On Predicating Proper Names. Philosophical Review 84: 471–498. Lormand, Eric. 1996. How to Be a Meaning Holist. Journal of Philosophy 93(2): 51–73. Losonsky, Michael. 1995. Embedded Systems versus Individualism. Minds and Machines 5: 357–371. Ludlow, Peter. 1995a. Social Externalism, Self-Knowledge, and Memory. Analysis 55: 157–159. ———. 1995b. Externalism, Self-Knowledge, and the Prevalence of Slow Switching. Analysis 55: 45–49. ———. 1995c. Social Externalism and Memory: A Problem? Acta Analytica 10: 69–76. ———. 1997. On the Relevance of Slow Switching. Analysis 57(4): 285–286. Ludlow, Peter and Norah Martin (eds.). 1998. Externalism and Self-Knowledge. Stanford: CSLI Publications. Ludwig, Kirk A. 1993. Externalism, Naturalism, and Method. In Philosophical Issues 4: Naturalism and Normativity, Enrique Villanueva (ed.). Atascadero: Ridgeview. ———. 1993. Causal Relevance and Thought Content. Philosophical Quarterly 44: 334–353. ———. 1996. Duplicating Thoughts. Mind and Language 11: 92–102.
486
Secondary Sources on Tyler Burge
Lycan, William G. 1987. Consciousness. Cambridge, MA: MIT Press. ———. 1995. We’ve Only Just Begun. Behavioral and Brain Sciences 18: 262–263. ———. 1996a. Layered Perceptual Representation. In Philosophical Issues 7: Perception, Enrique Villanueva (ed.), pp. 81–100. Atascadero: Ridgeview. ———. 1996b. Consciousness and Experience. Cambridge, MA: MIT Press. ———. 1996c. Replies to Tomberlin, Tye, Stalnaker, and Block. In Philosophical Issues 7: Perception, E. Villanueva (ed.). Atascadero: Ridgeview. Lyons, John. 1977. Semantics. 2 vols. Cambridge: Cambridge University Press. Macdonald, Cynthia. 1990. Weak Externalism and Mind-Body Identity. Mind 99: 387–404. ———. 1992. Weak Externalism and Psychological Reduction. In Reduction, Explanation and Realism, D. Charles and K. Lennon (eds.). Oxford: Oxford University Press. ———. 1995a. Anti-Individualism and Psychological Explanation. In Philosophy of Psychology: Debates on Psychological Explanation, Cynthia Macdonald (ed.). Oxford: Blackwell. ———. 1995b. Externalism and First Person Authority. Synthese 104(1): 99–122. ———. 1998a. Externalism and Norms. In Current Issues in Philosophy of Mind, Anthony O’Hear (ed.). Cambridge: Cambridge University Press. ———. 1998b. Externalism and Authoritative Self-Knowledge. In Knowing Our Own Minds, C. Wright, P. Smith, and C. Macdonald (eds.). Oxford: Oxford University Press. Macdonald, C., P. Smith, and C. Wright. 1998. Knowing Our Own Minds: Essays in Self-Knowledge. Oxford: Oxford University Press. Maloney, J. Christopher. 1985. Methodological Solipsism Reconsidered as a Research Strategy in Cognitive Psychology. Philosophy of Science 52: 451–469. ———. 1991. Saving Psychological Solipsism. Philosophical Studies 61: 267–283. Malpas, Jeff. 1994. Self-Knowledge and Scepticism. Erkenntnis 40: 165–184. Malt, Barbara. 1994. Water Is Not H2O. Cognitive Psychology 27: 41–70. Mandelkar, Steven. 1991. An Argument against the Externalist Account of Psychological Content. Philosophical Psychology 4: 375–382. Manfredi, Pat A. 1993. Two Routes to Narrow Content: Both Dead Ends. Philosophical Psychology 6: 3–22. Marr, David. 1982. Vision. San Francisco: W. H. Freeman. May, Todd G. 1995. The Limits of the Mental and the Limits of Philosophy: From Burge to Foucault and Beyond. Journal of Speculative Philosophy 9(1): 36–47. McClamrock, Ron. 1991. Methodological Individualism Considered as a Constitutive Principle of Scientific Inquiry. Philosophical Psychology 4: 343–354. ———. 1995. Existential Cognition: Computational Minds in the World. Chicago: University of Chicago Press. McCulloch, Gregory. 1988. Carruthers Repulsed. Analysis 48: 96–100. ———. 1990. Externalism and Experience. Analysis 50: 244–250. ———. 1992. The Spirit of Twin Earth. Analysis 52: 168–174. ———. 1995. The Mind and Its World. London: Routledge. ———. 1999. Content Externalism and Cartesian Scepticism: A Reply to Brueckner. In Transcendental Arguments: Problems and Prospects, Robert Stern (ed.). Oxford: Clarendon Oxford Press. McDermott, Michael. 1986. Narrow Content. Australasian Journal of Philosophy 64: 277–288. McDowell, John. 1977. On the Sense and Reference of a Proper Name. Mind 86: 159–185. ———. 1986. Singular Thought and the Extent of Inner Space. In Subject, Thought, and Context, McDowell and Pettit (eds.). New York: Oxford University Press.
Secondary Sources on Tyler Burge
487
———. 1994. The Content of Perceptual Experience. Philosophical Quarterly 44: 190–205. McGilvray, J. 1998. Meanings are Syntactically Individuated and Found in the Head. Mind and Language 13: 225–280. McGinn, Colin. 1982. The Structure of Content. In Thought and Object, A. Woodfield (ed.). Oxford: Oxford University Press. ———. 1989. Mental Content. Oxford: Blackwell. ———. 1991a. The Problem of Consciousness. Oxford: Blackwell. ———. 1991b. Conceptual Causation. Mind 100: 525–546. McKinsey, Michael. 1987. Apriorism in the Philosophy of Language. Philosophical Studies 52: 1–32. ———. 1991a. Anti-Individualism and Privileged Access. Analysis 51(1): 9–16. ———. 1991b. The Internal Basis of Meaning. Pacific Philosophical Quarterly 72: 143–169. ———. 1993. Curing Folk Psychology of Arthritis. Philosophical Studies 70(3): 323–336. ———. 1994a. Accepting the Consequences of Anti-Individualism. Analysis 54(2): 124–128. ———. 1994b. Individuating Beliefs. Philosophical Perspectives 8: 303–330. McLaughlin, Brian and Michael Tye. 1998a. Is Content-Externalism Compatible with Privileged Access? Philosophical Review 107(3): 349–380. ———. 1998b. Externalism, Twin Earth, and Self-Knowledge. In Knowing Our Own Minds, C. Macdonald, B. Smith, and C. Wright (eds.). Oxford: Oxford University Press. ———. 2000. Self-Knowledge, Externalism, and Skepticism. Aristotelean Society (suppl.) 74: 93–117. McLaughlin, Brian and Tye, Michael. 1998. The Brown McKinsey Charge of Inconsistency. In Externalism and Self-Knowledge, P. Ludlow and N. Martin (eds.). Stanford: CSLI Publications. Mehler, Jacques and Emmanuel Dupoux. 1994. What Infants Know. Oxford: Blackwell. Mercier, Adele. 1993. Normativism and the Mental: A Problem of Language Individuation. Philosophical Studies 22: 71–88. Mijuskovic, Ben Lazare. 1974. The Achilles of Rationalist Arguments. Dordrecht: Martinus Nijhoff. Michotte, A. 1963. The Perception of Causality. T. R. Miles and E. Miles (trans.). London: Methuen. (Trans. of 1946 French edition) Miller, George H. 1999. How Phenomenological Content Determines the Intentional Object. Husserl Studies 16: 1–24. Miller, Richard W. 1997. Externalist Self-Knowledge and the Scope of the A Priori. Analysis 57: 67–74. Millikan, Ruth Garrett. 1998. How We Make Our Ideas Clear: Empiricist Epistemology for Empirical Concepts. Proceedings and Addresses of the American Philosophical Association 72: 65–79. Miscevic, N. 1996. Computation, Content, and Cause. Philosophical Studies 82: 241–263. Montague, R. 1968. Pragmatics. Reprinted in Formal Philosophy: Selected Papers of Richard Montague, R. Thomason (ed.) 1974. New Haven: Yale University Press. Montgomery, R. 1995. Non-Cartesian Explanations Meet the Problem of Mental Causation. Southern Journal of Philosophy 33: 221–241. Moore, Gregory H. 1982. Zermelo’s Axiom of Choice: Its Origins, Development, and Influence. New York: Springer Verlag. Moran, Richard. 1997. Self-Knowledge: Discovery, Resolution, and Undoing. European Journal of Philosophy 5: 141–161. Moravcsik, Julius. 1990. Thought and Language. London: Routledge. Morton, P. 1993. Supervenience and Computational Explanation in Vision Theory. Philosophy of Science 60: 86–99.
488
Secondary Sources on Tyler Burge
Moya, Carlos J. 1998. Boghossian’s “Reductio” of Compatibilism. In Philosophical Issues 9: Concepts, Enrique Villanueva (ed.). Ridgeview: Atascadero. Noonan, Harold. 1984a. Fregean Thoughts. Philosophical Quarterly 34: 205–224. ———. 1984b. Methodological Solipsism: A Reply to Morris. Philosophical Studies 48: 285–290. ———. 1986. Russellian Thoughts and Methodological Solipsism. In Language, Mind, and Logic, J. Butterfield (ed.). Cambridge: Cambridge University Press. ———. 1990. Object-Dependent Thoughts and Psychological Redundancy. Analysis 51: 1–9. ———. 1993. Object-Dependent Thoughts: A Case of Superficial Necessity but Deep Contingency? In Mental Causation, J. Heil and A. Mele (eds.). Oxford: Oxford University Press. Normore, Calvin. 1986. Meaning and Objective Being: Descartes and His Sources. In Essays on Descartes’ Meditations, Amelie Rorty (ed.). Berkeley: University of California Press. Nuccetelli, Susana. (ed.). 2003. New Essays on Semantic Externalism and Self-knowledge. Cambridge, MA: MIT Press. Owens, Joseph. 1983. Functionalism and the Propositional Attitudes. Noûs 17: 529–549. ———. 1986. Synonymy and the Non-Individualistic Model of the Mental. Synthese 66: 361–382. ———. 1987. In Defense of a Different Doppelganger. Philosophical Review 96: 521–554. ———. 1992. Psychophysical Supervenience: Its Epistemological Foundation. Synthese 90: 89–117. ———. 1993. Content, Causation, and Psychophysical Supervenience. Philosophy of Science 60: 242–261. ———. 1994a. Psychological Externalism. In The Mind-Body Problem: A Guide to the Current Debate, Richard Warner (ed.). Oxford: Blackwell. ———. 1994b. Psychological Externalism and Psychological Explanation. Philosophy and Phenomenological Research 54: 921–928. Parsons, Charles. 1974. The Liar Paradox. Journal of Philosophical Logic 3: 381–412. ———. 1993. On Some Difficulties Concerning Intuition and Intuitive Knowledge. Mind 102: 233–246. Pateman, Trevor. 1987. Language in Mind and Language in Society. New York: Oxford University Press. Patterson, Sarah. 1990. The Explanatory Role of Belief Ascriptions. Philosophical Studies 59(3): 313–332. ———. 1991. Individualism and Semantic Development. Philosophy of Science 58: 15–35. ———. 1996. Success-Orientation and Individualism in the Theory of Vision. In Perception, Kathleen Akins (ed.). Oxford: Oxford University Press. Peacocke, Christopher. 1983. Sense and Content. Oxford: Oxford University Press. ———. 1987. Understanding Logical Constants: A Realist’s Account. Proceedings of the British Academy 73: 153–200. ———. 1992. A Study of Concepts. Cambridge, MA: MIT Press. ———. 1993a. How Are A Priori Truths Possible? European Journal of Philosophy 1: 175–199. ———. 1993b. Externalist Explanation. Proceedings of the Aristotelian Society 67: 203–230. ———. 1994. Content, Computation and Externalism. Mind and Language 9: 303–335. Reprinted in Philosophical Issues 6: Content, 1995, Enrique Villanueva (ed.). Atascadero: Ridgeview. ———. 1996. Entitlement, Self-Knowledge, and Conceptual Redeployment. Proceedings of the Aristotelian Society 96: 117–158. ———. 1999. Computation as Involving Content: A Response to Egan. Mind and Language 14: 195–202. Pereboom, D. 1995. Conceptual Structure and the Individuation of Content. Philosophical Perspectives 9: 401– 428. Pessin, Andrew and Sanford Goldberg (eds.). 1996. The Twin Earth Chronicles. New York: M. E. Sharpe. Petrie, B. 1990. Nonautonomous Psychology. Southern Journal of Philosophy 28: 539–559.
Secondary Sources on Tyler Burge
489
Pettit, Philip. 1986. Broad-Minded Explanation and Psychology. In Subject, Thought, and Context, P. Pettit and J. McDowell (eds.). Oxford: Oxford University Press. Pollock, John L. 1990. Understanding the Language of Thought. Philosophical Studies 58: 95–120. Prior, A. 1960. The Runabout Inference-Ticket. Analysis 21: 38–39. Pustejovsky, James. 1995. The Generative Lexicon. Cambridge, MA: MIT Press. Putnam, Hilary. 1981. Reason, Truth, and History. Cambridge: Cambridge University Press. ———. 1983. Computational Psychology and Interpretation Theory. In his Collected Papers, volume III. Cambridge: Cambridge University Press. ———. 1987. Fodor and Block on Narrow Content. In his (1988). ———. 1988. Representation and Reality. Cambridge, MA: MIT Press. ———. 1990. Is Water Necessarily H2O? In Realism with a Human Face, James Conant (ed.). Cambridge, MA: Harvard University Press. ———. 1996. Introduction. In The Twin Earth Chronicles: Twenty Years of Reflection on Hilary Putnam’s “The Meaning of ‘Meaning’,” Andrew Pessin and Sanford Goldberg (eds.). New York: M. E. Sharpe. Quillen, K. 1986. Propositional Attitudes and Psychological Explanation. Mind and Language 1: 133–157. Raffia, D. 1998. First-Person Authority and the Internal Reality of Beliefs. In Knowing Our Own Minds, C. Wright, B. Smith, and C. Macdonald (eds.). Oxford: Oxford University Press. Récanati, François. 1993. Direct Reference. Oxford: Blackwell. ———. 1994. How Narrow Is Narrow Content? Dialectica 48(3–4): 209–229. Rechenauer, Martin. 1993. How Not to Taxonomize Mental Kinds. Acta Analytica: Philosophy and Psychology 8: 134–141. ———. 1994. Intentionaler Realismus und Externalismus. Wurzburg: Konishausen and Neumann. ———. 1997. Individualism, Individuation, and That-Clauses. Erkenntnis 46: 49–67. Rey, Georges. 1992a. Sensational Sentences. In Consciousness: Psychological and Philosophical Essays, M. Davies and G. Humphreys (eds.). Oxford: Blackwell. ———. 1992b. Sensational Sentences Switched. Philosophical Studies 68: 289–331. ———. 1992c. Semantic Externalism and Conceptual Competence. Proceedings of the Aristotelian Society 66: 315–333. ———. 1997. Contemporary Philosophy of Mind. Oxford: Blackwell. ———. 2003. Chomsky, Intentionality and a CRTT. In Chomsky and His Critics, Antony and Hornstein (eds.). Oxford: Blackwell. Rhum, Michael. 1993. Understanding ‘belief’. Man 28(4). Richard, M. 1983. Direct Reference and Ascriptions of Belief. Journal of Philosophical Logic 12: 425–452. ———. 1990. Propositional Attitudes: An Essay on Thoughts and How We Ascribe Them. Cambridge: Cambridge University Press. Ricketts, T. 1996. Logic and Truth in Frege. Aristotelian Society (suppl. vol.) 70: 121–140. Robinson, William. 1998. Intrinsic Qualities of Experience: Surviving Harman’s Critique. Erkentniss 47: 285–309. Rosch, Eleanor. 1972. [E. R. Heider] Probabilities, Sampling and Ethnographic Method: The Case of Dani Colour Names. Man 7: 448–466. Rosenthal, David. 1991. The Nature of Mind. Oxford: Oxford University Press. Rowlands, Mark. 1995a. Externalism and Token-Token Identity. Philosophia 24: 359–375. ———. 1995b. Against Methodological Solipsism: The Ecological Approach. Philosophical Psychology 8: 5–24. ———. 1999. The Body in Mind. Cambridge: Cambridge University Press.
490
Secondary Sources on Tyler Burge
Rozemond, Marleen. 1994. Evans on De Re Thought. Philosophia 22: 275–298. Rudd, A. 1997. Two Types of Externalism. Philosophical Quarterly 47: 501–507. Russow, L. M. 1993. Fodor, Adams, and Causal Properties. Philosophical Psychology 6: 57–61. Saidel, E. 1994. Content and Causal Powers. Philosophy of Science 61: 658–665. Sawyer, Sarah. 1998. Privileged Access to the World. Australasian Journal of Philosophy 76: 523–533. Scales, Ronald. 1977. A Russellian Approach to Truth. Noûs 11: 169–174. Schiffer, Stephen. 1977. Naming and Knowing. Midwest Studies in Philosophy 2: 28–41. ———. 1987. The ‘Fido’-Fido Theory of Belief. In Philosophical Perspectives 1: Metaphysics, J. Tomberlin (ed.), pp. 457–480. Atascadero: Ridgeview. ———. 1989. Fodor’s Character. In Information, Semantics, and Epistemology, E. Villanueva (ed.). Oxford: Blackwell. ———. 1992a. Boghossian on Externalism and inference. Philosophical Issues 2: 29–38. ———. 1992b. Belief Ascription. Journal of Philosophy 89: 499–521. Schirn, Matthias. 1999. On Frege’s Theory of Non-Extensional Sentences. Theoria 14(34): 131–156. Schirn, Matthias (ed.). 1996. Frege: Importance and Legacy. Berlin: de Gruyter. Schofield, Robert. 1970. Mechanism and Materialism. Princeton: Princeton University Press. Seager, W. E. 1992a. Thought and Syntax. Philosophy of Science Association 1: 481–491. ———. 1992b. Externalism and Token Identity. Philosophical Quarterly 42: 439–448. Searle, John. 1983. Intentionality. Cambridge: Cambridge University Press. ———. 1992. The Rediscovery of the Mind. Cambridge, MA: MIT Press. Segal, Gabriel. 1989a. The Return of the Individual. Mind 98: 39–57. ———. 1989b. Seeing What Is Not There. Philosophical Review 97: 189–214. ———. 1991. Defence of a Reasonable Individualism. Mind 100: 485–494. ———. 2000. A Slim Book about Narrow Content. Cambridge, MA: MIT Press. Segal, Gabriel and Elliott Sober. 1991. The Causal Efficacy of Content. Philosophical Studies 63: 1–30. Seymour, Daniel. 1993. Some of the Difference in the World: Crane on Intentional Causation. Philosophical Quarterly 43(170): 83–89. Seymour, Michel. 1992. Les Expériences de Burge et Les Contenus de Pensée. Dialectica 46(1): 21–39. Shapiro, L. A. 1993. Content, Kinds, and Individualism in Marr’s Theory of Vision. Philosophical Review 102: 489–513. ———. 1997a. A Clearer Vision. Philosophy of Science 64: 131–153. ———. 1997b. Junk Representations. British Journal for the Philosophy of Science 48: 345–361. Shoemaker, Sydney. 1982. The Inverted Spectrum. Journal of Philosophy 7(79): 357–381. ———. 1988. On Knowing One’s Own Mind. Philosophical Perspectives 2: 183–209. ———. 1990. First Person Access. Philosophical Perspectives 4: 187–214. ———. 1991. Rationality and Self-Consciousness. In The Opened Curtain: A U.S.–Soviet Philosophy Summit, Keith Lehrer and Ernest Sosa (eds.). Boulder: Westview Press. ———. 1994a. Self-Knowledge and Inner Sense; Lecture III: The Phenomenal Character of Experience. Philosophy and Phenomenological Research 2(54): 291–314. ———. 1994b. Phenomenal Character. Noûs 28: 21–38. Shuster, John A. 1980. Descartes’ Mathesis Universalis 1619–28. In Descartes: Philosophy, Mathematics, and Physics, Gaukroger (ed.). Sussex: The Harvester Press. Silverberg, Arnold. 1995. Narrow Content: A Defense. Southern Journal of Philosophy 33: 109–127.
Secondary Sources on Tyler Burge
491
———. 1998. Semantic Externalism: A Response to Chomsky. Protosociology 11: 216–244. Silvers, S. (ed.). 1989. ReRepresentation. Dordrecht: Kluwer. Simmons, Keith. 1993. Universality and the Liar: An Essay on Truth and the Diagonal Argument. Cambridge: Cambridge University Press. Sosa, Ernest. 1991. Between Internalism and Externalism. In Philosophical Issues 1: Consciousness, Enrique Villanueva (ed.). Atascadero: Ridgeview. ———. 1993. Abilities, Concepts, and Externalism. In Mental Causation, John Heil (ed.). New York: Clarendon Press. Spelke, Elizabeth. 1990. Origins of Visual Knowledge. In An Invitation to Cognitive Science, II, D. N. Osherson, S. M. Kosslyn, and J. M. Hollerbach (eds.). Cambridge, MA: MIT Press. ———. 1995. Object Perception. In Readings in Philosophy and Cognitive Science, Goldman (ed.). Cambridge, MA: MIT Press/A Bradford Book. Stalnaker, Robert. 1981. Indexical Belief. Synthese 57: 129–152. ———. 1989. On What’s in the Head. Philosophical Perspectives 3: Philosophy of Mind and Action Theory, J. E. Tomberlin (ed.), pp. 287–316. Atascadero: Ridgeview. ———. 1990. Narrow Content in Propositional Attitudes: The Role of Content in Logic, Language, and Mind. Stanford: CSLI Publications. ———. 1993. Twin Earth Revisited. Proceedings of the Aristotelian Society 63: 297–311. ———. 1996. On a Defense of the Hegemony of Representation. In Philosophical Issues 7: Perception, E. Villanueva (ed.). Atascadero: Ridgeview. Sterelny, Kim. 1990a. The Representational Theory of Mind: An Introduction. Oxford: Blackwell. ———. 1990b. Animals and Individualism. In Information, Language, and Cognition, Philip Hanson (ed.). Vancouver: University of British Columbia Press. Stewart, Ian. 1996. From Here to Infinity. Oxford: Oxford University Press. Stich, Stephen. 1983. From Folk Psychology to Cognitive Science. Cambridge, MA: MIT Press. ———. 1991. Narrow Content Meets Fat Syntax. In Meaning in Mind: Fodor and His Critics, B. Loewer and G. Rey (eds.). Oxford: Blackwell. ———. 1996. Deconstructing the Mind. New York: Oxford University Press. Stoneham, Tom. 1999. Boghossian on Empty Natural Kind Concepts. Proceedings of the Aristotelian Society 99: 119–122. Strawson, Galen. 1994. Mental Reality. Cambridge, MA: MIT Press. Strawson, P. F. 1950. On Referring. Mind 59: 320–344. ———. 1952. Introduction to Logical Theory. London: Methuen. ———. 1959. Individuals. London: Methuen. Stueber, Karsten. 1997. Psychologische Erklärungen im Spannungsfeld des Interpretationismus und Reduktionismus. Philosophische Rundschau 44(4): 304–328. Sturgeon, Scott. 1994. Good Reasoning and Cognitive Architecture. Mind and Language 9: 88–101. Switankowsky, Irene. 1999. Dretske on Naturalizing Experience. Dialogue 38: 561–566. Taylor, Kenneth. 1989a. Narrow Content Functionalism and the Mind-Body Problem. Noûs 23: 355–372. ———. 1989b. Supervenience and Levels of Meaning. Southern Journal of Philosophy 27: 443–458. Thomas, J. 1996. Analogies and the Mind of the Replica: Sunburn, the Little Green Bug, and the Fake Plant. Philosophical Quarterly 46: 364–371. Tomberlin, James E. (ed.). 1989. Philosophical Perspectives, 3: Philosophy of Mind and Action Theory. Atascadero: Ridgeview. ———. 1997. Philosophical Perspectives 12: Language, Mind, and Ontology. Oxford: Blackwell.
492
Secondary Sources on Tyler Burge
Tuomela, R. 1989. Methodological Solipsism and Explanation in Psychology. Philosophy of Science 56: 23–47. Tye, Michael. 1994. Qualia, Content, and the Inverted Spectrum. Noûs 28: 159–183. ———. 1995a. Blindsight, Orgasm, and Representational Overlap. Behavioral and Brain Sciences 18: 268–269. ———. 1995b. Ten Problems of Consciousness. Cambridge, MA: MIT Press. ———. 1998. Externalism and Memory. Aristotelian Society Supplement 72: 77–94. Valdivia, Lourdes. 1987. Nombres Y Predicados: Algunas Consideraciones. Critica 19: 83–95. ———. 1995. Can Peter Be Rational? In Philosophical Issues 6: Content, Enrique Villanueva (ed.). Atascadero: Ridgeview. van Fraassen, Bas C. 1995. Belief and the Problem of Ulysses and the Sirens. Philosophical Studies 77: 7–37. van Gulick, R. 1989. Metaphysical Arguments for Internalism and Why They Don’t Work. In ReRepresentation, S. Silvers (ed.). Dordrecht: Kluwer. Vaughan, R. 1989. Searle’s Narrow Content. Ratio 2: 185–190. Vergara, Julia. 1995. La Distincion “Narrow Content-Wide Content.” Revista Latinoamericana de Filosofia 21(1): 131–144. Villanueva, Enrique (ed.). 1991. Philosophical Issues 1: Consciousness. Atascadero: Ridgeview. ———. 1992. Philosophical Issues 2: Rationality in Epistemology. Atascadero: Ridgeview. ———. 1993. Philosophical Issues 4: Naturalism and Normativity. Atascadero: Ridgeview. ———. 1995. Philosophical Issues 6: Content. Atascadero: Ridgeview. ———. 1996. Philosophical Issues 7: Perception. Atascadero: Ridgeview. ———. 1998. Philosophical Issues 9: Concepts. Atascadero: Ridgeview. Voltolini, Alberto. Is Narrow Content the Same as Content of Mental State Types Opaquely Taxonomized? In Analyomen 2, vol. III: Philosophy of Mind, Practical Philosophy, Miscellanea, Georg Meggle (ed.). Berlin: de Gruyter. Walker, Valerie. 1990. In Defense of a Different Taxonomy: A Reply to Owens. Philosophical Review 99: 425–431. Wallace, J. and H. E. Mason. 1990. On Some Thought Experiments about Mind and Meaning. In Propositional Attitudes, C. Anderson and J. Owens (eds.). Stanford: CSLI Publications. Walsh, D. M. 1998. Wide Content Individualism. Mind 107: 625–651. Warfield, Ted A. 1992. Privileged Self-Knowledge and Externalism Are Compatible. Analysis 52: 232–237. ———. 1995. Knowing the World and Knowing Our Minds. Philosophy and Phenomenological Research 55: 525–545. ———. 1997. Externalism, Privileged Self-Knowledge, and the Irrelevance of Slow Switching. Analysis 57: 282–284. ———. 1998. A Priori Knowledge of the World: Knowing the World by Knowing Our Minds. Philosophical Studies 92: 127–147. Warren, Dona. 1999. Externalism and Causality: Simulation and the Prospects for a Reconciliation. Mind and Language 14(1): 154–176. Wegner, Daniel M. and James W. Pennebaker (eds.). 1993. Handbook of Mental Control. Englewood Cliffs, NJ: Prentice-Hall. Weiner, Joan. 1995a. Realism “bei” Frege: Reply to Burge. Synthese 102(3): 363–382. ———. 1995b. Burge’s Literal Interpretation of Frege. Mind 104(415): 585–597. Weinstein, S. 1974. Truth and Demonstratives. Noûs 8: 179–184. White, Stephen. 1982. Partial Character and the Language of Thought. Pacific Philosophical Quarterly 63: 347–365.
Secondary Sources on Tyler Burge
493
———. 1992. Narrow Content and Narrow Interpretation. In his The Unity of the Self. Cambridge, MA: MIT Press. ———. 1995. Color and Notional Content. Philosophical Topics 22: 471–504. Williams, Michael. 1990a. Social Norms and Narrow Content. Midwest Studies in Philosophy 15: 425–462. ———. 1990b. Externalism and the Philosophy of Mind. Philosophical Quarterly 40: 352–380. ———. 1998. Bilgrami on Belief and Meaning: Is Fregean Externalism Possible? Philosophy and Phenomenological Research 58: 607–611. Williamson, Timothy. 1995. Is Knowing a State of Mind? Mind 104: 533–565. ———. 1998. The Broadness of the Mental: Some Logical Issues. Philosophical Perspectives 12: 389–410. ———. 1999. Skepticism, Semantic Externalism, and Keith’s Mom. Southern Journal of Philosophy 38: 149–158. Wilson, Robert A. 1992. Individualism, Causal Powers and Explanation. Philosophical Studies 68(2): 103– 139. ———. 1993. Against A Priori Arguments for Individualism. Pacific Philosophical Quarterly 74: 60–79. ———. 1994a. Causal Depth, Theoretical Appropriateness, and Individualism in Psychology. Philosophy of Science 61: 55–75. ———. 1994b. Wide Computationalism. Mind 103: 351–372. ———. 1995. Cartesian Psychology and Physical Minds: Individualism and the Sciences of the Mind. Cambridge: Cambridge University Press. Woodfield, Andrew. 1982a. Thought and the Social Community. Inquiry 25: 435–450. ———. 1982b. Thought and Object: Essays on Intentionality. Oxford: Oxford University Press. ———. 1986. Two Categories of Content. Mind and Language 1: 319–354. ———. 1997. Social Externalism and Conceptual Diversity. In Thought and Language, John Preston (ed.). Cambridge: Cambridge University Press. Woodfield, Andrew (ed.). 1982. Thought and Content. New York: Oxford University Press. Wyller, Truls. 1994. First-Person Authority and Singular Thoughts. Zeitschrift fur philosophische Forschung 48(4): 585–594. Yablo, S. 1997. Wide Causation. Philosophical Perspectives 11: 251–281. Yagisawa, Takashi. 1984. The Pseudo-Mates Argument. Philosophical Review 93: 407–418. Yalowitz, Steven. 1999. Davidson’s Social Externalism. Philosophia: Philosophical Quarterly of Israel 27: 99–136. Yolton, John. 1983. Thinking Matter. Minneapolis: University of Minnesota Press. ———. 1984. Perceptual Acquaintance. Minneapolis: University of Minnesota Press. Zemach, E. M. 1976. Putnam’s Theory on the Reference of Substance Terms. Journal of Philosophy 73: 116–127. Zimmerman, Dean. 1995. Theories of Masses and Problems of Constitution. Philosophical Review 104(1): 53–110.
Index
Ability-general representations, 305–307, 329n115–16 Abstraction, 310 Agency, cognitive/intellectual, 201, 206–208, 298, 301, 441. See also under Self-knowledge, authoritative Agent(s), 207, 222, 223, 298, 300, 301 Almog, J., 84–85, 93 Alternation(s), 118, 119, 130, 140–141 Anti-essentialism, 46 Anti-individualism, 350–351. See also Externalism(s); Nonindividualism; specific topics applies to conceptual elements in propositional attitudes, 348–351 argument for perceptual, 302–303 basis/grounds of, 341–342 Burge’s, xxix, 29, 32–33, 39–40, 52, 53 communities, individuation, and social, 55, 326–327 Davidson and forms of, 347–359 (see also Davidson) Davidson’s denial of, 49 (see also Davidson) defined, 15, 29, 453–454 epistemological problems associated with, 95 and failure of supervenience, 371–372 implications, 15–27 indexicality, character, and, 77–97 scepticism and, xv, 15–27, 324–325, 338–345 social, 356, 359 (see also Davidson; Externalism(s)) Antirealism, 46 Appearance concepts, 169 Appearance properties, 184 Application(s), 306–309 Apriori, 204, 324 logical axioms can be known, 144–145, 389 Apriori (inferential) connections, 315, 323 Apriori inferences, 425 Apriori knowledge, 120 as knowledge about thinker’s mind, 151n14 Apriori/necessity distinction, 81–82, 87 Apriori truths, 314 Apriori warrant, 315, 331–332n33, 460–461 Aristotle, 293 Arithmetic, 125, 147, 295, 385 “Arthritis” thought experiment, xiii, 39, 65, 72, 77–78, 91, 94, 250, 357–359, 364 compared with water thought experiments, 363 Artifacts, 319 Astronomy, 321, 322 Attention, 406, 407 Attitude-attribution/belief-attribution, 20–25, 211–212, 341–342, 345, 359n. See also Attribution
Attribution, 20, 21, 30, 36. See also Self-attributions third-person, 212 Authority, 2, 60. See also Experts; Self-attributions, authoritative; Self-knowledge, authoritative first-person, 32, 357 Autonomous psychology, xiii. See also Solipsism, methodological Autonomous representations, 307, 308 Awareness, 406, 407. See also Consciousness Axioms, 146–148, 150n1, 389. See also under Apriori can be known apriori, 144–145, 389 are necessary, 145 rationality of accepting logical, 390–392 Bach-y-Rita, 178–179 Bare concepts, 5, 6, 304, 308, 314, 323 Basic truths. See Self-evident/basic truths “Belief De Re” (Burge), xiv theses relevant to development of antiindividualism in, 347–350, 361n12 Belief formation, 210, 419 Belief relation, 211 Beliefs, 3, 23, 345, 451–452. See also Attitudeattribution/belief-attribution accuracy, 17 (see also Scepticism) as active vs. passive, 206–207 anti-individualism and the nature of, 342 as default position, 419 externally individuated, 153, 155, 156 hot and cold, 210 justification for (see Justification(s)) are mostly true, 21–23, 336, 343–344 perceptual, 337 spontaneous, 208–213 suspension of, 419 as veridical in nature, 21–22, 24 Black, Joseph, 261, 276 Block, Ned, 184, 185, 194 Bodies, 12 Bodily sensations, 175–177 “Body of doctrine,” 261, 262, 276 Boghossian, Paul, 78, 205–206, 213, 214–221, 226n15, 432n6 “platitude,” 215–217 Brain(s) in vat(s), 13n3, 18–19, 85, 86, 246, 250, 443–446. See also Twin-in-a-vat Brown, Deborah, 332n43 Burge, Tyler. See also specific topics writings “Belief De Re,” xiv, 347–350, 361n12 “Cartesian Error and the Objectivity of Perception,” xvi, 1, 291–303 “Frege on Sense and Linguistic Meaning,” 117
496
Burge, Tyler (cont.) “Individualism and Self-Knowledge,” xv, xxvi, 421 “Individualism and the Mental,” xi, xii, 33, 77, 348, 349, 363, 367, 368, 436, 438 “Intellectual Norms and the Foundation of Mind,” xiv, 34 “Mind-Body Causation and Explanation,” 59 “Other Bodies,” xiv, 33, 61, 77, 348, 366 “Russell’s Problem and Intentional Identity,” 329n15 Burge–Heil reconciliation story, 203, 427–428 doubts and discontents about, 203–206, 426–427 Loar’s objection, 221–224, 426 and smoldering self-knowledge, 213–215, 226n14 Buridan, Jean, 2–3 Calculus, 121–124, 138 “Cartesian Error and the Objectivity of Perception” (Burge), xvi, 1 mistakes in, 291–303 Categorization, 385 Causal externalism, 3, 5 Causal history, 45, 47, 51, 52 Causation (and causal relations), xv, 292, 347, 353, 424–425 higher-order rational, 207 mental, 158, 161–162, 397, 398 (see also Explanatory practice) Chair case, 384–385, 388–389 Character, xxi, 96, 97 anti-individualism, indexicality, and, 81–97 metaphysics of, 95 Character/content distinction, 82, 86, 88, 91, 96–97 Chomsky, Noam, 53, 54 “Innateness Hypothesis,” 279 Churchland, Paul, 276 Clear and distinct (ideas), 145–146, 389–390 Cogito, 9, 296, 429 Cogito cases impure, 417, 418, 421, 426, 428 pure, 418, 421, 426, 429 Cognitive science, 260–261 Cognitive value, 96 Coherence, 336–337 Coke machine example, 400–402 Color categories, 190–191 Color concepts, 195, 196–197n7, 405, 410–411, 414 Colors nature of, 197n13 seeing, 166–175, 186, 193, 196n3, 197n7, 238 (see also Mental paint)
Index
seeing red for the first time, 180–183 shifted spectra, 188–192 Color words, 190–191 Commitment, and spontaneous beliefs, 208–213 Common sense, 259 Communities. See under Anti-individualism Competence, 54, 381 Complement causes, 106–107 Concepts, 3–4 Burge’s vs. Davidson’s notion of, 47 conceptions and reflective understanding, 383–395 Descartes on, 5, 6, 11 individual, 329n15 mastery of, 38, 49 (see also Incomplete mastery) narrow, 236 nature of, 5 Ockham on, 5 perceptual, 3–4, 247, 250 role in reference and knowledge, 304–315 wide, 235 Conceptual-role theories, 132–136, 240 Consciousness, 298, 406 two perspectives on, 165 Constructivism, 119 Content-involving conceptions and explanations, 136, 137, 153 Content vs. character. See Character/content distinction Context-dependent representations, 306–309, 329n16, 330n18, 371, 380 Context sensitivity, 83, 88 Coordinate transformations, 108–109, 379–380 Cracks, 10–11 Crimmins, Mark, 124, 125 Cross-language question, 91, 92 Cross-language thesis, 93 Davidson, Donald, 106, 336–337, 339–341, 343, 372. See also under Externalism(s) on beliefs, 17, 18, 20–22 “A Coherence Theory of Truth and Knowledge,” 22 essential difference between Burge and, 54–55 and forms of anti-individualism, 347, 353–354, 356–359 learning argument, 43, 50–52 “The Myth of the Subjective,” 36 Debatable cases, 236 De dicto attitudes, 52, 347–350, 360n4 De dicto thoughts, 30, 48, 436 Defeating conditions, 210 Definition, 124–125 Deflationary notion of intentionality, 234, 235, 237 Deliberation, 208
Index
Demon(ic) worlds, 9 Demonstrative reference and truth, 113 normal forms for, 103–107 “Demonstratives” (Kaplan), 96–97, 98n3 Demonstratives/demonstrative concepts, 98n5, 239–241, 245, 247, 349, 350, 377–380, 423, 440, 446 Dennett, Daniel C., 185 Denotational theories of interpretation, 270–271 “Depth” problem, 37 De re beliefs/attitudes, 34, 52, 53, 347, 348–350, 359–360n1–4 De re thoughts, 30, 52 Descartes, René, 63, 207 and anti-individualism, 291–327 on belief formation, 419 blurring distinction between mathematical and physical kinds, 317 Burge on, 1–5, 291, 292–304, 315–320, 323, 327 Burge’s views contrasted with, 315–325 Cartesian metaphysics, 12 Cartesian sceptical concerns, 79 causal principle, 292 as externalist, 7, 8, 11 on ideas vs. judgments, 6 as individualist vs. anti-individualist, 291–293, 327 on love, 12–13 vs. modern externalists, 8 nominalism, 301 ontology, 317, 319 problem of two sons, 9–11 on reflection, 462 account of representation, 5–9 on will, 206, 207 Descartes’s demon, 9 Descriptional theories of names/descriptional model, 94 Diagonal propositions and diagonal strategy, 112 Diaphanousness, 244 of introspection, 405, 406 of perception, 254 Directedness, 241, 243, 247, 249, 251, 255, 440 and reference, xiii, 253–254 singular, 239, 240, 439, 446, 447 Direction of fit, 201–202, 209, 211 Direct self-knowledge. See Self-knowledge, authoritative Disjunctivists, 174, 175 Dispositional intentional properties, 252. See also Intentional properties Distinctions, conceptual, 299–301 Distortedness, 244
497
Donnellan, Keith, 461 Dualism/dualist, 261, 293, 296–297, 301, 302 Eliminativism, 183, 451, 468 Ellipsis, 102 Empirically identifiable objects, 459 Empirical theory and investigation, 461, 462 Empiricism, xiv, 143 Entitlement, 222–224, 427 Epiphenomenalism, 153, 154, 163, 397–398, 400, 401 (status of) threat from, 397, 398, 402 Epistemic warrant, 342, 343 Epistemology of belief, 83, 395 Error(s), 187, 323, 417. See also Referential error; Scepticism Burge on, 16, 17 Descartes on, 6–8 immunity from, 16, 17, 21, 23–24, 335 supposes background of veridicality, 22, 23, 338, 351 Essentialism, 46, 49, 435 Evans, Gareth, 108, 212, 330n23, 331n26 Evolution, 411, 445 Experts, 60, 68, 364. See also Authority Explanatory practice, xvi, 154, 160–162, 400, 402–403. See also Causation Explanatory role of content, 153 Explication (conceptual), 31, 47, 55, 121, 127, 139–140, 149, 350, 357, 377, 390 Explicative judgments, 384 Extensions, 79–81, 351, 352, 367 Externalism(s), 29, 83–84, 220, 435, 438–439. See also Anti-individualism agreements between Burge and Davidson, 29–32, 356 Burge’s theses, 4, 32–34 Cartesian, 7, 8, 11 causal, 3, 5 Davidson’s explanations of, 360n9 Davidson’s perceptual, 40–45 Davidson’s rejection of Burge’s arguments, 357 Davidson’s theses, 34–36 and explanatory role of content, 153 Kripke and contemporary, 3 perceptual, 29–31, 37 physical, 29, 31, 34, 36 a surfeit of, 45–52 semantic, 153, 155, 157 social, 29, 31, 33, 34 Davidson and Burge on, 47, 52–55, 356–357 from physical to, 36–39 Externalist memory, 183–186. See also under Memory
498
Externalist representationism, 168–170, 175, 182, 187, 188, 238, 245 Externally individuated thoughts, 153, 155, 156 Faculty of language (FL), 262–267, 275–276, 278 Fallible vs. infallible, 282 Falsity conditions, 134 Feedback, perceptual, 210 Finiteness, 150n7, 293 Fodor, Jerry, 161 Folk semantics, 266 Folk theories, 280 Forbes, Graeme, 85–86 Formally general representations, 306–310, 329n15–16 Foundationalism, 394 Fragmentation strategy, 218 Frege, Gottlob, xii, 122, 124, 147–149, 150n3, 390–392 Basic Laws of Arithmetic, 147 Fregean senses, 380, 381 “Frege on Sense and Linguistic Meaning” (Burge), 117 Frege’s constraint, 229 Functionalism, 165–167, 176, 193 representationism as form of, 179–180 Functional organization, 166 “Generality Constraint” (Evans), 212 Geometry, 294–295, 317, 318 God, 296, 306 Descartes on, 6–8, 292, 293, 296, 328n6 will of, 8 Hallucination, 174, 175, 239, 241, 242, 245–246, 248, 445 Hampshire, Stuart, 208, 212–213 Hardin, C. L., 189–190 Harman, Gilbert, 171–174, 177–179, 196–197n7, 242–243, 245 Heil, John. See Burge–Heil reconciliation story Hobbes, Thomas, 274, 275, 466 Holism, 452 “How”–“what” distinction, 240–241 Human body. See Bodies Human nature, science of, 259–261 Hume, David, 207, 261, 275 Idea(s) caused formally vs. objectively, 7, 9 Descartes’s account of, 5–7, 11 fictitious, vs. ideas of true/immutable natures, 7–8 object of an, 7
Index
sources of error in, 7 two ways of regarding, 5 Idealism, 8, 349 Identity theories, 32 Idiolects, 54, 55 I-language and I-linguistics, 263–268, 277, 279, 280, 283, 284, 453 Illusion, 238, 465 Images, 5–6. See also Representation(s) Implicit conceptions, xxiii, 117–125, 148, 149, 383–387, 390. See also Conceptual-role theories; Justification(s) defined, 383 deflationary readings rejected, 125–132 explanation by, 136–142 making them explicit, 121–124 Incomplete mastery/understanding, 65–68, 71, 350, 367, 368, 383 Indexical/character model, 94 Indexical expressions, 82 as determined by contextual factors, 81 explicit, 89–90 Indexical reference, 113 Indexicals, 79–82, 97, 105, 106. See also under Anti-individualism conflating kind concepts with, 448 implicit, 90–97 Kaplan’s theory of, 80–88, 96–97, 98n3–5, 99n12, 106 Indexical strategy, 63, 64, 79, 371–375. See also Anti-individualism, indexicality Individualism, 38, 40, 453 Burge on, 1–2 Descartes and, 291–293, 327 materialism and, 156 thought experiments and, 1–2, 59–60, 69 “Individualism and Self-Knowledge” (Burge), xv, xxvi, 421 “Individualism and the Mental” (Burge), xi, xii, 33, 77, 348, 349, 363, 367, 368, 436, 438 Individualistic theory of names, 94 Individuated states, nonintentionally, 88 Individuated thoughts, externally, 153, 155, 156 Individuation, 33, 35, 38–40, 43, 47–48, 51, 158, 160–162, 252 of attitudes, 351, 353 Descartes on, 301 of mental content, 229 of mental/intentional states, 350–352, 413 of perceptual states, 352 of thoughts, 292 Individuation conditions, 292 Inference/inferential rules, 118, 119, 141, 146, 147, 391–394
Index
Inferential abilities, 394 Inferential dispositions, 127 Inferential principles, 139 Inferential relations, 353, 392, 394 Infinitesimals, 121 Innateness Hypothesis (Chomsky), 279 “Intellectual Norms and the Foundation of Mind” (Burge), xiv, 34 Intension(s), 79–81 Intention, 209 Intentional aspects of visual qualitative representations, 443 Intentional content, 173, 202, 211, 245, 351, 352, 425, 435, 439, 446, 448 of memory, 213–214 narrow, 165–166 qualia and, 405–415 Intentional directedness. See Directedness Intentionality, 230–231, 347, 445 deflationary notion of, 234, 235, 237 externalism about, 231–233 how and what of, 240–242 internalism about, 231 “outer,” 441–442 perceptual, 448 phenomenal(ogical), 238–241, 246, 254, 441, 442 rejection of internalist theory of, 235–238 singular, 232, 441, 442 Intentional objects, 5, 178, 240–246, 251, 254 Intentional properties (of experience), 178–179, 229, 248, 252, 435 Intentional states, 2, 30, 51, 230, 397, 454. See also under Individuation Intentional success, 439 Intentional type, 212 Interface levels, interpretation of, 267–268 Internalism, 165–168, 453–456, 467. See also under Intentionality defined, 454, 468n2 Internalist-externalist opposition, xxviii Internalist representationism, 166–168, 175 Interpretation, 36, 37, 44–45, 54 anti-individualism and, 341–343 denotational theories of, 270–271 radical, 43 Interpretationism, 34–35, 56n3, 353 Introspection, 78, 173, 174, 178, 252 diaphanousness of, 405, 406 Intuition, 238, 254, 265, 451 intellectual, 328–329n14 Intuitive judgment, 265–266, 459, 460, 462 Inverted Earth (experiment), 175, 184–188, 196–197n7, 245, 246, 409–411 Swampman visiting, 197n9, 411
499
Inverted spectra and inverted worlds, 245, 409, 410 Inverted Spectrum, 175, 197n10, 246 Isolated brains, 245–257 I-sound and I-meaning, 264–266, 272 Jacobs, G., 189 Johnson-Laird, P. N., 138, 139 Judgments, 6, 205, 384, 418, 428, 430 content-involving, 136 explicative, 384 memory-based, 205 reflexive, 203–204, 418, 424 self-regarding, 205 Justification(s), 124, 208, 209, 222, 223, 338, 391, 392 climbing the ladder of, 224 vs. entitlement, 224 (see also Entitlement) rationality, understanding, and, 209 theory of, 146–150 Kant, Immanuel, 301, 302, 315, 337, 339n19–20, 387 “Refutation of Idealism,” 349 Kaplan, David, 104, 105, 107 theory of indexicals, 80–88, 96–97, 98n3–5, 99n12, 106 Kind-directedness, 249 Kind-independent intentionality, 247–251 Kinds/kind terms, 4, 74n3, 319, 443, 447–448, 456 eliminativism about, 451 natural, 11, 62, 64, 67, 69–71, 74n8, 80, 90–91, 318, 365, 366, 458 (see also Natural-kind environments) perceptual, 10, 11, 335 physical, 317 Kripke, Saul, 3, 94, 282, 461 Lange, Friedrich, 261 Language(s), 86, 102–103, 381. See also Demonstrative reference; I-language natural, 211 phases in the use of, 101–102, 106 Language-world relations, 466. See also Reference Learning argument, 43, 50–52 Leibniz, Gottfried Wilhelm, 121–124, 131, 138, 143–145, 150n2, 151n14, 385, 389, 390, 393 Lewis, David, 181 Lexical items, 264, 268–275, 284, 467 Limit-concept (calculus), 121–124, 138, 386 Lingua mentis, 277, 278, 283 Linguistic ascription, 438 Linguistic community/linguistic rules, 87, 358, 368
500
Linguistic labor, division of, 67–68 Linguistics. See I-language and I-linguistics; specific topics Loar, Brian, 225n4–5, 425–427 objection to Burge–Heil reconciliation story, 203–204, 221–224 Locke, John, 68, 145, 262, 274, 275 Logic, 460 Logical axioms. See Apriori; Axioms Love, Descartes on, 12–13 Lycan, William G., 175, 176, 178–181, 186, 194, 243–245 Marr, David, 463–465, 469n14 Materialism, 78, 156, 261, 262, 397, 400, 402–403, 435. See also Substances Mathematics, 8, 317–318, 349, 386, 460. See also Arithmetic; Calculus; Geometry Maximizing intelligibility, 131–132 Meaning, 451, 455 conceptual-role theories of, 132–136 notions of, 377–379 Memory, 214 externalism and, 183–188, 409 intentional content of, 213–214 phenomenal, 187–188 preservative, 214, 422, 431 Memory-based judgments, 205 Mental causation. See Causation, mental; Mentalistic explanations Mental content, 229. See also Individuation conceptual roles and, 233–235 externalist accounts of, 201, 202 (see also Antiindividualism) how it gets conveyed, 230, 381 vs. mental relation, 201 Mentalistic explanations, 153–163, 399–402 Mental Models (Johnson-Laird), 138–139 Mental oil, 175, 176, 180, 195n2 Mental paint, 172–173, 175, 178, 180, 184, 186. See also Colors Mental properties, 173–174 Mental relations, 210 types of, 201–202 Mental states, 35, 59, 202, 371. See also under Individuation responsibility for success of, 202 Swampman and, 354, 356 Metalinguistic strategy, 358, 368–369 Metaphysics, 160. See also Materialism anti-individualistic, 78, 79 of belief, 83 Cartesian, 12 Metarepresentations, 331n29
Index
Methodology, xv–xvi, 459–460 “Mind–Body Causation and Explanation” (Burge), 59 Mind–body interaction, 159–161, 163, 397, 403 Mind-to-world direction of fit, 201–202, 209 Misperception, 465. See also Perception “MIT mentalism,” 277–279, 281 Monetary-machine interaction, 160–163 Moore, George Edward, 338–339, 344 Moore’s paradox, 344 “Myth of the Subjective, The” (Davidson), 36 Names, 3, 437. See also Proper names individualistic vs. descriptional theories of, 94 Naming and Necessity (Kripke), 3 Natural-kind environments, 354–355. See also Kinds/kind terms, natural Natural-language sentences, 211 Nature(s), 8 Necessitation, principle of, 81–82 Necessity and validity, 82 Negation, 133–134, 140 Negative sentences, 3 Neitz, J., 189, 197n12 Neitz, M., 189, 197n12 Neo-Fregeanism, 230 Newton, Isaac, 121–124, 131, 138, 150n2, 261, 385 Nominal essence, 68 Nominalism, 301 Nonindividualism, 153, 154, 156, 158. See also Anti-individualism Normality, 335, 336, 377, 378 Object-dependence and object-independence, 253–254, 446 Object of thought, 83, 96 Oblique ascription, 436 Oblique content, 437 Oblique occurrences, 436–438 Oblique positions, 436 Ockham, William, 5 Orgasm, representational content of, 175–177, 179, 407 “Other Bodies” (Burge), xiv, 33, 61, 77, 348, 366 Overdetermination, 394 Paderewski example, 437 Pain, 179 Paint. See also Colors mental, 172–173, 175, 178, 180, 184, 186 phenomenal, 242–244 that points, 251 Paradigms, nature of, 68 “Particular nature,” 297–298
Index
Perception, 4, 337–338, 465. See also specific topics reliability, 15–16 (see also Scepticism) veridical, 463 Perception-attribution, 21 Perceptual anti-individualism. See under Antiindividualism Perceptual concepts, 3–4, 247, 250 Perceptual feedback, 210 Perceptual kinds, 10, 11, 335. See also Kinds Perceptual representations, 2, 10, 11, 312, 313, 330n18, 333n50, 335 Perceptual states, 4, 5, 10 Perceptual systems, 136–137, 310–313, 336–337 Performative element, 219, 220, 222, 224, 417, 418 Perry, John, 108 Personal systematic concepts, 251–253 Perspective(s), 304, 377, 381 conveyance of, 109, 381 preservation vs. loss of, 101–102, 377, 380 puzzles of, 109–112 tracking, 377–382 and truth, 112–113 Phenomenal character, 179, 195n3 of color experience, 186 concepts of (see Phenomenal concepts) constancies vs. differences in, 197n10, 410–414 externalism and, 168–170, 187, 188 externalist memory and, 183–186 internalism and, 165–168 vs. representational content, 165, 179, 186, 187, 190, 192, 194–195 supervenience and, 170–171 Phenomenal concepts, 179, 182, 183 Phenomenal directedness. See Directedness Phenomenality, reference and, 435–448 Phenomenal paint, 242–244 Phenomenal properties, 173, 179, 195–196n3, 238, 243 Phenomenal states, 234 “raw,” 202 Phenomenism, 165, 175, 180, 181, 192–193 and antiphenomenists, 180, 182, 407, 408 closet, 176 physicalist and dualist forms of, 166–168, 182 Phenomenon of New Principles, 133–135 Phlogiston, 316 Phonetic interface, 268 Phonetics, 267. See also Lexical items Phonetic Value (PV), 270–271 Phosphene-experiences, 177 Physicalism, 166, 168, 183 Physical kinds, 317 Physical sciences, 261, 355
501
Picasso pictures, 244–245 Pictures, the how and what of, 244–245 Plato, 66 Possibly true and the possible, the, 2 Predicative concepts, 241 Principal attributes (of mind), 294, 297–299 Projective-deflationist theory of reference, 237 Proof(s), 124, 146, 149 Proper names, 3, 4, 97, 331n32 description theory of, 74n10 Property-concepts, 247 Property-reference, 233, 247, 443–444, 447 Propositional attitudes, 30, 33, 34, 41, 70–71, 342, 355, 420, 436 anti-individualism applies to conceptual elements in, 348–351 nonmodular, 419 Proprioception, 417 Psychological abilities, 305–307, 452 Psychological generalizations (PGs), 260–261, 263 Psychological states, 83, 365, 453 causal roles, xv Putnam, Hilary, 13n3, 18, 61–63, 85–86, 314, 457, 4684n critique of “MIT mentalism,” 277–279 “The Meaning of ‘Meaning’,” 61, 62, 67, 77, 363 semantics and semantic rules of, 60, 62–65, 68–70, 367 thought experiments of, 59, 448 vs. Burge’s thought experiments, xiii–xiv, xx, 60–61 indexical element in, 60–62 Twin Earth, 59, 61–65, 67, 68, 71, 77, 79, 84–85, 97, 183, 195, 363–365, 369 (see also Twin Earth experiments) Qualia, 165, 168, 182, 234, 238. See also Representationist-qualiphile dispute acceptance vs. rejection of, 193, 243, 244, 246 color, 245, 246, 414 defined, 195n1 directed, 247 and intentional content, 405–415 phenomenal, 411, 412, 414, 443 postulation and awareness of, 406 scientific nature of, 195n1 visual, 243, 244, 246, 440, 442, 443 Quasi-representationism, 179, 180 Quine, Willard Van Orman, 17, 336 Rationalism, 389, 395, 462 supported, 142–146 Rationality, 149, 342–343, 389, 392
502
Reality degree of, 5 formal and objective, 5–7, 9, 292–293, 295 mental and physical, 261–262 Reason. See Rationality Recognition, 177, 181, 445 Recognitional capacity, 184–185, 253, 409, 412 Recognitional concepts, 181, 182, 247–250, 252, 253, 445–446 Recognitional dispositions, 179, 181 Reductionism, 78, 182–183, 319 Reference, 4, 109, 232, 235, 237, 238, 240, 250, 265, 279, 310, 351. See also Demonstrative reference contextual, 378, 380 Descartes’s view of, 315–316, 318 directedness and, xiii, 253–254 failure of (see Referential error) indexical, 113 intentionality presupposes, 232 magical theory of, 235 “outer,” 440–441, 446 perceptual, 310–312 phenomenality and, 435–448 problem of “inscrutability of reference,” 37 projective-deflationist theory of, 237 as social phenomenon, 280 in thought and perception, 308 Referential applications, 364 Referential dependence, 273–274, 364 Referential error/failure of reference, 248, 251, 313 Referentialists, 241, 242 Referential properties, 251 Referential success, 439 Referents, 37, 38, 40, 47, 86, 351, 352, 364, 380, 459 Reflection, 145 epistemology of, 387–388 Reflective judgment, 203, 459–462. See also Intuitive judgment Reflective understanding, concepts, conceptions and, 383–395 Reflexive consciousness/awareness, 298–300, 441 Reflexive judgments, 203–204, 418, 424 Reflexive performatives, 426–428, 430 Reflexive thoughts, 203–205, 223, 419, 424, 425, 429, 430 Reid, Thomas, 286n15 Representational abilities, 308 Representational character, 9, 30 Representational content, 171, 172, 196n3, 197n10, 316, 413. See also under Phenomenal character Bach-y-Rita experiment and, 178, 179 bodily sensations and, 175–177
Index
externalism and, 169, 170 externalist memory and, 184–186 phosphene experiences and, 177–178 supervenience and, 170, 171 Representational properties of experiences, types of, 183–184 Representational states, 351 Representationism, 172, 176, 177, 181, 186, 197n13, 238, 246, 407–409 argument against, 190–192, 194, 405 representationist objections to, 192–195 argument for, 405 externalist, 168–170, 175, 182, 187, 188, 238, 245 as form of functionalism, 179–180 vs. functionalism, 193 internalist, 166–168, 175 referentialist (see Referentialists) Representationist-qualiphile dispute, 251. See also Qualia Representationist view, 174–175, 405 and commitment to phenomenology, 242 Representation(s), 157–158, 165, 171, 305, 315, 319, 463–464. See also Images; Perceptual representations categories of, 305–306 Descartes’s account of, 5–9 formally general, 306–310, 329n15–16 implicit and explicit, 383 and misrepresentation, 4–5, 7 singular, 305–307, 329n14 (see also Contextdependent representations) Responsibility for success of mental state/event, 202 Rey, Georges, 182 Richard, Mark, 109–110, 112 Rosenthal, David, 225n12 Russell, Bertrand, 328n14, 329n16 “Russell’s Problem and Intentional Identity” (Burge), 329n15 Salmon, Nathan, 62 Sceptical concerns, Cartesian, 79 Scepticism, xvii–xviii, 27, 315, 324. See also under Anti-individualism antisceptical arguments, 340–345 doubt and suspension of belief as acts, 419 about nonperceptual beliefs, 18 reflections on, 335–345 about senses, 16 Scholasticism, 8 Science, 261 Searle, John, 276, 285n8, 331n28 Self-attributions, 211–212, 219, 426–430 authoritative, 201, 418, 420 performative character of, 219
Index
Self-consciousness, 215, 298 Self-evident/basic truths, 391, 393–395 Self-knowledge, xv, 79, 85, 205–206. See also Burge–Heil reconciliation story anti-individualism’s compatibility with, 78 authoritative, 17, 19, 78, 201, 202, 204, 205, 213, 215, 224 mental agency in, 417–432 basic, 202–206, 213, 214, 421 cold, 209 extended, 215 privileged, 213 smoldering, 213–215, 226n14, 420 strong/hot, 218 Self-verifying thought, 203, 205, 215–216, 221, 224, 417, 418, 420–421 Semantic externalism, 153, 155, 157 Semantic representations, 277, 278 Semantic rules, 67, 73, 87, 112–114, 147, 364–367. See also Demonstrative reference Semantics, 268, 466–467. See also Lexical items folk, 266 lexical/natural language, 267 referential, 466 Semantic theory, 68, 79, 112, 113, 381, 466–467 normal form for, 105 Semantic Value (SV), 271 Sense data, 349 Set, concept of, 386 Shadows, 10 Shaftesbury, 3rd Earl of, 275 Sharp grasp, 122, 135, 386, 388 Shifted spectra, 188–192 Shoemaker, Sydney, 175, 192, 195–196n3, 197n13, 208 Simulation, 120–121, 149 Singular directedness, 239, 240, 439, 446, 447 Singular reference, 235, 310, 440, 445–447 Singular representation, 305–307, 329n14. See also Context-dependent representations Skepticism. See Scepticism Slow-switching, 422. See also under Twin Earth puzzles ignorant vs. knowledgeable, 216–220 Social interdependence, 359 Socially deferential concepts, 250, 253 Sofa case, 184, 284, 364, 459 Solipsism, methodological, 79, 80. See also Autonomous psychology Space, traversing a, 398 Spatial concepts, 249, 250, 446 Spatial content, 130–131 Spatial location (in environment), 312–314, 435
503
Spatial relations and physical objects, concepts of, 237–238, 249–251 Spectra. See also Inverted spectra shifted, 188–192 Speech acts, 201 Spinoza, Benedict de, 419 Strawson, P. F., 212 Study of Concepts, A (Peacocke), 117, 139, 150n1 Substances, 12, 13 distinguished from their principal attributes, 299–301 Substitution strategy for removing puzzles, 111, 112 Supervenience, 163, 170–171 failure of, 32, 171, 302 anti-individualism and, 371–372 Supplemented and unsupplemented derivations, 146–147, 149 Surrealist sculpture, 244–245 Swampman case, 36, 41–45, 50, 170–171, 266, 354–356, 361n14. See also under Inverted Earth (experiment) Syntax, 267–268 Systematic error, 325 Tacit belief, 124, 125, 150n5 Tacit conceptions, 125 Telic, 202, 210–213, 219–224, 420 That-clauses, 229–230, 242, 436, 437 Theory of innate semantic representations (TISR), 277–278 Thetic, 202, 209, 223, 424 Thinker-dependence, 128–129 Thought experiments, 3–4, 59, 71–72, 166. See also Burge–Heil reconciliation story; Coke machine example; Swampman case; Twin Earth puzzles; specific topics anti-individualism and, 20–21 Cartesian, 1–2, 7, 9–13, 291 centering on diaphanousness, 406 Donnellan on, 59–73, 363–369 Donnellan on differences between, 363–365 folk theories and, 280–285 individualism and, 1–2, 59–60, 69 regarding internalist forms of representationism, 166–167 natural-kind (see Kinds) nature of, 460 Thought(s)/thinking, 5, 83, 153, 297 as active vs. passive process, 201 categories of, 296 externally individuated, 153, 155, 156 first- and higher-order, 202–203, 205, 211–213, 220–222, 419, 429
504
Thought(s)/thinking (cont.) mind as “producing,” 299 social factors in determining the nature of, 326–327 thinking as essential attribute of mind, 296, 299 Time, sense of, 375 Token-demonstrative elements, 348, 423 Token-identity theory of mental states and brain events, 59 Transcendental arguments, 260, 340, 342 Transcendental object, 447 Transcendental responses, 18–20 Transmodal vs. amodal concepts, 250, 256n16 Transparency of difference, 219, 226n15 Treasury Department example, 159–160 Triangulation argument, 36, 47, 55 Truth/satisfaction-conditions, 135–136, 138, 144–148, 231, 377–379, 391, 392, 439 Twin Earth experiments, xiv, 33, 83, 184–185, 266, 363–369, 388, 398–399. See also “Arthritis” thought experiment; Inverted Earth; Putnam; “Water” experiments; specific topics Burge’s central claim in, 371 concepts employed by twins as having indexical element, 79–86 Davidson on, 40, 46–48 Descartes on, 9–10, 319–321 difference in twins’ mental content, 438 difference in twins’ physical environments, 77 externalism and, 168–170 Kaplan and, 84–88, 97 Twin Sun and, 319–322 Twin Earth methodology, 456 Twin Earth puzzles, “slow-switching,” 206, 215–221, 226n13, 423 Twin-in-a-vat, 246, 250, 253. See also Brain(s) in vat(s) Tye, Michael, 180 Understanding, 379, 392 Union/unity, 12–13 projecting principles of, 318–319 Vat-in-the-image. See Brain(s) in vat(s) Virtual belief, 124, 125 Visual experience, 30, 166–170, 172, 234, 238–239, 401, 443. See also Colors; Qualia seeing as both a conceptual and, 183 Visual system, 178, 310–314, 320–321, 401, 440, 469n16 Wandell, Brian A., 469n16 “Water” experiments and examples, 31, 33, 60–64, 77–80, 84, 85, 87, 88, 91, 93–95, 168–170,
Index
183–184, 186, 206, 236, 266, 280, 281–282, 323–325, 351, 365–367, 422, 448, 456–458 Will, 8, 206, 207 Wittgenstein, Ludwig, 120, 141–142 Words, 105–106 intensions and extensions, 79–81 learning meanings of, 36 Word-to-world direction of fit, 201 World-to-mind direction of fit, 202, 209, 211 Wright, Crispin, 142 Zeno’s paradoxes, 398