ordinary objects
This page intentionally left blank
Ordinary Objects
Amie L. Thomasson
1
2007
1 Oxford Univers...
45 downloads
1288 Views
1MB 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
ordinary objects
This page intentionally left blank
Ordinary Objects
Amie L. Thomasson
1
2007
1 Oxford University Press, Inc., publishes works that further Oxford University’s objective of excellence in research, scholarship, and education. Oxford New York Auckland Cape Town Dar es Salaam Hong Kong Karachi Kuala Lumpur Madrid Melbourne Mexico City Nairobi New Delhi Shanghai Taipei Toronto With offices in Argentina Austria Brazil Chile Czech Republic France Greece Guatemala Hungary Italy Japan Poland Portugal Singapore South Korea Switzerland Thailand Turkey Ukraine Vietnam
Copyright # 2007 by Oxford University Press, Inc. Published by Oxford University Press, Inc. 198 Madison Avenue, New York, New York 10016 www.oup.com Oxford is a registered trademark of Oxford University Press All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior permission of Oxford University Press. Library of Congress Cataloging-in-Publication Data Thomasson, Amie L. (Amie Lynn), 1968– Ordinary objects / Amie L. Thomasson. p. cm. Includes bibliographical references and index. ISBN 978-0-19-531991-0 1. Object (Philosophy) 2. Common sense. 3. Ontology. I. Title. BD336.T46 2007 111—dc22 2006052471
1 3 5 7 9 8 6 4 2 Printed in the United States of America on acid-free paper
To Natalie
This page intentionally left blank
acknowledgments
The first full draft of Ordinary Objects was written with the kind support of the National Endowment for the Humanities, in the form of both a summer stipend (2002), and later a fellowship (2003–2004). I wish to extend to them my sincere thanks for making this work possible. Thanks also to the Centre for Consciousness at the Australian National University for having me as a Visiting Fellow (June–August 2005); there I enjoyed enormously fruitful discussions of these issues while revising the manuscript. Since its initial drafting, the book has gone through numerous revisions thanks to the helpful input of a great number of philosophers who have either read the manuscript or parts of it, or been present where parts of it were presented at conferences and colloquia. For reading and providing extensive feedback on the entire manuscript (at differing stages), special thanks go to Crawford Elder, Simon Evnine, Katherine Hawley, Dan Korman, Jonathan Lowe, Kay Mathiesen, Alan Sidelle, and Ted Sider. For helpful written comments and/or discussion of central portions of the manuscript, I am also grateful to Roberta Ballerin, Karen Bennett, Jacek Brzozowski, Ota´vio Bueno, David Chalmers, Matti Eklund, Liz Giles, George Graham, John Heil, Risto Hilpinen, Terry Horgan, Uriah Kriegel, Huaping Lu, Michael McCracken, Trenton Merricks, Brian Mondy, Michael Rea, Mark Sainsbury, Steve Savitt, Harvey Siegel, Susanna Siegel, David Woodruff Smith, Jason Turner, and Michael Tye. I have discussed and corresponded about these issues with so many people over the past few
viii
acknowledgments
years that I fear I am forgetting various people, and offer my sincere apologies to those whose names are inadvertently omitted here. I was fortunate to be able to present central portions of this material in colloquia at the University of British Columbia, the University of Delaware, Duke University, Durham University, Harvard University, the University of North Carolina–Greensboro, the University of South Carolina, the University of Texas–Austin, The University of Toronto, and Washington University in St. Louis. Portions were also presented at conferences including the American Philosophical Association Pacific Division, the Australasian Association of Philosophy, the ‘‘Criteria of Identity’’ conference in Leuven, Belgium, the Syracuse Workshop in Metaphysics, and in the On-line Philosophy Conference. My thanks go to the organizers of those events and to members of the audiences on each of those occasions for helpful discussions. Finally, I have had the pleasure of presenting parts of the work-inprogress in two graduate seminars at the University of Miami; one on ‘‘the world of common sense’’ and the other on metaontology. Special thanks go to the students of those seminars for ongoing discussions of the surrounding issues. Some of the material from chapters 1 and 4 has previously appeared as ‘‘Metaphysical Arguments against Ordinary Objects’’ (Philosophical Quarterly 56, 224 [ July 2006]: 340–59). Thanks to the editors for permission to reprint that material here. Thanks also to Ben Burgis for help preparing the index. My deepest thanks go to my husband, Peter Lewis, for offering his insight in ongoing philosophical discussions whenever and wherever needed, and especially for his encouragement, confidence, and love throughout the difficult process of producing this ordinary object.
contents
Introduction 1.
2.
3.
3
Problems of Causal Redundancy 9 1.1 The Threat of Causal Exclusion 10 1.2 The Threat of Overdetermination 15 1.3 Other Apparent Redundancies 20 1.4 The Difference between Baseballs and Minds
24
Analyticity and Conceptual Content 28 2.1 One Dogma of Quineans 29 2.2 Truth by Convention 32 2.3 Causal Theories of Reference and the Qua Problem 38 2.4 The Basis of Analytic Entailments 44 2.5 The Problem of Nonexistence Claims 45 2.6 Objections to Hybrid Theories of Reference
48
Identity, Persistence, and Modality 54 3.1 Existence, Identity, and Persistence Conditions 55 3.2 Modal Truths 62 3.3 Modal Conceptualism and Objectual Antirealism 63 3.4 Analyticity and Truth-Makers 68 3.5 Modal Properties 70
x
contents
4.
Problems of Colocation 73 4.1 ‘Nothing Over and Above’ 75 4.2 The No Coincidence Principle 78 4.3 Property Additivity 80 4.4 The Grounding Problem 81
5.
Problems of Vagueness 87 5.1 Sorites-Style Arguments 88 5.2 The Source of Vagueness 90 5.3 Solutions to the Problems of Vagueness 95 5.4 Vagueness and Contextual Semantics 100 5.5 Is There Vagueness in the World? 104 5.6 Can There Be Vague Objects? 107
6.
Handling Existence Questions 110 6.1 Specific Existence Questions 111 6.2 Generic Existence Questions 112 6.3 Bare Quantification 115 6.4 Quantifier Variance 118 6.5 The Number of Objects 119 6.6 Can We Revive the General Question?
121
7.
The Special Composition Problem 126 7.1 Uniform Answers to the Special Composition Question 127 7.2 Nonuniform Answers to the Special Composition Question 130 7.3 Behind the Special Composition Question 134
8.
Problems of Rivalry with Science 137 8.1 The Case for a Conflict 138 8.2 The Case for a Rivalry 144 8.3 Is There Really a Rivalry? 147
9.
Parsimony and Ontological Commitment 151 9.1 Parsimony’s Plausibility 152 9.2 Parsimony and Counting 154 9.3 The Case for Ordinary Objects 155 9.4 Trivial Transformations and Ontological Commitments 159 9.5 The Point of Paraphrase 168
contents
9.6 9.7
Are We Committed to Extraordinary Objects? 170 The Price of Avoiding Ordinary Objects 173
10. A Coherent Common Sense View 176 10.1 A Unified Diagnosis 177 10.2 The Basis for a Common Sense View 10.3 Too Many Objects? 183 10.4 The Specter of Antirealism 185
180
11. The Methods of Metaphysics 188 11.1 Identity and Persistence Questions 189 11.2 Existence Questions 192 11.3 Genuine Debates and Merely Apparent Debates 11.4 What’s an Ontologist to Do? 199 Notes
203
Bibliography Index
235
225
195
xi
This page intentionally left blank
ordinary objects
This page intentionally left blank
introduction
It has become increasingly common to see philosophers of various persuasions argue that ordinary inanimate objects such as sticks and stones, tables and chairs, simply do not exist.1 These arguments are often met with incredulous stares and claims that it is just common sense that such things do exist. But little has been done to address the arguments thoroughly and as a body, as I will in this volume. Some may think that detailed diagnoses of and replies to eliminativist arguments are unnecessary, holding (in the style of G. E. Moore’s replies to the skeptic, 1959, 226) that we are much more certain that there are tables, chairs, and other ordinary objects than we are of the soundness of any philosophical arguments to the contrary, so that we may assume that such arguments must have gone wrong somewhere, without needing to pinpoint where. But the past failure to provide unified and detailed responses to these arguments is unfortunate even if one thinks that entrenched common sense claims cannot be threatened by philosophical argumentation. For failure to address the concerns behind the various arguments against ordinary objects means failure to step up to the challenge of showing how our common sense worldview can be developed in a way that avoids conflict or rivalry with science, while also avoiding internal contradictions or conflicting with plausible general metaphysical principles and demands. Showing how, reflectively, we can make sense of our unreflective common sense worldview is arguably one of the chief tasks of philosophy, and we owe a debt to those eliminativists who show us the challenge of doing this 3
4
ordinary objects
well by drawing vividly into relief the dangers that lie on all sides of this conceptual territory. Arguments against ordinary objects have taken a great variety of forms and arise from diverse impulses, ranging from naturalist inclinations to accept only the ontology yielded by our best physical theories to pure a priori arguments based on apparent contradictions within our ordinary concepts. Causal redundancy arguments are based on the claim that accepting ordinary objects would require us to accept the existence of material objects that lack causal powers, in conflict with our common sense view that baseballs, chairs, and the like may have causal impact, and in violation of Alexander’s dictum that ‘‘to be real is to have causal powers.’’ Anti-colocation arguments against ordinary objects arise from the fact that if we accept ordinary objects such as chairs and statues as well as the lumps of stuff that constitute them, we wind up having to accept that there are two (or more) objects that completely overlap in their spatial location and fundamental parts, in apparent violation of common sense. Moreover, it seems that we must then accept that these objects differ in their sorts and modal properties despite overlapping in all their basic (nonmodal, nonsortal) properties, landing us in the grounding problem: How can these sortal and modal differences then be explained? Still other arguments are based on problems of vagueness. Sorites arguments are often used to argue that the very concepts of ordinary objects are selfcontradictory, so that nothing can correspond to them. Others argue that no sense can be made of ontological vagueness, and that in consequence there can be no objects corresponding to our vague concepts. Other reasons commonly given for rejecting ordinary objects are based on taking a more global approach and arguing that we can get a more unified, less ad hoc, better epistemically justified, or more parsimonious ontology overall by rejecting ordinary objects. The special composition question—of when simples compose some further object—has lately been wielded against ordinary inanimate objects in arguments that suggest that any acceptable (and nonarbitrary) answer to the question must omit inanimate composite objects. Others have argued that a common sense ontology that includes ordinary objects conflicts with, or is a rival to, a scientific ontology, and that (given the superior epistemic credentials of the latter) the scientific ontology must win this contest. Finally, many who are reluctant to accept ordinary objects cite grounds of parsimony, urging that we can offer a better, because sparer, metaphysics without them.
introduction
5
A central contention of this book is that, despite their diversity in origin and superficial form, these arguments against ordinary objects have much in common. In fact, I will argue that a small cluster of interrelated and independently plausible theses about reference, analyticity, and modality enables us to diagnose the problems behind all of the arguments against ordinary objects and to resist these arguments successfully. While I provide some direct, bottom-up arguments in favor of this cluster of views in chapters 2, 3, and 6, the most important evidence for it comes from its unified ability to dissolve a wide range of metaphysical puzzles regarding ordinary objects. It also enables us to meet the challenge of showing how we can make sense of our common sense worldview without internal contradiction, violation of plausible metaphysical principles, or rivalry with a scientific ontology. Thus this book pursues not merely the defensive goal of showing how to defuse arguments against ordinary objects but also the constructive goal of showing how to develop a defensible common sense ontology. Since this book’s central goals are to demonstrate a way to resist the diverse arguments of eliminativists and defend a common sense ontology, its intended audience is not primarily eliminativists themselves. For the ways in which I diagnose the problems behind eliminativist arguments and build up a coherent common sense ontology are based on certain fundamental theses about meaning, reference, and modality that many of my opponents would reject. While I defend these views against some of the more persistent objections and give reasons for preferring them to the starting-places of my opponents, that is of course not the same as showing that one must accept my starting points rather than theirs, and I expect few converts from among the hard-core eliminativists. Nonetheless, I do hope that the work of the book will help pinpoint exactly where the superficial disagreements between eliminativists and common sense ontologists lead back to, and thus fruitfully refocus debate on the underlying issues about reference, modality, quantification, and the methods of metaphysics. The primary intended audience instead includes those who approach these issues from a neutral standpoint, and those who are inclined to accept a reasonably common sense ontology but find the multitude of arguments wielded by eliminativists worryingly persuasive. To the latter I hope to offer some ammunition in resisting these arguments, and tools for showing how a certain sort of common sense view may be coherently defended. To both of these groups, I hope to make evident what other tacit philosophical beliefs may be involved in arguments for rejecting or accepting a common sense ontology, as
6
ordinary objects
the basis for making a more informed global choice based on seeing more thoroughly what is at stake in each of these approaches. I begin the main work in chapter 1, which is designed to motivate the work that follows by taking a preliminary look at a sample argument against ordinary objects: the causal redundancy argument. A key problem behind this argument (and some later arguments), I argue, lies in accepting as universally applicable a principle that does not apply when the claims substituted into it are analytically interrelated. Since this diagnosis relies on the idea that there are analytic entailments among our claims, the response to causal redundancy arguments cannot be fully justified until later work is done in defense of analytic entailments. Nonetheless, the preliminary examination of causal redundancy arguments here provides a clue as to where we should turn next in looking for a global defense of ordinary objects. The very idea of analyticity has been somewhat out of fashion in the wake of Quine’s criticisms of the analytic/synthetic distinction, and the associated idea that our terms have conceptual content that can ground these analytic interrelations is widely rejected by those who accept a pure causal theory of reference. I thus turn in chapter 2 to address these criticisms and show how, notwithstanding, we can develop a defensible understanding of analytic entailments. I suggest a route for defending the idea that there are analytic entailments against Quinean criticisms of analyticity, and also argue that in light of the qua problem and problems of handling nonexistence claims, even those inclined to causal theories of reference have reason to accept that our nominative terms have determinate reference only to the extent that they are disambiguated by competent speakers associating them with a basic category of entity to be referred to. This view of reference leads to some interesting conclusions, especially that the most basic claims about existence conditions, identity conditions, and persistence conditions, as well as the most basic modal claims, are analytic. In chapter 3 I draw out these consequences, show how they are related to the conclusions of chapter 2, and defend them against some common objections bred in misunderstandings. In chapter 4 I show how the theses of chapters 2 and 3 provide a clear way of seeing what’s gone wrong in arguments based on problems of colocation. I argue that, like the causal redundancy arguments of chapter 1, most of the alleged difficulties with colocation arise from accepting completely general metaphysical principles that should in fact be restricted to cases in which the claims involved do not involve analytic entailments. The grounding problem is based on the demand
introduction
7
to explain how there could be differences in modal properties for objects that (at some level of decomposition) share all their parts as well as their nonmodal (and nonsortal) properties. But, I argue, the view of modality argued for in chapter 3 enables us to see why this demand is inappropriate. In chapter 5 I argue that the views developed in chapter 2 about how application and coapplication conditions for our customary terms are set up also enables us to understand the source of vagueness. Once we see the source of vagueness, I argue, it becomes clear why it should not be considered a problem that should lead us to reject ordinary objects. Chapter 6 lays out some further consequences of the theses of chapters 2 and 3, in particular for determining what existence and counting questions are well-formed and answerable, and how they are to be answered. In short, I argue that while there are various ways of addressing questions about what ‘things’ exist or how many ‘objects’ there are, we have reason to think that any fully meaningful, answerable form of these questions requires that we (tacitly or explicitly) specify what category or categories of object we are enquiring about. However, (as I argue in chapters 7, 8, and 9, respectively) arguments against ordinary objects based on the special composition question, alleged rivalry with a scientific ontology, and concerns with parsimony all rely on a completely category-neutral approach to existence or counting questions. Thus the hypothesis that category-neutral versions of these questions are ill formed provides the basis for a unified diagnosis of the common problem underlying all of these arguments. The basic cluster of views developed in chapters 2, 3, and 6 provides the grounds from which to argue that there is no rivalry between the ontologies of science and common sense, no pressure to deny ordinary objects to provide a more parsimonious or less ad hoc ontology, and that the eliminativists’ arguments do not provide genuine cases in which legitimate general metaphysical principles or demands conflict with accepting the existence of ordinary objects. If that is the case, so much the better for all of us, for then the drive to make sense of our common sense ontological judgments is not after all in conflict with the need to accept the best scientific theories or maintain plausible metaphysical principles and theories. The alternative approach to existence questions developed in chapters 2 and 6 also provides a simple way of arguing for the existence of ordinary objects, as I will show in chapter 9, where we finally arrive at a constructive argument for ordinary objects. One important positive result of this study lies in demonstrating how the claim that there are
8
ordinary objects
ordinary objects can form part of a coherent, reflective metaphysical view built up out of our common sense way of looking at the world, in a way that avoids the philosophical problems that have long been feared to plague a common sense ontology. In chapter 10 I come back to make the case that the views defended in chapters 2, 3, and 6 are able not only to reveal the problems with arguments against ordinary objects but also to form the basis for a common sense ontology— further adding to the indirect evidence in favor of these basic views about reference, analyticity, and modality. But the most important result of addressing these eliminativist arguments is not merely avoiding their conclusions or even developing a positive common sense view. Diagnosing the common problems behind the eliminativist arguments requires reopening issues about meaning, reference, and conceptual truths, and quantification, counting, and existence questions, with results that may have application in many other areas of philosophy. Careful study of these arguments for eliminativism is thus useful even for those convinced in advance of the existence of ordinary objects, since examining their failings brings into question certain widely held assumptions about what uses of metaphysical principles are appropriate, what metaphysical demands are answerable, and how we should go about answering such fundamental questions as ‘What exists?’ As a result, as I will argue in closing in chapter 11, the payoff of this study lies not just in a defensible common sense ontology but also in a better understanding of the methods and limits of metaphysics.
C c
one
problems of causal redundancy
To set the stage for the work that follows, I will begin with a preliminary look at one major form of argument raised against the existence of ordinary macroscopic objects such as tables and chairs, sticks and stones: the causal redundancy argument. The basic strategy of this argument is to show that all the causal work allegedly done by macroscopic inanimate objects may be accounted for by the causal powers collectively contributed by their microscopic parts, properly arranged, so that ordinary inanimate objects (if there were any) would be epiphenomenal.1 But this conflicts with the dictum that ‘‘to be real is to have causal powers’’—a metaphysical principle2 that, as Trenton Merricks argues (2001, 81), is at least quite compelling for macroscopic physical objects (whatever one thinks of its applications to abstracta). As Peter van Inwagen writes, all the activities apparently carried out by shelves and stars and other artifacts and natural bodies can be understood as disguised cooperative activities [of simples properly arranged]. And, therefore, we are not forced to grant existence to any artifacts or natural bodies. (1990, 122)
Typically the alleged causal redundancy of ordinary inanimate objects is used only as a way of suggesting that we need not posit their existence. Merricks (2001), however, develops the argument in a stronger form, arguing that the causal redundancy that ordinary inanimate objects would exhibit (if there were any) does not merely relieve us from the need to posit such objects—it demonstrates that there are no 9
10
ordinary objects
such objects, since it shows the very idea of such objects to be inconsistent. Using the example of a supposed case of a baseball shattering a window, Merricks develops what he calls the ‘‘overdetermination argument’’ (see x1.1 below) to show that the baseball, if it exists, does not cause the window to shatter. But, he adds, if there were baseballs, they would have causal powers: If there were baseballs, they would break windows, they would injure batters, they would cause visual sensations (and so be seen), and they would cause tactile sensations (and so be felt). . . . But given the Overdetermination Argument . . . if there were such objects, they would not have causal powers, so there are no such objects. (81)
Since positing the existence of baseballs would force one to conclude that they both have and do not have causal powers, we can conclude that there are no baseballs. One could of course avoid the conclusion that positing baseballs leads us into inconsistency, but only at the cost of accepting that baseballs, if they existed, would be epiphenomenal, leaving the way open for eliminativists to maintain that we need not posit them, and indeed have no reason to posit them. Even this weaker conclusion would be (at the least) unpalatable for the defender of ordinary objects.
1.1 The Threat of Causal Exclusion But must we accept that baseballs, if they exist, would lack causal powers? The crucial argument for this conclusion is the overdetermination argument (Merricks 2001, 56): (1) The baseball—if it exists—is causally irrelevant to whether its constituent atoms, acting in concert, cause the shattering of the window. (2) The shattering of the window is caused by those atoms, acting in concert. (3) The shattering of the window is not overdetermined.
Therefore, (4) If the baseball exists, it does not cause the shattering of the window.
Merricks in turn defines the causal ‘‘irrelevance’’ in (1) as: O is causally irrelevant to whether the xs cause E just in case none of the following four situations holds (59–61):
problems of causal redundancy (i) (ii) (iii) (iv)
11
O is one of the xs O is a partial cause of E alongside the xs One or more of the xs causes O to cause E O causes one or more of the xs to cause E.
The license to move from (1–3) to (4) is based on a principle that lies in the background of this argument and is essential to its validity. Merricks simply calls it the ‘‘causal principle’’: Suppose: O is an object. The xs are objects. O is causally irrelevant to whether the xs, acting in concert, cause a certain effect E. . . . The xs, acting in concert, do cause E. And E is not overdetermined. It follows from all this that O does not cause E. (2001, 58)
The causal principle may sound unobjectionable at first glance, and Merricks illustrates it with an example that makes it seem uncontroversial: Suppose . . . the members of an unruly mob, cause the vandalism of a park. Suppose also that the vandalism of the park is not overdetermined. And finally, suppose that I am ‘causally irrelevant’ to whether those members cause the vandalism . . . I am not myself one of the members. Second, I am not a ‘partial cause’ of the vandalism alongside the members. . . . Third, I am not an intermediate in a causal chain between the members and the vandalism; that is, the members do not cause the vandalism by causing me to do something by which I, more proximately, cause the vandalism. And, finally, I do not cause any of the members to cause the vandalism . . . [then] I do not cause the park to be vandalized. (2001, 57–8)
And indeed it is unobjectionable, as long as we implicitly limit the discussion to entirely separate and independent objects. This is the case with Merricks’s example of himself and the mob (since he is not part of the mob) and is normally assured of being the case in scientific causal explanations, where the discourse is limited implicitly to the question of which particle, or which microbe, and so on, caused the relevant effect. But the question at issue, whether or not baseballs (if there are any) are rendered epiphenomenal in virtue of the causal work of their parts, is not of this form. The vandalism case could be made parallel only if one considered not whether a separate bystander caused the vandalism of the park, but whether the mob caused the vandalism of the park. The causal principle would also yield the conclusion that (although the members of the mob cause the vandalism) the mob does not cause the park to
12
ordinary objects
be vandalized.3 But this hardly has the immediate plausibility required to motivate us to accept the causal principle. In fact, it is just as plausible (or implausible) as the conclusion that the baseball does not cause the window to shatter, and so cannot be used to garner independent support for the causal principle that can then be used to argue that the baseball does not cause anything. I will return below to the question of whether we should accept the causal principle as applying quite generally—even to entities that are not separate and independent. For now, let us provisionally accept it, for the sake of argument, noting that if we accept it, the argument is valid. But is it sound? At first glance, the overdetermination argument seems to show that the baseball doesn’t cause the window to shatter. Premise 2 seems fine, since it seems that the collective action of the atoms is legitimately described as causing the shattering.4 Premise 1 looks okay, at least given Merricks’s definition of ‘causal irrelevance’, since the baseball is clearly not one of its atoms, nor is it a mediating item in the causal chain— either being caused by its atoms to cause the shattering of the window or causing its atoms to cause the shattering (by downward causation), and it would seem wrong to say that ‘‘alongside’’ its constituent atoms, the baseball is a partial cause of the window’s shattering (2001, 59).5 Premise 3 initially looks plausible, since we don’t want to say that every time a baseball hits a window, the shattering is overdetermined. Merricks defines ‘‘overdetermination’’ as follows: ‘‘an effect is overdetermined if the following are true: that effect is caused by an object; that object is causally irrelevant to whether some other—i.e. numerically distinct—object or objects cause that effect; and the other object or objects do indeed cause that effect’’ (2001, 58). Claims of overdetermination thus seem to involve making a conjunctive causal claim, so that ‘‘one would have a reason for believing that the shattering of the window is overdetermined only if one had a reason for believing that both the baseball and the atoms arranged baseballwise caused it’’ (72, italics original). But while we might say that a bullet through the heart and another through the head both caused a death (thus overdetermining it), we would hesitate to say that the baseball and the atoms arranged baseballwise both caused the shattering (thus overdetermining it). Clearly something is amiss with saying that the baseball and the atoms arranged baseballwise both caused the window’s shattering, and it seems to be our disinclination to accept that that makes us inclined us to accept premise 3. But the problem with the claim does not seem to
problems of causal redundancy
13
be that it is false (making its negation, premise 3, true). Instead, what seems to be wrong with saying the baseball and the atoms arranged baseballwise both caused the shattering is that it involves conjoining claims about a whole and its parts in a single list (saying both the parts and the whole did the causing). Other statements of the same form seem similarly problematic—even if they mention nothing of causation. Ryle (1949/1984) identifies claims of this form, such as ‘‘he bought a left-hand glove and a right-hand glove and a pair of gloves’’ (22) and there was ‘‘a parade of batteries, battalions, squadrons and a division,’’ (17) as prime examples of absurd statements involving category mistakes. But what exactly is wrong with statements of this kind, and why are they inappropriate?6 The reason they feel inappropriate seems to be that conjoining items in a list with ‘and’ (especially where this is reinforced with ‘both’ or ‘all’) normally presupposes that the items conjoined are separate and independent, but that presupposition is violated in cases like these. This is closely related to the constraints of the Gricean conversational maxim of brevity (Grice 1989, 27) since, provided the listeners know of the relation between the two clauses, it would be pointless to assert the second once the first has been asserted.7 Nonetheless, the constraint does not seem to be just audience-relative; even speaking to an audience of children of civilians (who could not be expected to know the relation between squadrons and divisions and so could not infer claims about one from claims about the other), it would be inappropriate and misleading to say ‘I saw both squadrons and a division’; indeed putting things this way (rather than as, say, ‘I saw the squadrons that make up the division’) would lead them to suppose that squadrons and divisions were separate and independent entities of a similar sort. So in short it seems that it feels wrong to say that the baseball and its constitutive atoms are both causes of the shattering not because it is false, but rather because either asserting or denying the claim presupposes that the baseball and the atoms arranged baseballwise are separate and independent. Asking, of two rocks or two people, whether they both caused a window to break (by each hitting the window forcefully at the same moment) is perfectly sensible. But, as Merricks himself says, ‘‘the baseball and the atoms are not—according to anyone— relevantly analogous to two rocks jointly shattering the window. . . . For while two rocks can do more work than one, a baseball and its constituent atoms cannot do any more than those atoms all by themselves’’ (2001, 59).
14
ordinary objects
If our disinclination to say that the shattering is overdetermined by the baseball and its constituent atoms both causing it (and thus our inclination to accept premise 3 of the overdetermination argument) results from the fact that the presupposition for claiming that both are causes is violated, the question remains whether the overdetermination claim is thus lacking in truth-value, or merely misleading, though capable of being true or false. Some treat cases where presuppositions are violated as neither true nor false, for example, Ryle (1949/1984) treats them as absurd. While they do not discuss presuppositions of conjunctive claims like those above, Strawson (1950, 330) and Austin (1962, 50–1) treat sentences where a referential presupposition fails as cases where the question of truth or falsity doesn’t arise, the statement (as Austin says) being ‘‘void.’’ Grice (1989, 1–21) argues that at least in many cases of presupposition failure, the statements are capable of being true, strictly speaking, although they are misleading, and Stalnaker (1973, 451) remains neutral on the issue. It seems more natural to consider conjunctive claims involving failure of the independence presupposition as capable of being true (though misleading), but in any case we need not decide the issue here. For either way, the overdetermination argument is in trouble. Suppose we hold that statements violating the independence presupposition are neither true nor false. Then the overdetermination claim that the baseball and the atoms arranged baseballwise both cause the shattering is neither true nor false. But since negations of statements with presupposition failure also suffer from presupposition failure (Soames 1989, 558; Stalnaker 1973, 448), the (premise 3) claim that the shattering is not overdetermined must also on this assumption be neither true nor false. But then the overdetermination argument is not sound. Suppose that, instead, we say that such statements are capable of being true, but are misleading (since their assertion may lead hearers to suppose that the speaker presumes that the conjoined items are independent). If so, then it seems we should accept that the overdetermination claim that the baseball and its constituent atoms both cause the shattering is true (but misleading). (After all, each has a prima facie claim to causation, and Merricks endorses the atoms’ claim to causation in premise 2 and can’t—in this premise—deny the baseball’s claim without begging the question.) But then premise 3—the claim that the shattering is not overdetermined—would have to be false, and so again the argument is unsound and does not establish that the baseball does not cause the shattering. In either case, we can explain away our
problems of causal redundancy
15
hesitancy to accept the overdetermination claim without accepting its negation in premise 3.
1.2 The Threat of Overdetermination It might look as though this move saves the realist about ordinary objects from one problem only to land in another: accepting widespread and systematic overdetermination in order to save ordinary objects, in violation of another plausible metaphysical principle—that we should always resist systematic causal overdetermination (Merricks 2001, 67). Thus far I have been using Merricks’s definition of ‘overdetermination’ as applying whenever an ‘‘effect is caused by an object; that object is causally irrelevant to whether some other—i.e. numerically distinct— object or objects cause that effect; and the other object or objects do indeed cause that effect’’ (58). But is this so-called overdetermination real overdetermination of the sort that is supposed to be objectionable? The worry with overdetermination, more traditionally understood, is that it involves an unwarranted ‘doubling up’ of the causal story, by positing two (each separately sufficient) causes of an event where there is no special evidence for two, and one would suffice to do the explanatory work alone. But while positing two diseases, or two particles, as both causing a given effect may involve a superfluous ‘doubling up’ of the causal story needed, it is not at all obvious that accepting the causal claims of the baseball and the baseballwise arranged atoms does any such thing.8 In Jaegwon Kim’s original formulation of the problem of explanatory exclusion, he stated it as the principle that ‘‘two or more complete and independent explanations of the same event or phenomenon cannot coexist’’ (1993, 250, italics mine). But this independence seems to be lacking between the causal claims of the baseball and the atoms arranged baseballwise. So it is not at all obvious that, in cases in which independence does not hold between objects A and B, A and B either provide double the amount of causation or are rivals for causal influence. As Stephen Yablo puts it in a similar context, ‘‘in this competition wholes and parts are not on opposing teams; hence any principle that puts them there needs rethinking’’ (1992, 183).9 Developing this reply, however, requires offering a more specific account of when the independence condition fails, and why such failures of independence should be thought to undermine genuine causal redundancy or rivalry. I will describe at least one kind of case where
16
ordinary objects
I think it is clear that the relevant sort of independence fails, and clear why such failures of independence undermine causal doubling or rivalry claims. But while this will be sufficient to do the trick in this case, I do not mean to suggest that this is the only sort of nonindependence that can block causal redundancy claims—others would simply have to be considered separately.10 One clear case in which the independence condition fails for entities x and y is when claims of x’s causal relevance analytically entail claims of y’s causal relevance. I use the expression ‘analytically entail’ to mean ‘entail in virtue of the meanings of the expressions involved and rules of inference’, so that a sentence (or set of sentences) j analytically entails a sentence c just in case, given only logical principles and the meanings of the terms involved, the truth of j guarantees the truth of c. Thus where j analytically entails c, given knowledge of the truth of j, as well as grasp of the meanings of the terms and reasoning abilities, a competent speaker may legitimately infer the truth of c on that basis alone.11 Thus, for example, ‘Jones bought a house’ analytically entails ‘Jones bought a building’, as the truth of the latter is guaranteed given the truth of the former, as any competent speaker (and reasoner) who understands the meanings of ‘house’ and ‘building’ can see. (In chapter 2 I return to defend the idea that there are such analytic entailments.) It is easy to see why, if causal claim j analytically entails causal claim c, the two are neither additive nor rivals. Consider what is involved in analytic entailments. If claim j analytically entails claim c, then competent speakers can infer the truth of c merely by knowing the truth of j and knowing the relevant meanings of terms (and being competent reasoners). But if this is the case, then clearly c requires no more of the world for its truth than j already required—sufficient truth-makers in the world for j are also sufficient truth-makers in the world for c, they just make a new claim c true.12 If the truth of c does not require anything more of the world than the truth of j requires, then clearly it does not require any extra causal action beyond what was averted to in j, and there is no doubling of or competition between the two claims. The claims ‘the atoms arranged baseballwise are causally relevant to the shattering’ and ‘the baseball is causally relevant to the shattering’ seem to be of just this form. Phrases such as ‘arranged baseballwise’ were introduced by eliminativists like Merricks and van Inwagen to provide a way of paraphrasing talk apparently about ordinary objects in a way that will be committing only to simples, in order to provide a sense in which the ordinary person’s claims about chairs, baseballs, and
problems of causal redundancy
17
the like may still be true (van Inwagen 1990) (or ‘‘nearly as good as true’’, Merricks 2001, 171) even if there are no chairs or baseballs. So the truth-conditions for ‘there are simples arranged chairwise here’ are supposed to mimic those for ‘there is a chair here’ as closely as possible without incurring commitment to chairs. Van Inwagen (105, 109) makes the effort to spell these out, giving conditions for when, for example, simples are arranged chairwise that will enable us to preserve the sense in which claims about chairs are (in the ordinary person’s mouth) true (102). Merricks simply leaves a placeholder, defining the clause ‘atoms arranged statuewise’ as meaning that those atoms ‘‘have the properties and also stand in the relations to microscopica upon which, if statues existed, those atoms’ composing a statue would nontrivially supervene’’ (4).13 That is, if you think that human intentions to make a statue (pursued through rearranging some physical material) are necessary for there to be a statue, consider that a human intending to make a statue (and rearranging atoms accordingly) is also required for atoms to be arranged statuewise. Similarly, if you think that the rules and practices that make up the game of baseball, and intentions of those who rearrange atoms into appropriate spherical shapes, are necessary for there to be baseballs, consider that for atoms to be arranged baseballwise requires atoms tightly bonded in a spherical shape of suchand-such diameter, such that they are jointly capable of fulfilling the functions of baseballs, bonded by people with intentions to make baseballs that meet major league regulations, are usable and to-be-used in such games, and so on. Given this sort of understanding of ‘arranged baseballwise’, any competent speaker and reasoner who knows how to use terms like ‘baseball’ and ‘caused’, is given the rules for applying the term ‘atoms arranged baseballwise’, and knows it is true that atoms arranged baseballwise are causally relevant to the shattering need investigate the world no further to infer that a baseball was causally relevant (and wonder why this was put in such a roundabout way). This gives us prima facie reason to think that the claim that atoms arranged baseballwise caused the shattering analytically entails that a baseball caused the shattering. Of course, fans of the causal redundancy argument might deny that there is such an analytic entailment—either by denying the existence of such entailments (and analyticities) altogether or (even if they accept that there are some analytic entailments) denying that there are, in this case, the relevant analytic interrelations between claims about atoms arranged baseballwise and claims about baseballs. The response to the causal redundancy argument, thus, cannot be complete
18
ordinary objects
until I have defended the general idea that there are such entailments (as I do in chapter 2) and the specific claim about the entailments that hold between claims about simples arranged p-wise and claims about ps (as I do in chapter 9). Nonetheless, even this initial intuitive examination has given us reason to think that there are such analytic interrelations, motivating the next step of defending the very idea of analytic interrelations, and giving us reason to suspect that these may be a clue to what has gone wrong in various arguments against ordinary objects. For provided there are such analytic interrelations, there is no rivalry between the atoms arranged baseballwise’s claim to be causally relevant to the shattering and the baseball’s claim to causal relevance, and so acknowledging the atoms’ causal role gives us no reason to deny the baseball’s—indeed, noting the analytic relations among these claims gives us reason to assert the baseball’s causal efficacy if we assert the collective causal efficacy of its constituting atoms.14 Nor does accepting the claim of each mean accepting real ‘overdetermination’ in the sense of positing causes that do ‘double’ the work necessary. Merricks actually considers a similar reply: the reply that the baseball and the atoms, if both acknowledged as causes, would not really overdetermine the shattering; it would be mere pseudo-overdetermination, for ‘‘what it is for the baseball to shatter the window is for its parts— such as its atoms—to shatter the window’’ (2001, 68). He rejects this move for the following reasons. Consider a case of acknowledged pseudo-overdetermination: the supposed overdetermination that comes from supposing that both an object and the event it participates in are causes of a single effect. So, for example, ‘‘a window’s shattering is pseudo-overdetermined if it is caused by a baseball and caused by the baseball’s striking the window’’ (67). In such cases, he argues, object O’s causing an effect E may be ‘‘analyzed as O’s participating in the appropriate way in an event that causes E’’ (68). But for this analysis to avoid circularity, there must be a different kind of causation involved in the claim of object causation and the claim of event causation. This is plausible, Merricks argues, in the case of object and event causation, ‘‘since events differ categorically from objects’’ (68). But, he argues, such an analysis could not avoid circularity in the case of the baseball and its atoms, since they belong to the same category, and so the same kind of causation must be involved on both sides of the ‘analysis’. But it is unclear why (in either case) one must suppose that the claim that ‘what it is for the baseball to shatter the window is for its
problems of causal redundancy
19
atoms to shatter the window’ is supposed to provide anything like a reductive, noncircular analysis of the statement that the baseball shattered the window. We need not say that the former provides a reductive analysis of the latter to say that the claim that the relevant atoms (arranged baseballwise) caused the window to shatter analytically entails that the baseball they compose caused the window to shatter. We no more need to suppose that ‘cause’ refers to a different relation in each case for that to be true than we need to suppose that ‘burned’ refers to a different event on each side for it to be true that the fact that the house burned down analytically entails that a building burned down. In fact, it is unclear (to me at least) what it even would mean to say that events are engaged in ‘a different kind of causation’ than objects are. The actual reason that there is no real overdetermination involving an object and the event in which it participates, I would suggest, is the same as the reason that there is no real overdetermination involving a baseball and the atoms arranged baseballwise: the claims of causal relevance in each case are not independent—the event’s causal relevance (the baseball’s striking of the window causing the latter to break) analytically entails the object’s (the baseball’s) causal relevance, and the parts’ collective causal relevance analytically entails the whole’s. We are now in a position to reexamine Merricks’s argument. As I have argued above, if we accept Merricks’s definition of overdetermination, it seems that the claim that the shattering is overdetermined is either true or neither true nor false, so that in either case premise 3 is not true, and the argument (though valid) is not sound. But while the baseball ‘and’ its properly arranged atoms fit Merricks’s definition of ‘overdetermination,’ they do not seem to provide a case of real overdetermination of a kind that was supposed to be worrying. If we limit discussion to real overdetermination, it seems, premise 3 is true but the argument is not valid, since the inference enshrined in the causal principle in its general form (as applied unrestrictedly to any O and any xs) is then illegitimate. The causal principle can be written out as a long conditional: (A) O is causally irrelevant to whether the xs cause E, and (B) the xs do cause E, and (C) E is not overdetermined, Then O does not cause E.
If
Thus it entails that, where the xs and O both make causal claims, if O isn’t causally relevant to the work of the xs (O and the xs are not
20
ordinary objects
operating sequentially (in a single causal chain) or cooperatively (each as a merely partial cause), then either they are rivals (so that one’s claim would rule out the other’s) or accepting both means ‘doubling up’ the causal story, in a case of worrying overdetermination. This is reasonable if the claims of O and the xs are independent—and it is by considering such restricted applications (i.e. Merricks and the mob) that we come to think that the causal principle is plausible. But this plausibility does not carry over when ‘O’ and ‘the xs’ are terms for entities whose causal claims are analytically interrelated. For even if one accepts, for example, that the baseball is not causally relevant to whether its relevantly arranged atoms do the shattering, the claim that the baseball does the shattering is analytically related to the claim that the atoms arranged baseballwise do the shattering, and so the baseball is causally related to the shattering. So accepting that the atoms arranged baseballwise caused the shattering does not show that the baseball did not cause the shattering (or that the shattering was double-caused) any more than saying that only battalions, batteries, and squadrons have marched past demonstrates either that no division marched past or that there was a double-supply of marching past (compare Ryle 1949/1984, 17–23). And so, if we would speak of real overdetermination, we should reject the causal principle taken in its completely general form, and thus reject the inference used in the overdetermination argument. But then we have no reason to think that either the very idea of baseballs is inconsistent or, as they would be epiphenomenal, the dictum ‘to be real is to have causal powers’ gives us reason to avoid postulating them. Nor do we have reason to think that accepting the causal work of baseballs requires us to accept objectionable systematic overdetermination. Nonetheless, some might suggest that, if all the causal work still can be accounted for by the activities of the atoms, without need to mention the baseball,15 we have no need to postulate the baseball, and so parsimony requires us to reject it (keeping just the atoms). I will return to this reformulated objection in discussing parsimony in chapter 9.
1.3 Other Apparent Redundancies I have thus far been responding directly to Merricks’s argument against composite inanimate objects based on their alleged casual redundancy. But even if you accept the above diagnosis, you may suspect that some
problems of causal redundancy
21
of the problems are peculiar to Merricks’s argument, and so may not be completely satisfied that there is no problem in the vicinity about causal redundancy. My diagnosis of the problem above relied on the idea that Merricks replaces claims about baseballs and other composite inanimate objects with claims about atoms arranged baseballwise. Given the proper understanding of ‘arranged baseballwise’, I have argued, the fact that atoms arranged baseballwise are causally relevant to a shattering analytically entails that a baseball is. But there are other apparent causal redundancies (regarding ordinary objects) where there aren’t such obvious and straightforward analytic entailments. That is, suppose that one does not speak of atoms arranged baseballwise (where this might be thought to include all the relevant intentions, social practices, etc. that friends of baseballs might consider necessary to the existence of baseballs) but only of atoms arranged in certain purely physical formations, causal relations, and so on (the same ones as the friend of baseballs would say a baseball was in if she limited herself to speaking of its situation purely in the language of physics). Such a situation does not seem to analytically entail that there is a baseball, let alone that the baseball causes anything since, by the folk ontologist’s own account, the existence of baseballs requires that there be people with certain intentions and practices. So one might worry that even if we can remove apparent redundancies between baseballs and atoms arranged baseballwise by appeal to the analytic entailments between claims involving these terms, other redundancies will still arise between, for example, a collection of atoms in a certain physical arrangement (or simply, a physical lump of stuff ) and a baseball. This line of worry provides good occasion to return to the idea of analytic entailments and to elucidate some of the different forms they may take. It is true that the mere presence of a physical lump of stuff (or atoms in certain merely physical formations) does not analytically entail the existence of a baseball, that claims about the two are not synonymous, and that baseballs cannot be reduced to such mere lumps of stuff without violating some of the basic beliefs and practices about baseballs that are constitutive of the proper use of the term.16 Nonetheless, there are other analytic entailments involved in two directions between claims about baseballs (or other constituted artifacts) and claims about the physical stuff that (according to the friend of ordinary objects) forms their constitutive basis. First, there are entailments in a top-down direction: One of the application conditions for the term ‘baseball’ is that there be a roughly spherical, medium-sized,
22
ordinary objects
unified physical lump of stuff. Thus the presence of a baseball analytically entails the presence of such a physical lump of stuff (although, as I will discuss in chapter 8, it leaves quite open what the precise microphysical nature of that stuff is—for example, whether it is atoms bonded in certain ways, a plenum of matter, and so on—provided it can fulfill the characteristic functions of baseballs). And any physical claims about the baseball analytically entail claims about the lump; more particularly, claims about the baseball’s causal relevance analytically entail claims about the lump’s. That is, a competent speaker of English who knows the meanings of the terms need know no more than that a baseball shattered a window to know that a certain medium-sized, roughly spherical lump of stuff shattered the window. Given such entailments, there again can be no competition between claims of the baseball and claims of the physical lump to causal relevance.17 Second, we can recover bottom-up analytic entailments of claims about baseballs by considering analytic entailments not just between individual claims but also involving sets of claims. A set of claims F may analytically entail a claim c, so even though claim j1, ‘There is a physical lump of stuff three inches in diameter, with the following physical properties . . . that shattered a window,’ does not on its own analytically entail that there is a baseball (much less that a baseball had any causal role), combined with other claims—j2, ‘That lump was arranged in a factory designed to make baseballs,’ j3, ‘As a result of its controlled design in manufacture, that lump is capable of performing the characteristic functions of baseballs,’ and so on—the set F of claims j1, j2, j3, and so on does analytically entail the claim c: that a baseball shattered the window. Again, a competent speaker and reasoner knowing the meanings of the terms and the truth of the members of F need investigate the world no further to infer that a baseball broke a window. It is a reasonable principle that not only is there no rivalry between (or doubling up of ) individual claims when one analytically entails the other; so similarly, there is no rivalry between claim c and any members (j1 . . . jn) of a set of claims F, where F is a minimal set of claims that analytically entails c. (By a ‘minimal’ set of claims I mean that each member (jn) of F is such that, were it to be removed from F it would not be the case that F analytically entails c.)18 There can be no rivalry between c and any member (jn) of F, since rivalry suggests that at most one can be true; but clearly both can be true, since the truth of (jn), in conjunction with certain other claims, entails that c is true.
problems of causal redundancy
23
To see this, consider some simpler cases: The claims (j1) that Jones bought a left-hand glove, (j2) that Jones bought a right-hand glove, and (j3) that those gloves match, taken together as a set of claims F, analytically entail c, that Jones bought a pair of gloves. Of the claims that are members of F, only (j1) and (j2) are even candidates for ‘competition’ with c in their claims to be what Jones purchased. But clearly there is no real competition between the claim (j1) that Jones bought a left-hand glove and the claim (c) that Jones bought a pair of gloves, nor is there any additional buying attested to in the former claim not described by the latter. Similarly, that there is a man living in a certain house does not itself analytically entail that a bachelor lives there. But the minimal set F of claims, (j1) that there is a man living in that house and (j2) that that man is unmarried, does analytically entail c that there is a bachelor living in that house. But again, the claim of the man to live in the house (j1) neither rivals nor adds to the claim c of the bachelor to live in that house—we need not conclude that either one of the claims is false, or that the man and the bachelor are roommates. So, similarly, since the claim that a certain physical lump of stuff (with various physical properties and capabilities) caused the window to break may serve as one of a minimal set of claims that jointly analytically entails that a baseball caused the window to break, the two claims are neither rivals nor (if jointly accepted) contributors to a kind of double-causation overdetermination. So while claims about mere physical lumps of stuff in certain arrangements certainly are not synonymous with claims about baseballs, nor are claims about one paraphrasable solely in terms of claims about the other (or reducible to such claims), there nonetheless seem to be analytic interrelations between the sets of claims that preclude there being any rivalry between (or additivity resulting from) their claims to causation. In general, when two claims j and c are such that either (1) j analytically entails c or (2) c analytically entails j or (3) j is a member of a minimal set of claims F such that F analytically entails c or (4) c is a member of a minimal set of claims C such that C analytically entails j, then I will speak of there being ‘analytic interrelations’ between claims j and c. These may not be the only sorts of philosophically interesting analytic interrelations between claims, but they will suffice for present purposes. If the presence of such analytic interrelations between claims prevents the claims from being additive or rivals, then we need not worry about the baseball’s causal action being redundant with respect to either simples-arranged-baseballwise or a mere physical lump of
24
ordinary objects
stuff. If successful, this shows that worries about causal redundancy, in whichever way they are couched, should not lead us to deny the existence of such ordinary objects. Uncovering the relevant analytic interrelations between claims about ordinary objects and claims about properly arranged simples will also play a key role below in defusing other appeals against ordinary objects such as those based on the idea that surely there is nothing ‘over and above’ the simples there (see x 4.1), those that arise from worries about colocation, and those arising from certain appeals to parsimony. It is also essential to the positive argument for accepting that there are ordinary objects (see x 9.3). Of course, for these arguments to be complete, the idea that there are analytic interrelations among our claims must be defended against some common objections—a task I will turn to in chapter 2.
1.4 The Difference between Baseballs and Minds Before leaving the topic of causal redundancy, it is worth examining what impact, if any, the sort of analysis provided above could have on debates about epiphenomenalism in philosophy of mind. For Merricks’s overdetermination argument bears a strong resemblance to causal exclusion arguments developed in the philosophy of mind— indeed such arguments have come to the center of philosophical debate more for their role in arguments in philosophy of mind than common sense ontology. Causal exclusion arguments in the philosophy of mind have been used, prominently by Jaegwon Kim (1993), to argue that the nonreductive physicalist about mental properties faces the dilemma of either accepting that mental properties are causally efficacious—and with it accepting widespread overdetermination and (if one allows them to have not just mental but physical effects) the violation of the causal closure of the physical—or denying them any causal powers, and thus treating them as epiphenomenal, in violation of the principle that ‘to be real is to have causal powers’. Merricks himself argues that despite its resemblance to causal exclusion arguments in philosophy of mind, his overdetermination argument does not apply to show that persons are causally redundant once human organisms are taken into account, nor does he think we should accept the parallel arguments in philosophy of mind. For (he holds) our conscious mental properties do not supervene on the natures of our parts, and do have causal powers not imputable to those
problems of causal redundancy
25
parts (2001, chap. 4). According to Merricks, mental properties have nonredundant causal impact on our behavior by causing it by way of causing the relevant physiological chains of events (110), in (admitted) violation of the causal closure of the physical (140), but without overdetermination (since there is but one causal chain, with the mental events causing the causally relevant physical events). Of course, many would resist Merricks’s way out, since it involves abandoning the causal closure of the physical, and so it is worth examining other possible escape routes. Can the argument utilized in defense of ordinary objects be pressed into service here as a way of showing that accepting that both, say, a mental property such as feeling pain and a physical property of the brain (e.g. C-fibers firing) may be considered causes for someone moving her hand, without that action being overdetermined in any worrying sense (and without accepting downward causation in violation of causal closure)? The way out of the problem in the case of the baseball and its atoms was to note that the causal principle loses its plausibility when we substitute in terms for entities whose causal claims are not (analytically) independent. Since the causal work of the atoms arranged baseballwise analytically entails the causal work of the baseball (I have argued) there is no rivalry between these claims, and positing ‘both’ does not mean accepting double the causation needed to explain the phenomenon. But the same does not seem to apply to the case of a mental property and its physiological basis. For the very fact that makes nonreductivism about the mental appealing is that there do not seem to be the relevant analytic entailments between claims about physiological states and those about mental states. A speaker who knows the meaning of the ordinary term ‘baseball’ and of the introduced term ‘atoms arranged baseballwise’ may infer that the baseball caused the shattering just in virtue of knowing that the atoms arranged baseballwise caused it. But, it seems, a competent user of mental terms like ‘pain’ can not infer that a pain caused the motion simply from knowledge of the meanings of such mental kind terms and knowledge that the C-fiber firings caused the motion.19 This apparent ‘logical gap’ between terms for physical and mental properties is, of course, precisely what makes possible explanatory gap arguments, zombie arguments, and knowledge arguments that provide the main support for nonreductive views in philosophy of mind.20 Thus there does seem to be a crucial difference between the case of mental properties versus purely physical properties, and the case of ordinary inanimate objects versus simples-arranged-ordinary-object-wise
26
ordinary objects
(where the latter is expressly defined in such a way as to include whatever intentional, cultural, contextual, or other properties play a role in our ordinary judgments about whether, for example, there is a table or baseball). That difference is, crucially, that in the latter but not the former case, there do seem to be the relevant analytic entailments. As David Chalmers writes, ‘‘conscious experience is almost unique in its failure to supervene logically’’ (1996, 71). One might wonder, however, whether the strategy of section 1.3 could be successfully wielded in the mental case. For we have seen that although the mere existence of a physical lump of a certain sort does not analytically entail that of a baseball, there still are analytic interrelations between claims about the two—including both top-down entailments (from claims about the baseball to claims about the lump) and bottom-up entailments (from claims about the lump combined with other claims about its intentional creation, social context, functional capabilities, and so on, to claims about the baseball). The first (topdown) direction does not seem promising for mental states, since (as Descartes argued) it does not seem that a competent speaker/reasoner who knows the existence of a pain or a thought and the meanings of terms may on that basis alone infer the existence of some physical state. What about the collective bottom-up direction? Can one combine statements about brain states with other statements to get a minimal set of claims from which one can analytically infer the existence of pains or other mental states? While this cannot be settled here, there are significant reasons for doubting that the move of section 1.3 will apply straightforwardly. To get bottom-up analytic entailments about baseballs that started from mere physical claims about a lump, we had to add claims about intentions, social practices, and the like.21 But one cannot simply do this to find analytic entailments about the mental that will save it from causal rivalry with the physical, for intentions, social practices, and the like are themselves mental (or dependent on the mental), and the issue here is precisely whether or not the mental taken as a whole is causally redundant with respect to the physical. But it is not at all obvious that from any set of merely physical claims one can get analytic entailments to any claims about consciousness, so it is doubtful whether the modified argument strategy of section 1.3 may be applied here. Whether or not there really is such a ‘logical gap’ between talk about the physical and talk about the mental is, of course, a matter of substantial controversy in the philosophy of mind, and one that, thankfully, we need not resolve here. It is enough to note that, given
problems of causal redundancy
27
that controversy, we are not entitled to assume that the above reply to causal exclusion arguments against ordinary objects can also be wielded in defense of mental states. In fact, the nonreductive physicalist cannot adopt the argument strategy above to demonstrate that mental states may be considered causally efficacious without accepting real overdetermination, without undermining the basis for the most central arguments in favor of her position. This is not, of course, to suggest that there is no good reply to the causal redundancy arguments available to the nonreductive physicalist about mental properties, only to say that such arguments must go at least somewhat differently from that proposed above for ordinary objects.22 One could attempt to develop a similar reply to that above by arguing that the causal principle is only plausible when the causal claims in question are not only analytically but also nomologically independent, or when the entities involved do not stand in a constitution relation, or . . .23 But this case would have to be made separately, and would take us well outside the current theme. So for now it is enough to note that accepting the above reply to overdetermination arguments against ordinary objects does not dictate what we say about causal exclusion arguments in philosophy of mind, though it may provide some suggestions worth investigating.
C c
two
analyticity and conceptual content
So far I have considered one sample problem alleged to plague those who accept ordinary objects: the problem of causal redundancy. In diagnosing the mistakes behind the causal redundancy argument, I have relied on the idea that some metaphysical principles do not properly apply when there are certain analytic interrelations between the terms substituted into them, and have argued that, for example, the causal principle should be rejected when claims about the causal involvement of one entity analytically entail claims about the causal involvement of the other.1 At this stage I will pause from directly considering arguments against ordinary objects to defend the idea that there are such analytic entailments. The analytic entailments at issue here are based not on logic alone (at least if one thinks of logic as topic-neutral, or as a mere syntactic system) but also on the meanings of the particular terms involved, so that, for example, the claim that I bought a house analytically entails the claim that I bought a building, and the claim that atoms arranged baseballwise caused a window to shatter analytically entails the claim that a baseball caused a window to shatter. It is in part constitutive of the meaning of ‘house’ that all houses are buildings, so that the truth of ‘X bought a house’ is sufficient for the truth of ‘X bought a building’: if we know the truth of the first, the meanings of the terms, and have reasoning abilities, we can infer the truth of the second claim on that basis alone. Thus the idea of analytic entailments I have been using relies on the idea that there are meanings, and relations among 28
analyticity and conceptual content
29
meanings, such that one may analytically infer the truth of one sentence merely from the truth of another and knowing the meanings of the terms involved (as well as having reasoning abilities).2 The idea that there are meanings with analytic entailments among them has, however, come under pressure on the one hand in virtue of Quine’s attacks on the very idea of meanings and analyticity and on the other in virtue of causal theories of reference. In this chapter I will confront both of these concerns head-on and suggest responses to the most prominent lines of objection. But the task here is not merely defensive. In defending the idea of analytic entailments, I will also argue for accepting a hybrid theory of reference that acknowledges that our general and singular nominative terms have certain sorts of frame-level conceptual content established by competent speakers. Accepting this sort of hybrid approach to reference not only provides us with the sorts of conceptual content needed to underpin analytic entailments, it also leads to some interesting consequences about metaphysical claims of identity, persistence, and modality, as I will argue in chapter 3. The views of reference, identity, and modality defended here involve enormous and controversial philosophical issues, each deserving book-length treatment in its own right, and the present arguments do not pretend to settle them. Nonetheless, here I will at least preliminarily defend them against some common objections, in hopes of avoiding peremptory dismissal. The chief virtues of these closely interrelated views will come out later (chapters 4–10), as we see their remarkable ability to provide a unified diagnosis of the mistakes behind all of the apparently diverse arguments against ordinary objects, and to help ground a coherent common sense view subject to none of the alleged problems. So whatever plausibility these views may acquire in these initial chapters, I hope, may be greatly enhanced by seeing their payoffs in the work below.
2.1 One Dogma of Quineans Quine’s (1953/2001) critique of the very idea of analyticity in ‘‘Two Dogmas of Empiricism’’3 opens by considering various ways of drawing the analytic/synthetic distinction and arguing that each is inadequate. One natural way to attempt to draw the distinction (Kant’s way) is to say that ‘‘a statement is analytic when it is true by virtue of meanings and independently of fact’’ (21). But talk of meanings as ‘‘obscure
30
ordinary objects
intermediary entities’’ may be replaced with more straightforward talk about the synonymies of pieces of language (22), and so Quine’s central attack on the idea of analyticities as applied to natural (as opposed to artificial) languages is based on arguing against the Fregean attempt to define ‘analyticity’ in terms of synonymy—that is, that analytic statements are those that are logically true (e.g. all unmarried men are unmarried) or that may be turned into logical truths by replacing terms with their synonyms (e.g. all bachelors are unmarried) (22–3). I should say at once that the kinds of analytic entailments that I discuss throughout this book are not typically based on synonymies; they may obtain (in one direction or the other) between terms that are not synonymous, and there may be analytic entailments even for terms that are not definable in terms of any set of necessary and sufficient conditions. (I will have more to say about the nature and grounds of these analytic entailments below.) So a defense of these sorts of analytic entailment need not rely on a defense of synonymy. Nonetheless, discovering a route to defend the existence of synonymies against Quine’s attacks will provide the route to a useful defense of the idea of analytic interrelations more generally, whether these involve direct synonymies or looser relations of analytic entailment. In ‘‘Two Dogmas,’’ Quine argues that the attempt to define ‘analyticity’ in terms of synonymy leaves the notion of synonymy in need of explication, and the natural approach to this is to treat synonymies as involving terms that are interchangeable by definition.4 However, Quine rejects the attempt to explicate analyticity in terms of synonymy, and synonymy in terms of definition. If by ‘definition’ we mean the sorts of statement appearing in dictionaries, this cannot be the analysis we sought, for, as Quine puts it: Clearly this would be to put the cart before the horse. The lexicographer is an empirical scientist, whose business is the recording of antecedent facts; and if he glosses ‘bachelor’ as ‘unmarried man’ it is because of his belief that there is a relation of synonymy between those forms, implicit in general or preferred usage prior to his own work. (1953/2001, 24)
The definitions appearing in dictionaries are merely descriptions of the synonymies already in an extant language, and so definitions, so understood, cannot ground synonymies. There is, however, another sense that might be attributed to the idea of something being true ‘by definition’. That is the sense in which a definition is not a description of an existing fact in a natural (or
analyticity and conceptual content
31
artificial) language, but rather a stipulation of a certain meaning (or use) of a novel term; that is, a ‘definition’ taken as an act or process of defining—of setting up the rules of use of a term—rather than as a descriptive report of linguistic fact. As Quine writes: There does, however, remain still an extreme sort of definition which does not hark back to prior synonymies at all: namely, the explicitly conventional introduction of novel notations for purposes of sheer abbreviation. Here the definiendum becomes synonymous with the definiens simply because it has been created expressly for the purpose of being synonymous with the definiens. Here we have a really transparent case of synonymy created by definition; would that all species of synonymy were as intelligible. For the rest, definition rests on synonymy rather than explaining it. (1953/2001, 25–6)
Quine later (in ‘‘Carnap and Logical Truth’’) will call this ‘‘legislative’’ definition, and accept that it ‘‘institutes truth by convention’’ (1966/ 1976, 118). The crucial point is that we must distinguish between definitions in the sense of descriptive reports versus definitions in the sense of performances or legislations (a distinction parallel to that between names being reported on in the phone book versus being established through baptismal ceremonies and the like, and to that between marriages being reported on and being entered into through different linguistic acts). While descriptive reports can only describe conventions previously established for the use of words, performances or legislations establish such conventions by fiat.5 In ‘‘Two Dogmas,’’ Quine only considers legislative cases in which new terms and their definitions are introduced merely ‘for purposes of abbreviation’ by a single individual who explicitly lays out the definition, and the only collective case he treats involves reports on conventions given by lexicographers. But, as Richard Creath writes, for Quine’s central interlocutor, Carnap, the central philosophic issue is not over the factual claims that the speaker adopts or that most speakers adopt such conventions. . . . Rather, what ought to be said is: ‘I propose that we adopt such and such convention.’ This is not a factual claim, and the sort of argument that would be appropriate on behalf of such a proposal would be [a] pragmatic argument, i.e., one which tries to show that such conventions would be useful. Quine’s alternatives simply miss the point. (2000, 334)6
This distinction between legislating and reporting on definitions is orthogonal to the distinction between individually versus collectively defining terms. Rules governing behavior may be established by a single
32
ordinary objects
individual (e.g. a ruler explicitly proclaiming a law) or collectively (e.g. by the differential punishment/reward practices of members of a society). So, for example, the rules of a game may be explicitly stipulated by an individual or committee, or they may be tacitly established by the practices of reward and punishment, praise and rebuke, engaged in by a single powerful individual or the group of ‘players’ at large. But all of these ways of establishing rules are quite distinct from such norms being reported on by anthropologists. So, similarly, rules governing proper use of an expression—when it may be successfully applied, what sorts of inferences we may make between and among claims of different types, and so on—may be established either individually or collectively (and in either case, this may be done either explicitly or tacitly). The difference is only one of degree between the case of a logician who explicitly introduces a novel term of her own invention along with its stipulated meaning and cases in which speakers of a common natural language ground the use of a new term by associating it with certain conditions that must be satisfied if it is to refer to anything at all, and if it is to be used to refer to the same thing again in other circumstances (and go on to rebuke, correct, or express bewilderment at those who might use the term improperly). In any case, however, this legislative sense of ‘definition’ as a process of speakers establishing rules of use for their terms must be distinguished from the sense of ‘definition’ in which the relevant rules are reported on by lexicographers.7 Granted, the cases in which it is clearest that a definition is being established and what precise definition is established are those cases in which a single individual performs a sort of ceremony saying ‘‘I shall hereby introduce term ‘T’ to mean . . .’’, thereby establishing the synonymy of ‘T’ with the construction in ellipses. Understanding the tacit and collective case and pinning down precise rules may be much more difficult than doing so in the case of an explicit introduction by an individual, but nonetheless if the latter case is (as Quine suggests) intelligible in principle as a way of grasping the basis of certain special cases of synonymies, the former should be intelligible as well as a way of understanding the basis for synonymies generally.8
2.2 Truth by Convention The view that synonymies and other analytic interrelations are in some sense grounded in definitions, where these are thought of as (perhaps tacit, collective) legislations of the rules of use for our terms, however,
analyticity and conceptual content
33
resembles classic conventionalism about analyticity—a view widely rejected. ‘Conventionalism,’ however, is a term used in many different ways (see section 3.2), and debates about conventionalism have more often focused on conventionalism about logical truths as such than about these derivative analytic truths, which Quine calls ‘‘truths by essential predication’’. I should emphasize again that I am only dealing here with analyticities that do depend on the particular meanings of words (or concepts) involved, not with pure logical truths as such, and the account I provide of analyticity (of this sort) is consistent with a variety of positions on purely logical truth (and inconsistent with some claims that go under the name ‘conventionalism’—see chapter 3).9 Nonetheless, it is worth examining part of the classical debate about logical truth by convention—since there Quine explicitly considers a position parallel to that I have suggested above. In ‘‘Truth by Convention,’’ Quine argues that those who would take all logical truths to be true by convention face the problem that, even if we can explicitly state certain basic conventions, from which the infinite range of logical truths may be inferred, ‘‘logic is needed for inferring logic from the conventions’’ (1966/1976, 104), so that we can’t take all logical truths to be established by accepting certain explicitly stated conventions. But he acknowledges that one way to circumvent the problem is for the conventionalist to allow that conventions may be tacitly and collectively established through behaviors, rather than being directly proclaimed in the form of explicit rules. The conventionalist may hold that we can adopt conventions through behavior, without first announcing them in words. . . . It may be held that the verbal formulation of conventions is no more a prerequisite of the adoption of the conventions than the writing of a grammar is a prerequisite of speech; that explicit exposition of conventions is merely one of many important uses of a completed language. So conceived, the conventions no longer involve us in a vicious regress. (105)
Of this tacit collective legislation of conventions, Quine immediately adds: ‘‘It must be conceded that this account accords well with what we actually do. We discourse without first phrasing the conventions; afterwards, in writings such as this, we formulate them to fit our behavior’’ (105). So, we might ask, what would be wrong with accepting that something like tacit, collective acts of definition for ordinary language may legislate that certain words are to be used interchangeably (or with
34
ordinary objects
other interrelations, e.g. that one is guaranteed to successfully apply if the other does), making synonymy and analyticity intelligible through understanding how the rules of proper use for our natural language are (tacitly, collectively) established? Quine follows up his endorsement of this move (in the case of logical truth) by explaining his reservations: On the other hand it is not clear wherein an adoption of the conventions, antecedently to their formulation, consists; such behavior is difficult to distinguish from that in which conventions are disregarded. . . . In dropping the attributes of deliberateness and explicitness from the notion of linguistic convention we risk depriving the latter of any explanatory force and reducing it to an idle label. We may wonder what one adds to the bare statement that the truths of logic and mathematics are a priori, or to the still barer behavioristic statement that they are firmly accepted, when he characterizes them as true by convention in such a sense. (1966/1976, 105–6)
In short, the bottom line of Quine’s complaint against this suggestion is that, at least in the tacit and collective cases, the supposed distinction between those uses of a term that are legislative and constitutive of its meaning and those that are not cannot be grounded in observable differences in overt linguistic behavior. Thus understood, his reason for rejecting this suggestion is much the same as the underlying basis of his objection to the analytic/synthetic distinction itself. For as Richard Creath has argued (2004, 49), the real problem Quine finds with the analytic/synthetic distinction is that behavioral criteria for applying the distinction are unavailable. The call for a behavioral criterion becomes still more evident in many of Quine’s later writings, for example, in ‘‘Truth by Convention’’ and ‘‘Carnap and Logical Truth’’ (1966/1976, 105, 129; see Gibson 2004a, 185). In his own ‘‘Two Dogmas in Retrospect,’’ Quine writes explicitly: ‘‘Repudiation of the first dogma, analyticity, is insistence on empirical criteria for semantic concepts: for synonymy, meaning’’ (1991, 272). Quine’s rejection of analyticity has been so widely taken on board that it is difficult to use the term ‘analyticity’ without provoking skeptical glares. But if this is the ultimate foundation for rejecting the very idea of analyticity and all that goes along with it, it warrants reevaluation. Indeed one lesson of this book will be just how great a difference the acceptance or rejection of analyticity makes throughout metaphysics, so that the move to reject analyticity should not be taken
analyticity and conceptual content
35
lightly, nor on the authority of Quine or even of the bulk of a generation of analytic philosophers who followed him. First, note that if this is the true foundation of the move to reject analyticity, those who do not embrace Quine’s behaviorism have no need to follow him in rejecting the analytic/synthetic distinction, or rejecting the possibility of understanding this in terms of tacit and collective legislation of the proper rules of language use (as distinct from descriptions of extant linguistic rules). A first-pass way to characterize the former would involve appeal to a complex of factors naturally appealed to in explaining the differences in force of various utterances, including speakers’ intentions and norms to which they hold themselves and others. Such first-pass analyses clearly use terms that the behaviorist would reject unless they could be given a behavioral analysis. Whether or not such behavioral analyses could be given of course runs far beyond the scope of this book, and is beside the point for those who reject the behavioristic demand in the first place. Second, it is worth noting that the prospects for some behaviorally relevant criteria to distinguish analytic and synthetic statements, or legislative versus descriptive definitions, are not nearly as bleak as Quine’s brief remarks in these early papers suggest. As Strawson and Grice (1956) argue, since the analytic/synthetic distinction not only is in common philosophical use (where those who use it consistently apply it in the same way to new cases) but also is interdefinable with the use of ordinary expressions such as ‘means the same as’ (applied with equal consistency by ordinary speakers of English), there is a strong presumption in favor of there being some distinction marked by these differential uses (146–8).10 And, as Strawson and Grice again suggest, we do respond differently to a speaker who misuses a term than to one who simply asserts something false or highly implausible: in the latter case, we might deny the claim, refuse to believe it, while in the former we express bafflement or deny that the person’s remarks had sense (150–1), or (e.g. in interacting with foreigners) suggest that a word has been mislearned. As Quine later acknowledges, ‘‘often in talking with a foreigner we recognize some impasse as due to his having mislearned an English word rather than to his having a bizarre view of the subject matter. This is a bit of practical psychology at which we are all adept’’ (1991, 270). In his later writings (1974, 1990, 1991) Quine himself suggests a behavioral criterion for distinguishing analytic truths: that analytic sentences are those that speakers learn the truth of just in coming to
36
ordinary objects
understand the relevant words (Creath 2004, 59). This, of course, is rather parallel to the criterion for analytic entailments I suggested above: that A analytically entails B just in case competent reasoners may infer the truth of B simply by knowing the truth of A and understanding the relevant terms. Ultimately, Quine explicitly acknowledges that ‘‘analyticity undeniably has a place at the commonsense level’’ (1991, 270) and that there are, especially in natural language, some clear cases of analyticities, though he still denies that even his proposed behavioral criterion can provide any general rule for demarcating analytic versus synthetic sentences across the whole domain of sentences (271). Summing up his reflective position, Quine writes: In short, I recognize the notion of analyticity in its obvious and useful but epistemologically insignificant applications. The needs that Carnap felt for the notion in connection with mathematical truth are better met through holism. Beyond its manifest cases I find analyticity less help than hindrance. It begets an uncritical notion of meaning, or synonymy, that can induce a false sense of understanding. (271)
Quine’s central concerns of course are quite different from ours, and I will not here touch on the notion of mathematical truth. One moral of the work to follow, though, is just how significant the notion of analyticity—broadened to cover analytic entailments, not just synonymies—in its ‘obvious and useful’ applications in natural language is, if we turn our concerns to understanding common sense ontology and the variety of arguments against it. Much more, obviously, needs to be said to provide a fuller account of how analyticities may be founded through collectively establishing certain rules (or conventions) of proper use of our language and terms. The attempt to do so would take us far afield from our main theme to a general study of the source of linguistic meanings.11 So this is not the place for it; here we have metaphysical fish to fry.12 Thus far, I have merely tried to suggest a way of grounding synonymies and other analyticities that takes the same direction as one Quine acknowledges to be plausible, and is enough to suggest that the whole idea of there being analytic statements, and other relations among claims (such as what I have been calling ‘analytic entailments’) that hold in virtue of the meanings of the natural language terms involved is not bankrupt or illusory, as many have taken the early arguments of Quine to show. It is also, I have argued, not implausible to think even that behavioral differences may mark these distinctions,
analyticity and conceptual content
37
subtle and diffuse though they may be. It is thus, I hope, enough to dispel or at least suspend the skeptical glares that come with mention of analyticity, while we see what the payoff may be in terms of unraveling a variety of arguments against ordinary objects and formulating a defensible common sense ontology. As I will argue in section 2.3, the proposed account of the grounds of analyticities also fits naturally with the claim that, even in the best case scenarios for direct reference theories, there is some conceptual content to the meanings of terms that is established stipulatively in attempts to ground the reference of a term. It is also worth noting, as Strawson and Grice (1956, 156–7) do, that accepting that there is an analytic/synthetic distinction does not preclude us from retaining Quine’s positive insight in ‘‘Two Dogmas’’ that confirmation is holistic, so that anything is in principle open to revision.13 Revisions in the case of analytic statements must merely be considered to be revisions based on proposed reconfigurations of our conceptual system rather than revisions based on direct conflict with experience. We do not discover that they are false, based on some kind of experience, but we may decide to alter them, in order to make the system as a whole work better. ( Just as officials do not discover the rules of NCAA basketball to be false, but may decide to alter them in order to make the games more exciting, efficient, etc.—and may decide to do so on the basis of empirical results, e.g. that show fans to be turning off the television after a point.) So this understanding of analyticity is quite consistent with the idea that empirical results may make conceptual changes of these sorts desirable in order to simplify the system as a whole, and the case for holism gives us no decisive reason to prefer the Quinean view over a more Carnapian approach. In sum, then, on this model synonymies may come to hold either because one term is (whether tacitly or explicitly, individually or collectively) defined in terms of another, or because two or more terms are defined in the same way. In either case, definitions—in the sense relevant to explicating synonymies—should be understood as performatively establishing relations among meanings rather than reporting on preexisting relations of synonymy.14 More generally, analytic interrelations other than synonymy may come to hold in virtue of various (explicitly stipulated or implicitly established) rules for proper use of terms, for example, that this term can be correctly applied only if that one applies, or is guaranteed to be applicable if that one is (it is these rather than strict synonymies that play the central role in most cases of analytic entailment).
38
ordinary objects
2.3 Causal Theories of Reference and the Qua Problem The idea that our terms have some conceptual content has also been threatened by causal theories of reference. For causal theories of reference, at least in their extreme or pure form, have led many to hope that the meanings of terms can be understood as determined not at all by the concepts of competent speakers, but rather purely by a real causal relationship to things in the world (an individual in the case of names, or a sample in the case of kind terms). If a pure theory along these lines could be made to work, the idea that there are analytic relations among meanings of different terms such as those I have invoked above might seem entirely wrongheaded, for on this view meanings are not discoverable by conceptual analysis at all, nor are there conceptual relations among meanings that can ground the kind of analytic entailments I have relied on. But pure causal theories face two well-known problems: the qua problem and the problem of handling nonexistence claims. I will argue that both of these problems may be resolved by moderating pure causal theories, holding that reference is only unambiguously established to the extent that our nominative terms are associated with a high-level conceptual content establishing what category of entity is to be referred to by the term, if it refers at all.15 Accepting this hybrid approach to reference in turn has important consequences for preserving the idea of analytic entailments, and for understanding the proper methods and limits of metaphysics, all of which I will return to below. Pure causal theories may seem plausible enough as long as we implicitly confine our discussion, for example, to names of people. But the qua problem arises once we acknowledge that there are terms of many different sorts that at least purport to refer to many different sorts of things, for example, artifacts, lumps of matter, spatial or temporal parts of objects, events, and so on. For those attempting to ground the reference of a new singular term, it will be radically indeterminate what the term refers to (or even whether or not it refers) unless they have some very basic concept of what sort of thing (broadly speaking) they intend to refer to, if the reference grounding is to succeed.16 While the case is most often made with proper names, the same qua problem clearly arises for any attempts at referring to individuals— whether by way of names, demonstratives, or pronouns (see Dummett 1973/1981, 570–1; Wiggins 2001, 52–3)—and arises similarly for general nouns.
analyticity and conceptual content
39
Note that raising the qua problem does not require any special ontological assumptions: we need not assume that there are entities of all the possible categories to get the qua problem going, only that speakers may at least intend to refer to items of various sorts. For the relevant ambiguities arise regarding not merely which object (if any) the term refers to but also whether or not a given singular term refers at all.17 There is no doubt that we at least intend our singular terms to refer to different sorts of things—that is why the eliminativist view that we often fail in such attempts is supposed to be surprising. Even those who accept a minimal ontology, say, accepting the existence of only a single ‘level’ of basic objects capable of being referred to, must have some way of distinguishing those singular terms that do successfully refer (e.g. ‘lepton a’) from those that (according to them) do not (e.g. ‘Big Ben’).18 Devitt and Sterelny argue that a hybrid descriptive/causal theory is needed to avoid the qua problem and explain how the reference of our terms may be grounded, and suggest that those who would ground the reference of a term must (consciously or unconsciously) conceive of the item to be referred to (causing the experience) ‘‘under some general categorial term like ‘animal’ or ‘material object.’. . . The grounding will fail if the cause of the perceptual experience does not fit the general categorial term used to conceptualize it’’ (Devitt and Sterelny 1999, 80). They say little about what is involved in a categorial term, but reexamining the particular difficulties for pure causal theories of reference can provide further insight into what sort of categorial term or conception is required to provide the needed disambiguation, and thus about the particular form a hybrid theory of reference should take. To successfully disambiguate whether or not the reference of a term is grounded, and if so to what, it seems that nominative terms must be associated with a sortal or, more generally, categorial concept that does at least two things. First, it must establish certain very basic conditions under which the attempted grounding would or would not be successful in establishing reference. So, for example, if I attempt to ground the name ‘Orky’ as the name for an animal (swimming near my boat), my attempt to ground the reference may fail if all that has perturbed the water near my boat is a large clump of seaweed, or a strange event in the ocean current causing an unusual wave. Call these the ‘frame-level application conditions’ for the term—‘frame-level’ since they involve conditions that are conceptually relevant to whether or not reference is established, not all the conditions that may be
40
ordinary objects
empirically discovered as relevant.19 In fact, the frame-level application conditions may defer to further real-world conditions to be fleshed out, for example, ‘Orky’ may only apply if there is life here, but other necessary conditions such as the presence of certain sorts of chemical structures necessary for life may only be discoverable empirically, not read off the frame-level application conditions of the term. Some associated conceptual content is also needed to supply framelevel coapplication conditions for our nominative terms—that is, rules that (supposing the term to have been successfully applied) specify under what conditions the term would be applied again to one and the same entity. (For singular terms—unlike sortals—these are of course just the same as conditions under which it could be successfully applied again at all.) For it is only this that disambiguates, for example, the attempt to refer to an animal from the attempt to refer to a mass of cells, or a time-slice of an animal, and so on. These coapplication conditions (like the application conditions) may incorporate a great deal of deference to the world (i.e. it is the same animal only if death is not undergone, but what empirical conditions establish death may be left for discovery), but nonetheless, at the basic, frame level, this conceptual content establishes what basic conditions are and are not relevant to when the term may be reapplied to one and the same entity.20 It is important to distinguish criteria of application from criteria of coapplication. Nonsortal terms, such as adjectives, may supply only criteria of application but no criteria of coapplication (Dummett 1973/1981, 75). Thus, to use Dummett’s example, ‘smooth’ has criteria of application (there are situations in which it is and is not appropriately applied) but comes with no criteria that enable speakers to say whether this smooth is the same as that. Another important reason for distinguishing criteria of application and coapplication is that there may be cases in which conditions for successfully applying two terms are the same (or at least in which they share a certain set of sufficient conditions) but the conditions diverge regarding when (supposing it to have been successfully applied) each term may be reapplied to one and the same thing. Thus, for example, as Dummett (74–5) points out, the conditions under which the term ‘book’ (in the sense of a physical copy) may be correctly applied are the same as the conditions under which the term ‘book’ (in the sense of a literary work) may be applied. But the conditions under which we may refer again to one and the same copy versus to one and the same work are quite distinct—the first, for example, demanding spatio-temporal continuity while the second does not. Much the same can be said for aggregates of trees versus
analyticity and conceptual content
41
copses (to borrow an example from Wiggins 2001, 52–3) or for performances and works of music. Since criteria of application and coapplication are relatively independent in these ways, we may ask whether a sortal term’s application conditions are fulfilled independently of inquiring after the conditions under which, supposing it to have been successfully applied twice, it would have been applied to one and the same thing each time. (The application conditions for a sortal term are mastered by learning under what conditions that general term is to be applied or refused, and may be learned apart from learning the conditions under which the term may be twice applied to one and the same entity.21) As a result, it seems we also must allow that, at least in some fundamental cases, the application conditions for our terms need not take the form of requiring that there be some individual/thing/object, such that it has the following structure or properties. Instead, we should think of application conditions as being fulfilled or unfulfilled by the way the world is; if they are fulfilled, the term is guaranteed to apply, but application conditions alone do not determine what individual(s) it refers to, or even if it refers to things at all (since the term may be a stuff term rather than a singular or sortal term). That disambiguation comes only with the addition (or failure of addition) of the coapplication conditions that yield identity conditions for the thing(s) (if any) referred to. The associated application and coapplication conditions may be more or less determinate, more or less vague, so we should not take association with such conditions as an on/off matter, with which reference is determinate and without which it is radically indeterminate. For while all sortal terms (including categorial terms) are alike in providing application and coapplication conditions, sortal terms differ among themselves in terms of specificity, and may be arranged in hierarchies of increasing generality. What I have been calling ‘categorial’ terms (and concepts) are just highly general sortal terms (and concepts).22 The application of a sortal S1 to any entity x may analytically entail that another sortal, S2, also applies to x, as, for example, the application of ‘dog’ to Fido guarantees the application of ‘animal’ to Fido. Where this occurs, I will say that S2 is a ‘genus-sortal’ with respect to S1, and S1 is a ‘species-sortal’ with respect to S2. (The two sortals may, nonetheless, have different application conditions, as, for example, something may satisfy the application conditions for ‘animal’ without satisfying those for ‘dog’ though the reverse is not the case.)
42
ordinary objects
We can then arrange sortal terms in grouped hierarchies, calling high-level genus-sortals ‘categorial terms’, and the high-level sorts ‘categories’, noting that different sortal terms may be of the same category (e.g. ‘dog’, ‘cat’, and ‘horse’ all are of the category ‘animal’; ‘fork’, ‘table’, and ‘shoe’ are all of the category ‘artifact’, etc.). More specifically, two sortal terms A and B are of the same category if and only if either: 1. B is a genus sortal with respect to A, or: 2. A is a genus sortal with respect to B, or: 3. There is some C, such that C is a genus sortal with respect to A, and C is a genus sortal with respect to B.
So categorial terms, like all sortals, have application conditions and coapplication conditions,23 and the application conditions for the categorial term are guaranteed to be fulfilled provided those for any of its species-sortals are. In sum, while our terms may be associated with conditions of application and coapplication that vary greatly in specificity, arguments from the qua problem suggest that reference to individuals (whether via singular or sortal terms) is determinate only to the extent that the term is associated with determinate application conditions and coapplication conditions, via association—at a minimum—with a certain sort or category of entity to be referred to. As a result, attempts to refer using a name, demonstrative, or other singular term cannot be disambiguated simply by conjoining these terms with a general term such as ‘individual’, ‘object’, or ‘thing’, if these general terms are not associated with the application and coapplication conditions needed to provide disambiguation.24 To avoid some potential misunderstandings and objections, however, more should be said here about what is involved in associating a term with a category of entity to be referred to, and who must possess the relevant categorial conception. So far I have focused on the case of those who ground the reference of terms, insisting that whether and if so to what their terms refer is only disambiguated to the extent that they possess a high-level categorial conception, but what of other language users? Can they use the relevant terms successfully without sharing the categorial concept grounders used to establish reference (either because they lack any categorial conception at all, or have the wrong one)? Devitt and Sterelny are clear that they are arguing only for a hybrid theory of reference grounding, retaining a purely causal theory of reference borrowing: ‘‘[A speaker] can pick up a name on a
analyticity and conceptual content
43
very slender basis, wrongly inferring all sorts of things about its bearer. Perhaps it names a university yet she believes it to name a person, a cat or a river. She is linked into the causal network for the name and so there seems to be no good reason to deny that she uses the name to designate the university’’ (1999, 79). We can remain neutral here about whether or not later speakers lacking any categorial conception, or possessing the wrong one, can still use terms to designate the relevant entities by borrowing their reference from others. The important point from the current perspective is that we can still distinguish those speakers who are competent users of the term, who grasp the frame-level application and coapplication conditions (enabling them to say, of various actual and hypothetical situations, whether the term would apply or reapply) from those who, at best, refer purely parasitically. Only the former may be said to grasp the meaning of the term in the sense relevant for grasping analytic entailments. We have good reason to think that the framelevel categorial conception must be passed on to a reasonable number of later speakers who may be indoctrinated into the proper rules of use for the term and help avoid reference shifts. For those who mistake the basic frame-level conditions that establish the category of the item or kind to be referred to (e.g. who take ‘Aleph Zero’ to name a person, or ‘tenacity’ to be a place-term) are a prime source of reference shifts.25 Accepting that these terms have determinate reference only to the extent that those who ground the term’s reference (and later speakers who are competent users of the term) associate them with a disambiguating categorial conception clearly leaves room for the sorts of speakers’ ignorance and error that originally helped motivate adopting causal theories. It is, for example, still true on this view that competent users of the name ‘Go¨del’ may be completely ignorant or mistaken about any facts about Go¨del’s life, characteristics, or achievements, provided they at least take the term to be a name for a person and not, for example, for a kind of undergarment or lumbering practice. Nor does the need for competent users of the relevant term to associate it with frame-level application and coapplication conditions mean that they must be able to articulate what the frame-level application and coapplication conditions for the relevant categorial term are. Disambiguation requires associating the singular term with a relevant category of thing (e.g. with being a name for an animal, a nightclub, or a holiday), where this involves a tacit understanding of what it takes for there to be such things, and when we can properly say they
44
ordinary objects
are the same. Such a tacit understanding requires the ability (on the whole) to apply and reapply terms of that sort properly, to evaluate whether or not they would be properly applied or reapplied in various situations, to rebuke others for misuse, and so on—not the ability to explicitly recite the relevant conditions; just as understanding the rules of soccer requires ability to play properly, rebuke others when they break the rules, and so on—not an ability to recite the rules in discursive form. Indeed a term may have application conditions even if there is no statement of these conditions available in other terms at all: for a term to have application conditions is for competent speakers to be able to evaluate, with respect to various hypothetical situations (ways the actual world could turn out to be), whether or not the term would apply—not for competent speakers, or philosophers, to be able to provide an explicit statement in other terms of what those conditions are.26 Much the same goes for coapplication conditions: speakers must be able to evaluate whether or not the term would be applied to one and the same entity in various hypothetical circumstances, not to explicitly state identity conditions for the entity (if any) referred to. Nor must such conditions provide a complete set of necessary and sufficient conditions for a term’s application and coapplication; instead, they may, for example, only involve various sufficient conditions, and may be highly incomplete.27
2.4 The Basis of Analytic Entailments If, to avoid the difficulties of the qua problem, we accept that our singular and sortal terms come with at least this sort of minimal conceptual content, we have the basis we need to ground the kinds of analytic entailment I rely on in diagnosing problems with various arguments against ordinary objects. Given the frame-level application conditions associated with singular and sortal terms, for any terms ‘p’ and ‘q,’ where the application conditions for ‘p’ are also sufficient conditions for ‘q’ to apply, claims such as ‘ (A) p exists’ analytically entail claims that ‘ (a) q exists,’ for example, the application conditions for ‘house’ in a situation are sufficient to ensure the application of ‘building,’ so ‘There is a house’ analytically entails ‘There is a building.’ While this guarantees that there are analytic interrelations between claims that there is something of a given sort and that there is something of the subsuming category, such analytic entailments hold generally for any terms with the relevant relation between their
analyticity and conceptual content
45
application conditions, whether or not these are terms of the same category. In other cases, the application of ‘p’ may be a necessary condition for the application of ‘q,’ as, for example, the application of ‘physical lump of stuff’ is necessary for the application of ‘baseball,’ so that the claim ‘There is a baseball’ analytically entails ‘There is a lump of stuff’ (since the former could not apply unless the latter also applied). In still other cases, the application of ‘p,’ taken alone, may not be sufficient for the application of ‘q,’ but ‘p exists’ might nonetheless form part of a minimal set of statements that does ensure that the application conditions for ‘q’ are fulfilled and so does analytically entail that ‘ (a) q exists,’ and so on. As mentioned above, such relations among meanings are often much looser relations than synonymy, so to claim that there are analytic entailments between sentences involving terms ‘p’ and ‘q’ is not to claim that ‘p’ and ‘q’ are synonymous, nor that either may be paraphrased without loss in terms of the other, and however close the application conditions of terms ‘p’ and ‘q’ may be, ps may still not be reducible to qs on account of the different coapplication conditions associated with the corresponding terms (see chapters 3 and 11). The above are just a few examples of ways in which the conceptual content associated with our terms can form the basis for analytic interrelations among our sentences. Providing details of the multifarious ways in which elements of conceptual content of terms (combined with logic) can lead to analytic interrelations of our sentences would be an enormous undertaking, and cannot be done here. So below (as above) in making judgments about analytic entailments, I will mostly rely on the initial criterion of what a competent speaker/ reasoner who grasps the meanings of the terms, and of the truth of a first sentence (or set of sentences), could legitimately infer from that basis alone.
2.5 The Problem of Nonexistence Claims Another well-known problem facing direct reference theories is handling nonexistence claims—and the key to solving the problem again lies in rejecting a purely causal theory, and accepting a hybrid theory holding that well-formed singular and sortal terms have at least a high-level category specifying conceptual content that plays a role in determining the truth-conditions of sentences involving them.
46
ordinary objects
Singular claims of nonexistence such as ‘Atlantis doesn’t exist,’ ‘Sherlock Holmes doesn’t exist,’ and the like (at least as uttered in some contexts) seem pretheoretically to be meaningful and true. But if we take them to be simple subject-predicate sentences, it seems that direct reference theories must hold that for them to be meaningful at all, the singular terms in question must refer to some entity. But then it seems all singular nonexistence statements must be false if they are meaningful, for the entity in question (Holmes, Atlantis) must exist if it is to be referred to and its existence is to be meaningfully denied. One common response to this problem for direct reference theorists has been to take a metalinguistic approach to nonexistence (and existence) statements.28 So according to Keith Donnellan’s metalinguistic account, for example, direct reference theorists should treat nonexistence claims as having the following truth-conditions:29 If N is a proper name that has been used in predicative statements with the intention to refer to some individual, then ‘N does not exist’ is true if and only if the history of those uses ends in a block. (Donnellan 1974, 25)
A name use chain ends in a ‘block’ when, for example, it ends with the introduction of a name in a work of fiction, via a mistake, an act of imagination, and so on (23–4). But I have argued elsewhere (2003c) that what counts as a ‘block’ cannot be defined so generically; what counts as a block depends on what sort of individual speakers in the relevant name use chain intended to refer to. So, for example, nonexistence claims involving fictional names may be made in a variety of contexts. The usual context presupposed when we pretheoretically think of claims such as ‘Sherlock Holmes doesn’t exist’ as true is the context in which someone has mistakenly supposed ‘Holmes’ to be a name for a person, perhaps attempting to use it in contexts like ‘I would like to hire Holmes to solve this crime’—in this context, showing someone that the chain from which she has learned the name ends in a work of literature rather than a baptism of a person is enough to show the speaker her mistake. In the context of discussions of works of literature, however, existence and nonexistence claims involving fictional names may be used to affirm or deny the existence of certain fictional characters, as we might say, for example, that the character Holmes does exist, while the character of Holmes’s psychiatrist Schmidt does not exist. Donnellan’s simple metalinguistic approach to nonexistence statements cannot preserve this kind of distinction, since both names (‘Holmes’ and ‘Schmidt’) end in what he terms a ‘block’. Yet intuitively, in such
analyticity and conceptual content
47
literary contexts, tracing ‘Holmes’ back to a work of literature confirms that it does refer (to a fictional character) while the fact that ‘Schmidt’ cannot be traced to a work of literature, but only to a mistake, shows the latter name use chain (if intended to refer to a fictional character) to end in a block. The issues here aren’t unique to fictional discourse, but arise wherever there may be category ambiguities in our terms. A sentence like ‘Atlantis doesn’t exist’ may be true if speakers intended to refer to an ancient city, and the name use chain ends in a creatively decorated nightclub mistaken by a stoned clubber for an ancient city. But if the users and grounder intended to refer to a nightclub, no mistake would have been made; the usage chain would not be blocked but properly grounded, and the nonexistence claim (presupposing uses with that intention) would be false. Or again consider a claim such as ‘The great blue whale no longer exists.’ This may be true if uttered in response to a child’s request to go see the great blue whale in the Museum of Natural History, when their taxidermed specimen was recently destroyed in a fire. But it is false if it is uttered in response to claims that great blue whales are the largest living creatures, based on the mistaken view that the species is extinct. In short, the truth-conditions for nonexistence (and existence) claims seem to vary systematically depending on what category of entity prior speakers intended to refer to. The need to accept that what counts as a block (and thus what the truth-conditions are for nonexistence claims) varies according to the category of entity prior speakers intended to refer to can even be seen in Donnellan’s original formulation. For blocks were supposed to occur where the name use chain ends in a mistake, but what counts as a mistake by prior speakers clearly depends on what category of entity they meant to refer to: If the prior speakers intended to refer to a nightclub with the name ‘Atlantis,’ they made no mistake; if they intended to refer to a city, a mistake was made somewhere along the line. This gives us reason to think that the truth-value of claims of existence and nonexistence depends on the category of entity speakers intend to refer to, and cannot be evaluated except with respect to some presupposed category or other. Where it is ambiguous what category of entity was the presupposed referent, existence claims may not be truth-evaluable. If we accept that the intentions of speakers regarding what broad ontological sort of thing their term should refer to are relevant to determining whether or not their use of a term refers, we can see how to generalize Donnellan’s suggestion to avoid the above problem in handling nonexistence claims:
48
ordinary objects If N is a proper name that has been used in predicative statements with the intention to refer to some individual of category C, then ‘N does not exist’ is true if and only if the history of those uses does not lead back to a grounding in which the application conditions associated with C are met.
Thus whether or not a claim like ‘Holmes doesn’t exist’ or ‘Atlantis doesn’t exist’ is true depends on what category of entity is the intended referent in the presupposed range of prior predicative statements—if the presupposed category is a person or city, the chain must end in a mistake and thus a block, making the claim true; if on the other hand the category presupposed is that of a fictional character or nightclub, no similar mistake is being made, the necessary conditions are met, and the nonexistence claims would be false (while the corresponding existence claims would be true) (see my 2003c, forthcoming d). The issues again run in parallel for sortal terms. Claims of nonexistence such as ‘Unicorns don’t exist’ seem to presuppose a real kind of causal connection for the sortal ‘unicorn’ to be meaningful, thus making such claims apparently false if meaningful. Such difficulties can similarly be overcome by treating sortal claims of existence and nonexistence as presupposing a prior range of predicative statements intending to refer to entities of a certain category, so that ‘Ks don’t exist’ is true just in case the history of those uses does not lead back to a grounding in which the relevant application conditions for terms of that category are met. Again, where it is indeterminate what category of entity was the intended referent in the presupposed uses, the claim may not be truth-evaluable. In short, we can provide a natural solution to both the qua problem and to the problems of handling claims of existence and nonexistence by mitigating pure causal theories of reference, accepting that it is determinate whether our terms refer, and if so to what, only to the extent that the terms are associated with disambiguating application and coapplication conditions that establish what it will take for reference to be grounded and under what conditions the term may be used again to refer to the same object.
2.6 Objections to Hybrid Theories of Reference Hard-line causal theorists of reference, however, have objected to such hybrid approaches on grounds that (like pure descriptivist theories)
analyticity and conceptual content
49
they allegedly cannot sufficiently account for possible error and empirical discovery. First, it is often said, such theories cannot account for the fact that, even if what I have called the frame-level application conditions for our terms are not met, we would still often allow that the terms refer, concluding only that we were wrong about what it would take for the term to refer. Thus, in criticism of such hybrid theories, Richard Miller writes: It is open to some of the same objections that have been directed against purely descriptive theories of reference. If kangaroos had turned out not to be animals at all but rather Putnam’s Martian robots then, on the current hypothesis, ‘kangaroo’ couldn’t refer to kangaroos since kangaroos turned out not to be a species. But of course we do refer to kangaroos whether or not they turn out to be animals. (1992, 427)
The evidence against the hybrid approach is thus supposed to lie in the intuition that we would allow that ‘kangaroo’ could refer, even if we discovered that what was hopping across the hills was a group of Martian robots, showing that, even where there are such associated frame-level application conditions, these are not stipulatively established conditions for reference to succeed, but instead are fallible assumptions ( just as externalist thought experiments suggested that other aspects of our associated concepts were fallible); what it takes for the term to refer is then supposed to be entirely externally determined. Note, however, that in other cases it seems that we wouldn’t say that a term referred if it turned out that its basic application conditions weren’t met—for example, if the name ‘Orky’, as above, is introduced as an animal name, it seems an even clearer intuition that we would say that there was no Orky (not that Orky turns out to be a patch of seaweed) if there was only a patch of seaweed by my boat when I attempted to ground the reference of the name. Even in quite similar cases to Miller’s, our intuitions can easily go the other way. If ornithologists coin the term ‘Key sparrow’ to name a new race of sparrows apparently discovered to be living in the Florida Keys, only later to find that all of the supposed exemplars observed were sophisticated animatronics planted by a glory-hungry birdwatcher, it seems we would say that there are no Key sparrows (since the things observed were not birds at all), not that it turns out that Key sparrows are little robots. So what lesson should we draw from these intuitions? The right lesson seems to be not that reference survives any failure of associated basic application conditions (so that even the most basic application conditions must be empirically discovered), but rather that
50
ordinary objects
where such conditions fail, we have to make a decision about what to do with the term based on our surrounding practices and concerns. In some cases (e.g. ‘Orky’ and ‘Key sparrow’, where our interests lay in there being animals of a certain kind at a certain place) we might retain the term’s original meaning but accept that the term failed to refer. In those other cases in which it seems we might continue to use the term despite the fact that the most basic, frame-level application conditions were not fulfilled, it seems reasonable to hold that this is the result of a semantic decision to alter the basic meaning-content of the term to adopt it to a new use (just as we may adopt other nonfunctioning artifacts to new uses). It is easy to see how such changes might be affected, even in the more extreme cases—at first we might naturally use the term implicitly as scare-quoted. (‘Your friend ‘‘Orky’’ seems to be coming apart! Look, now a school of fish is swimming right through him!’ ‘These so-called Key sparrows are just robots.’) Ultimately (when there is no further danger of confusion) we might simply adopt the new term (which after all is handy and familiar) in a way that replaces the earlier requirements with new ones reflecting this discovery.30 So even where we do or would continue to apply a term despite the fact that the presumed frame-level application conditions turn out not to be fulfilled, that need not be taken as a sign that these were never part of the meaning of the term. A similar line of objection holds that hybrid theories like that proposed above cannot, in general, account for the epistemic role of empirical discovery, either in discovering to what kind the things to which we refer belong, or regarding what the basic categories of things are (Schroeter 2004). The first point arises from intuitions some have that, driving past a field, I might seek to name that thing (sticking up) ‘Klut’ without any category-specific intentions. Should curiosity get the better of me, I can park the car, wander out a few hundred meters, and discover whether in fact Klut is a kangaroo bent over eating, or a pile of hay, or an irrigation device, and so on. Thus (the intuitions go) I can refer to Klut without any categorial intentions, and empirically discover the basic category to which it belongs (e.g. animal or artifact). It is true enough that we can discover, for example, whether there is (or is not) a kangaroo or a pile of hay there, and so (if we’d left it open which of these our term is to refer to) we may discover which (if either) of these Klut is. But notice what we cannot discover: if there is a kangaroo there, we can’t—at least not by just walking through the field and checking it out—discover whether Klut is a kangaroo or a mereological sum of cells.31 That ontological disambiguation must
analyticity and conceptual content
51
again be delivered by our referential intentions. It is not difficult to analyze what’s going on here: our categorial intentions may in certain cases be conditional on what application conditions are fulfilled in the world, for example, if the application conditions for terms of the category ‘animal’ are met, ‘Klut’ is to refer to an animal; if those for an inanimate artifact are met, ‘Klut’ is to refer to an inanimate artifact.32 As long as these application conditions are distinct, we can in a sense discover what category of item we have referred to by discovering which of the categories (on our disjunctive list) has its application conditions fulfilled in the situation. But this disambiguation procedure works only among possible categories that have mutually exclusive application conditions. Terms may also, however, share application conditions (either because these are the same, or because the application conditions for one are a subset of those for the other), even though they diverge in their coapplication conditions. So, for example, ‘kangaroo’ and ‘mereological sum of cells’ both apply in the situation in which a kangaroo is in the field, and no amount of empirical investigation can tell us which of these Klut is. So even in these cases, our referential intentions must include ways of disambiguating which category of entity is to be referred to, contingent on the world turning out in certain ways (e.g. if these rather than those application conditions are met). Categorial intentions still are needed and play a crucial disambiguating role in establishing the category of entity to be referred to, although, given the possibility of deferring to the world about what the facts are, there may nonetheless be room in some cases for empirical discoveries about which of a limited range of (disjunctively intended) categories our referent turns out to belong to. The second objection arises since, if we accept the hybrid view, even if competent speakers are fallible regarding specific features of the essence of the kinds picked out, they still are not fallible about the basic categories of kinds that are picked out by our terms, should these terms refer at all. Yet, as Laura Schroeter puts it, ‘‘scientific inquiry is no less relevant to discovering appropriate taxonomic categories than it is to determining the nature of particular objects and properties’’ (2004, 436)—so the hybrid view is said to be inadequate to account for the role that empirical discovery may play in revising our basic category schemes. But it is highly misleading to say that hybrid theories of reference cannot account for the role of empirical discovery in revising our basic categories and epistemic strategies. For although, on this view, competent speakers cannot, as a whole, turn out to be mistaken about
52
ordinary objects
what basic category of entity their term (as actually used now) refers to, if it refers at all, that does not leave our linguistic and conceptual schemes unresponsive to empirical discoveries. On the model I have been describing, setting up frame-level conditions of application and coapplication for terms is analogous to establishing other sets of rules for behavior, such as those that govern games. We cannot simply discover that the rules of NCAA basketball set up by the commissioners are wrong; these are constitutive of what it is (at that time) to properly play the game. Nonetheless, as mentioned above, that does not mean that these rules cannot be revised in light of practical needs, and even of later empirical discoveries. So if it is discovered that adopting a particular rule ultimately leads to games that last longer than viewers’ attention spans (or television stations’ tolerance), or to a prevalence of injuries among players, or other undesirable consequences, we may use those empirical discoveries to revise the rules (though this is not a case of empirically discovering that they were false). Similarly, given our investigative interests, our empirical discoveries about the world may lead us to revise the systems of categories we care about, conduct scientific research in terms of, or associate with our favored terms— but again this is a matter of empirical discoveries giving us incentive to change the basic rules of use; it is not a matter of discovering that the application conditions prescribed by those rules are mistaken. Thus the apparent role empirical investigations play in shaping the basic categories we use and associate with our terms does not require us to hold that the associated frame-level application and coapplication conditions are fallible (with all meaning externally determined) rather than simply revisable in light of both empirical discoveries and our purposes in using the terms. In fact, the divergences in intuitions about what to do in different cases provide some reason in favor of the view that such discoveries push the need for a semantic decision rather than making evident a new semantic fact. Given the need to accept that there is basic conceptual content to our terms to overcome the qua problem and disambiguate reference, it seems, we have still more reason to accept the latter interpretation of the role of empirical discoveries in affecting the most basic application and coapplication conditions associated with our terms. So these common objections do not give us reason to abandon the two linguistic claims I have provided preliminary arguments for in this chapter: first, that there are analytic entailments among our statements, and second, that the reference of our nominative terms is determinate only to the extent that these terms are associated with disambiguating
analyticity and conceptual content
53
frame-level application and coapplication conditions, fixing the category of entity to be referred to (if the term refers at all).33 The original goal of this chapter was to show how it is possible that there be analytic entailment relations among our sentences that are not just a matter of purely logical relations. Of course various theories of reference, including purely descriptive theories as well as the suggested hybrid causal theory, preserve the kinds of analytic entailments needed to respond to the causal redundancy argument. So the diagnosis I have offered of the causal redundancy argument—and some of those I will offer below of anti-colocation arguments (xx 4.1, 4.2, and 4.3) and arguments from parsimony (x 9.1)—may stand somewhat independently of the details of a theory of reference, as long as it allows for analytic entailments. Nonetheless, one of the major sources of resistance to the idea of analytic entailments comes from causal theories of reference. So I have tried here to motivate the claim that even those inclined toward causal theories of reference have reason to think that the reference of our nominative terms is only determinate to the extent that it is associated with category-specifying forms of conceptual content. What is important to note is that even conceptual content as minimal as that can ground the analytic entailments needed. For given the role application conditions for terms play in establishing the truth conditions for sentences, relations among meanings of terms can lead to analytic entailment relations among sentences. The work of this chapter, I hope, provides enough detail to at least show how such analytic interrelations may be grounded in the conceptual content of our terms—even on quite a minimal view of what conceptual content they have—and thus to give us at least preliminary reason to accept that there are such analytic entailments.
C c
three
identity, persistence, and modality
I have argued above that the reference of singular and sortal terms is determinate only to the extent that the term in question is associated with disambiguating frame-level application and coapplication conditions that establish what category of entity the term is to refer to, if it succeeds in referring. This thesis about reference has important consequences for our understanding of metaphysical claims about identity, persistence, and modality, which I will draw out in this chapter. I will begin in section 3.1 by laying out the relationship between the linguistic criteria for application and coapplication of our singular and sortal terms, and the criteria of existence, identity, and persistence for the entities (if any) these terms refer to. The most important consequence of the hybrid approach to reference defended above is that the most basic conditions of existence, identity, and persistence for the objects we refer to are discoverable by a kind of conceptual analysis, and the most basic claims about these conditions are analytic. In section 3.2 I will show how this in turn leads to the conceptualist view that the most basic modal claims are likewise analytic. But modal conceptualism is a much misunderstood view, so I will defend it against some common objections in sections 3.3 and 3.4, and discuss its implications for modal metaphysics in section 3.5. These results will play a crucial role in diagnosing problems with ordinary objects based on the grounding problem (x 4.4) and vagueness (chapter 5). They also have implications regarding what ontological questions are well formed and how we can sensibly approach existence 54
identity, persistence, and modality
55
and counting questions—issues to which I will return in chapter 6, the results of which will enable us to reexamine other arguments against ordinary objects based on the need for a nonarbitrary answer to the special composition question (chapter 7), on a supposed rivalry with scientific ontology (chapter 8), and on the need for parsimony (chapter 9).
3.1 Existence, Identity, and Persistence Conditions According to the view I have argued for above, a nominative term’s reference is only disambiguated to the extent that competent speakers intend it to refer to a certain category of entity, where this involves establishing at the frame level the conditions under which their term refers, and under which it may be used to refer again to the same entity. Criteria of application and coapplication are rules for the proper employment of terms of our language, and if they were to be stated, they would have to be stated in the metalanguage, with the terms in question mentioned. Nonetheless, all metaphysical claims must be expressed using language, and these rules for the proper application and coapplication of our terms play a central role in establishing the truthconditions for metaphysical claims that use those terms in claims about when individuals, or objects of a given sort, exist, are identical, and persist. I have already argued in section 2.5 that the truth-conditions for singular claims of nonexistence should be understood as: ‘N does not exist’ is true if and only if the history of uses of the name doesn’t lead back to a grounding situation in which the application conditions for terms of the associated category are met. Conversely, ‘N exists’ will be true if the relevant application conditions are met in the grounding situation. So, given that ‘N exists’ is true just in case N exists, we can also say that the (frame-level) application conditions for the relevant terms fix the (frame-level) existence conditions for the entities (if any) named by them. Similarly, in the case of sortal terms (rather than singular terms), the criteria of application involve conditions that are relevant to whether or not a sortal term ‘S’ may be applied correctly in a given situation, enabling us to truly say ‘That is an S’ or ‘There is an S,’ and so again the application conditions for the sortal term also fix the conditions under which it is true that there is something of the sort, and so (since ‘There is an S’ is true just in case there is an S) may be said to fix the (frame-level) existence conditions for things of that sort.
56
ordinary objects
Much the same goes for criteria of coapplication. These lay out conditions C under which (supposing it to have been successfully applied on two occasions) a sortal term may be reapplied to one and the same thing, enabling us to truly say ‘This is the same S as that,’ and so conditions C also fix the conditions under which any Ss are identical. These results carry over likewise for identity claims involving singular terms, such as ‘Sham is identical to Flam’. The identity claim only has a determinate truth-value if the names are associated with conditions that are sufficiently disambiguating to make the claim truth-evaluable (by making it determinate which entities are referred to). Suppose they are—suppose, for example, that ‘Sham’ is intended as an animal name, thus associating it with frame-level conditions under which it is true that Sham exists (that the application conditions for ‘animal’ are met in the original grounding situation) and frame-level conditions under which ‘Sham’ may be coapplied to one and the same thing, for example, that there be an animal spatio-temporally continuous (etc.: fill in with the proper account of identity conditions for animals) with that originally baptized as ‘Sham’. This then gives us truth-conditions for ‘Sham is identical to Flam’: it is true, and so Sham is identical to Flam, if and only if Flam is an animal and is spatio-temporally continuous (etc.) with Sham. In short, the coapplication conditions for terms of the category associated with the name also fix the truth-conditions for any identity claims made using the relevant names, and so fix (frame-level) identity conditions governing the objects (if any) referred to by those names. Indeed, what I above have been calling ‘coapplication’ conditions are typically simply called ‘identity conditions’ in the previous literature (e.g. Dummett 1973/1981, Lowe 1989), since they fix conditions of identity for anything falling under them. Nonetheless, since coapplication conditions are clearly conditions that apply to our terms, while identity conditions govern the things (if any) those terms refer to, to avoid confusion it seems best to keep these distinctions at the forefront, using ‘application’ and ‘coapplication’ to name the conditions on terms; ‘existence’ and ‘identity’ to name the corresponding conditions governing the objects (if any) referred to. As discussed in section 2.3, application conditions and coapplication conditions are relatively independent, insofar as two terms may share (at least a set of sufficient) application conditions, while diverging in their coapplication conditions—and thus diverging in the identity conditions that govern the things (if any) they refer to. This is an important point, since it suggests ‘that there be an object possessing certain identity conditions’ (as it is often, somewhat misleadingly, put) cannot
identity, persistence, and modality
57
be understood as among the conditions required of the world for a given name or sortal to apply—or else sortals with distinct coapplication conditions could not have the same application conditions. Application conditions give us rules for saying, in various hypothetical situations, whether or not a term would successfully be applied (and may be present without any coapplication conditions, e.g. for stuff terms). Coapplication conditions provide rules for when a successfully applied term may be reapplied to the same entity—rules that then fix the conditions of identity that govern the object (if any) referred to. So the basic application conditions (as I am using the term) cannot include that there be an object governed by certain identity conditions. Which identity conditions govern the object (if any) referred to is determined separately, by the associated coapplication conditions that give the rule for successfully reapplying the term (supposing the term is a wellformed sortal and does apply). There are two other general conclusions to draw about identity claims. First, identity claims are only well formed and truth-evaluable if the terms flanking the identity statement are associated with a certain category of entity each is to refer to, which disambiguates the reference of each term and the criteria of identity applicable to each.1 Second, identity claims are only true if the entities referred to are of the same category, and meet the criteria of identity appropriate for things of that category. One consequence of this is that all sortals of the same category must yield the same identity conditions for the things (if any) falling under them.2 For (given our definitions of species sortals and genus sortals in section 2.4), for every individual x of a species sort S, it is guaranteed that there is some y of genus sort S’ such that x ¼ y. But x and y could not be identical unless S and S’ have the same conditions of identity. So each species sort has the same conditions of identity as its genus sort, and thus all species sorts of the same genus sort must also have the same conditions of identity as each other (though their existence conditions may vary). Although individuals picked out by names associated with different categories cannot be identical, sortals of the same category may pick out the same individual, for example, a certain mammal may be identical with a certain cat (see Lowe 1989, 18).3 The existence conditions and identity conditions for individuals of the associated category also jointly determine the persistence conditions for the individual named, if the naming succeeds. Provided we allow that a name only continues to apply at all if it continues to apply to one and the same individual, an individual n successfully baptized by the name ‘n’ (associated with the categorial term ‘C’) continues to
58
ordinary objects
exist only provided ‘C’ continues to apply, and n continues to exist at a time only provided that some ‘n’ (of category C) exists at that time, where ‘n’ and n meet the identity conditions for Cs.4 Similarly, for a sortal term ‘S’, the conditions under which any individual S, say s1, persists at a time are just those conditions under which there is at that time an S (say, s2) and s1 and s2 meet the identity conditions for Ss. Given the role of conditions of application and coapplication in establishing the truth-conditions for claims about existence, identity, and persistence, where a term fails to be associated with sufficiently disambiguating conditions, claims about the existence, identity, or persistence of its referent (if any) may be simply indeterminate in truthvalue, and questions about these issues may be unanswerable. So, for example, some direct reference theorists might want to allow that the name ‘Smod’ may be used to pick out ‘that (distant) thing in on the hillside (affecting my eyes, whatever it might turn out to be)’, without specifying whether it is to be a name for a pile of rocks, an animal, a mass of molecules, a time slice of an animal, and so on. But even if that is allowed, the point here is that given the indeterminacies in the term’s meaning, many questions such as ‘Did Smod survive the fire?’ will have no determinate answer unless it is specified that, for example, ‘Smod’ was to be a name for a mass of molecules rather than for an animal. Given the above, the idea that we can discover the basic identity conditions governing things we refer to (as we might be able to discover their chemical structure or temperature) by simply finding them and empirically investigating them is misguided. The question ‘What are the basic identity conditions for n?’ (where ‘n’ is a name) is not well formed and answerable unless ‘n’ is associated with disambiguating application and coapplication conditions that enable us to determine whether and if so to what ‘n’ refers to. But if it is, the beginnings of an answer are already contained in the question, since if ‘n’ refers at all it must refer to something with the frame-level identity conditions associated with that category.5 So if we grasp the frame-level conceptual content of the term, we can gather the frame-level identity conditions suitable for n (if ‘n’ refers) by some form of conceptual analysis.6 (More, obviously, might be said about what exactly conceptual analysis involves, but here I will leave that open.)7 Similarly, if we ask ‘What are the identity conditions for Ss?’ where ‘S’ is a sortal term, the answer is not to be found by first discovering the Ss and then investigating them. Instead, it is to be found by analyzing the coapplication conditions associated with the relevant sortal term ‘S’. These establish the frame-level conditions under which any
identity, persistence, and modality
59
two things that could be referred to using that sortal term would be identical. Since persistence conditions are the joint product of existence and identity conditions, the question ‘Under what (basic) conditions does n persist?’ is similarly only well-formed and truth-evaluable if ‘n’ is associated with a categorial conception to disambiguate what it refers to. If no such conceptual content is associated with the name ‘n’ then the question is ill formed, since we cannot say which individual is being inquired about. But if there is such conceptual content, then we need not turn to investigate the world to determine what the frame-level persistence conditions for n are; if ‘n’ refers it must refer to something with the frame-level persistence conditions that are entailed by the existence and identity conditions suitable for n. In sum, on this view there are constitutive rules for proper use of our singular and sortal terms (involving their proper conditions of application and coapplication). These rules of use for the terms may be expressed in object-language claims about conditions of existence and identity for the things (if any) the terms apply to, where these objectlanguage claims use rather than mention the terms in question. Basic truths about frame-level identity and persistence conditions, stated in the object language, then turn out to be analytic, where this is understood in the broad sense as illustrations of the rules of use for the terms involved.8 So, for example, if ‘Margaret’ is a person-name and ‘Orky’ is an animal name, ‘Margaret (if she exists) is identical with any person P only if P developed from Margaret’s actual biological origin’9 and ‘Orky (if he exists) cannot survive death’ are analytic, since given the conceptual content of the singular terms used in making these statements, if those terms refer at all, they must refer to things with the relevant frame-level identity and persistence conditions. The same goes for generalized basic claims about identity and persistence conditions for sorts of things, for example, ‘Rocks (if there are any) cannot survive (at a world) any conditions that (at that world) result in their liquification,’ since the terms involved have categorial content that fixes the frame-level identity and persistence conditions governing anything that may fall under them. The view that analytic statements are in a sense illustrations of the rules of use for our terms has, however, come under heavy fire as part of criticisms of certain forms of conventionalism about analyticity. A first objection (Boghossian 1997, 336; Sider 2003a, 199–200) is that conventionalism makes ‘‘the truth of what is expressed [by an analytic claim] contingent, whereas most of the statements at stake in the present
60
ordinary objects
discussion are clearly necessary’’ (Boghossian, 336). For if analytic statements were ‘‘actually about language use’’ (Sider, 199)—if, for example, ‘All bachelors are men’ meant ‘It is a linguistic convention that ‘bachelor’ is to be applied only where ‘man’ is applied’—then it would clearly be contingent, since we might have adopted other linguistic conventions as constitutive rules for proper use of these symbols. This approach to analyticity has to be formulated very carefully to be tenable, and there is no doubt that some past formulations of the view that have gone under the name ‘conventionalism’ led to the difficulties seized on by critics. But the view proposed here is clearly not subject to this difficulty. Rules of use for our terms, if explicitly stated, must be formulated in the metalanguage, for example: apply ‘bachelor’ where and only where ‘unmarried man’ applies. But analytic statements are typically in the object language, where they illustrate the rules of use for our terms, where these are among the constitutive rules of proper language use. Illustrations of rules, however, are not the same as descriptive reports of rules, so it is not the case that ‘All bachelors are men’ means ‘It is a linguistic convention that ‘bachelor’ is to be applied only where ‘man’ is applied.’ While the latter is clearly about language, and contingently made true by the adoption of the convention, ‘All bachelors are unmarried men’ is not. Instead, the latter illustrates, rather than describing, the rules of use, and is in some sense about bachelors and men, rather than about language (since it uses these terms rather than mentioning them). (I will return to this issue below.) Classical conventionalist treatments of identity and persistence conditions also classically provoke accusations that we are (attempting to) illicitly ‘stipulate’ metaphysical facts about identity or persistence. So, for example, Katherine Hawley writes: The conventionalist suggestion is that the same physical processes of unraveling [a sweater], with the same qualitative results, can result either in the persistence of a single object, or in the destruction of an object, depending on what our conventions are: it is because we have the concept ‘sweater’ that there is something which does not survive unraveling. But this is to say that we can, simply by agreement, and from a distance, make it the case that the material object I see before me today either will or will not exist tomorrow. This would be a remarkable metaphysical party trick, if only we could arrange it. But unless we are idealists, we cannot suppose that objects exist only courtesy of our conventions. (2001, 148)
identity, persistence, and modality
61
But (whatever their virtues as directed at more extreme or incautious past views that have gone under the name ‘conventionalism’), it is also clear that these criticisms do not apply to the above view.10 I will have more to say below about why the view I am recommending is not subject to accusations of objectual antirealism. For now, note that it is not particular facts about persistence that are stipulated, but rather certain facts about language (at bottom, facts about the frame-level conditions under which a given term may be successfully applied, and under which it may be applied again to the same entity). And it should be no greater surprise, nor any more implausible, that facts about proper language use are established stipulatively through human practices than that facts about the proper playing of baseball, proper behavior at formal functions, and so on, are so established. But these facts about language have import for metaphysics, since all questions about, for example, existence and identity have to be asked linguistically, and claims about these issues have to be made using language. Another common objection to the view that basic identity and persistence conditions are fixed in fixing what (if anything) our terms are to refer to is that it ignores the fact that we may make surprising discoveries about identity and persistence. But on this view, room remains for several sorts of discovery about identity and persistence. First, since the relevant conditions for ordinary terms are typically established collectively and diffusely (and perhaps more through differential practices and reactions to certain uses as ‘right’ and ‘wrong’ than any internalized set of rules), conceptual analysis may be required to ferret out from our practices exactly what conditions are established, and to make these conditions explicit. Second, empirical work is required to discover particular facts about identity and persistence, for example, that Mark Twain is Samuel Clemens: whether or not Twain and Clemens fulfill the conditions relevant for people (among them spatio-temporal continuity) is a matter for empirical discovery, although which conditions are relevant for Twain and for Clemens is stipulatively established by those who ground ‘Twain’ and ‘Clemens’ as person names. We stipulate the rules of use for our terms, not whether (in a particular circumstance) they apply, or may be applied again, just as we stipulate (more formally and explicitly) the rules of baseball, but not whether or not Sosa is out at first base. Third, the frame-level conditions may appeal to empirical facts to fill them out (e.g. regarding what Margaret’s actual biological origin is, or what conditions result in the liquification of rocks at various
62
ordinary objects
worlds), providing more complete, derivative identity and persistence conditions. So given these frame-level conditions, we can go on to investigate their relations to empirical facts and thus to discover further empirical criteria for identity or persistence (this is probably what gives rise to the feeling that we can investigate identity conditions empirically by studying the things referred to). I will return to this issue in discussing the parallel issues for modal truths below.
3.2 Modal Truths If the most basic truths about identity and persistence conditions are analytic, so are the most basic conditional modal truths, for talk about identity and persistence conditions is a matter of talking about what sorts of changes an individual could undergo (or what variations it could tolerate) while still existing as one and the same. This is not, however, to say that on this view all modal truths are analytic and a priori; some modal truths may still be a posteriori, though these, too, are based on the frame-level analytic truth, with the empirical details it appeals to filled in appropriately. For example, with insertion of the relevant empirical facts, we can derive such claims as that Margaret Truman is necessarily Bess’s daughter, or that (in our world) rocks cannot survive being heated to 5,000 degrees Celsius. The view that all modal truths are ultimately based on analytic truths is generally discussed under the label ‘modal conventionalism,’11 a view prominently defended (in a somewhat different form) by Alan Sidelle (1989). Sidelle argues that we can account for the necessary a posteriori by maintaining that while the most general principles of individuation are analytic, these may appeal to actual empirical facts to fill them out in detail. For example, while it is not analytic that water is H2O, it is analytic that—if there is water—whatever water’s chemical composition actually is, water is of that kind necessarily. Empirical research enables us to fill in the details of what the chemical composition actually is, and our analytic principle then tells us that if water is actually H2O, it is necessarily H2O (35–7, 43–4). This enables us to retain the idea that modal truths are ultimately grounded in analyticities while still accounting for the fact that not all modal truths are knowable a priori (since the empirical filler-facts are not). In current discussions, the term ‘modal conventionalism’ is associated with at least three theses: (1) that all modal truths are ultimately based on analytic truths in the sense that modal truths are either analytic truths or
identity, persistence, and modality
63
based on combining an analytic truth with an empirical truth;12 (2) that it is not the case that modal properties are real or intrinsic features of the world;13 and (3) that the truth of modal propositions depends on our adopting certain linguistic conventions.14 Alan Sidelle apparently takes all three theses to be part and parcel of his conventionalist view, and takes theses 2 and 3 to be independently motivated (personal correspondence). But the approach to modality I am interested in here is best expressed simply in thesis 1, since it is that view that the above views on reference, identity, and persistence apparently commit me to. The name ‘modal conventionalism’, however, most naturally brings to mind thesis 3, not thesis 1, so to avoid perpetuating confusion, I will instead adopt the term ‘modal conceptualism,’ defining this simply as the view that modal truths are either analytic truths or based on combining an analytic truth with an empirical truth.15 So understood, it is clear that the view argued for above—that basic identity and existence conditions are fixed analytically in fixing reference—leads to modal conceptualism. The question to be considered here is: Is that a problem or a benefit? Critics regularly reject the conceptualist approach on alleged grounds that it requires us to accept thesis 3, while that in turn is said to lead to antirealism about objects, enabling us to accept (at best) an ontology of mere undifferentiated ‘stuff ’, since all ‘things’ in the world turn out to be mind-dependent. Since the latter view is generally taken to be noxious, this is often presented (Elder 2004; Rea 2002) as a sufficient reason to reject the view. In section 3.3 I will directly address two forms of argument that the conceptualist approach to modality leads to antirealism, argue that it does not, and attempt to diagnose the underlying error. Thus the central reason for rejecting this view may be discarded, and both the modal conceptualist view and the hybrid theory of reference that leads to it may be retained. In fact, if it is not subject to these criticisms, the conceptualist approach to modality is independently appealing (as I will argue in x 3.5), as it eases the way to avoiding some of the ontological and epistemic problems of modality.
3.3 Modal Conceptualism and Objectual Antirealism The view that modal truths are ultimately based on analytic truths is widely held to lead to antirealism about objects. Sometimes the view
64
ordinary objects
is accused of being antirealist, since it is said to hold that all objects depend on (or ‘are constructed by’) human conventions, language, or mental representations. Other times, while noting that it is not antirealist insofar as it may acknowledge the existence of mind-independent ‘stuff’, the view is nonetheless presented as committed to an ontology of mere stuff, not things, since again objects (things, individuals) are allegedly held to exist only given our conventions, language, or concepts.16 Alan Sidelle presents the latter as the preferred interpretation of his view:17 Rather than selecting from among the many objects out there waiting to be referred to, the conventions articulate (or create or construct— but ‘articulate’ seems to me better) objects from the independently inarticulate world. We need these conventions not because there are too many items out there to refer to otherwise—but because there are too few: in fact, none. (1992a, 284–5)
While Sidelle defends the resulting ‘stuff’ ontology, critics of the approach often take the alleged fact that it leads to objectual antirealism to give us grounds for rejecting it. Thus, for example, Crawford Elder closes his critique of Sidelle’s conventionalism with the remark ‘‘Conventionalism, I contend, ultimately founders on its refusal to allow that any objects in the world possess mind-independent existences’’ (2004, 20).18 Accusations of objectual antirealism often come from the thought that thesis 1, that all modal truths are analytic or based on combining an analytic truth with an empirical truth, leads to thesis 3, that the truth of modal propositions depends on our adopting linguistic (or other) conventions.19 Michael Rea takes Sidelle to make this move and does not criticize it;20 he simply argues that it leads to antirealism about material objects. His argument may be reconstructed as follows. 1. All modal truths are either analytic or derived from a conjunction of an analytic truth with an empirical truth (definition of modal conceptualism). 2. If modal conceptualism is true, then what makes modal propositions true ‘‘is the fact that we have adopted linguistic conventions governing the use of the words in [the sentence expressing a modal proposition] according to which that sentence has to be true, regardless of the empirical facts’’ (Rea 2002, 85–6). 3. It follows from (1) and (2) that ‘‘modal properties are exemplified in a region only if the matter in that region stands in particular contingent relations to human beings and their mental activity’’ (86), making all modal properties extrinsic.
identity, persistence, and modality
65
4. If all modal properties are extrinsic, so are all sortal properties, since ‘‘what a thing can and cannot survive depends on what kind of thing it is; and what kind of thing it is depends on what it can and cannot survive’’ (95). 5. But if all sortal properties are extrinsic, realism about material objects is false (95). 6. Therefore, modal conceptualism entails that realism about material objects is false.
But does it follow from the basic conceptualist view (premise 1) that what makes modal propositions true is the fact that we have adopted certain linguistic conventions (as premise 2 alleges)—that is, that modal propositions are only mind-dependently true? There is a sense in which (conceptualism aside) the truth of any sentence could be said to be mind-dependent—even the truth of a simple nonmodal sentence about stuff such as ‘There is gold (in the world)’ could be said to be mind-dependent insofar as it would not be true if the term ‘gold’ were meaningless (say if there were no English language) and it would be false if by ‘gold’ we meant what we mean by ‘magical fairy dust.’ That is, minds obviously play a role in determining whether sentences are true or false by establishing the meanings of the sentences that contribute to their truth-conditions. And, of course, if a series of marks or noises had different meaning, it might have a different truth-value. But this is not what anyone has in mind by the truth of a sentence being mind-dependent. In a world without minds (the response goes), there would still be gold, it just wouldn’t be called ‘gold’ (since there would be no linguistic conventions); to think otherwise is to make a usemention mistake. And so the truth of ‘There is gold’ is not minddependent in the relevant sense: the meaning of the sentence is minddependent, but that’s no surprise; once the meaning is fixed, however, the meaningful sentence may be true independently of the existence of minds. In short, when we ask whether a statement is mind-independently true, we must be asking whether, holding the meaning of that statement fixed, the existence of minds is a necessary condition for its truth. Is the proposition expressed by ‘Rocky could not survive being heated to 5,000 degrees Celsius’ only mind-dependently true on the conceptualist view? According to the conceptualist view, the name ‘Rocky’ has frame-level conceptual content that disambiguates among possible intended referents by establishing the sort of thing the name is to refer to (a rock) via frame-level conditions that must be fulfilled for the name to be successfully grounded, and other frame-level
66
ordinary objects
conditions that must be fulfilled to properly apply the name (to the same thing) again. Together, these determine the frame-level conditions under which the thing referred to persists (i.e. under which one and the same rock remains available for reference). Once the meanings involved establish what sort of thing Rocky must be, it is established that Rocky (if there is one) is the sort of thing that could not survive liquification (e.g. preservation of solidity is plausibly one frame-level necessary condition for the persistence of rocks), so that if (in a given world) certain conditions result in liquification, (in that world) Rocky could not survive those conditions.21 This is a persistence condition for Rocky, and does not yet establish whether or not Rocky actually would survive the heating in question (i.e. whether that condition is fulfilled in the situation described). To establish the truthvalue of the claim, we must check the (actual) world to see if being heated to 5,000 degrees Celsius results in liquification of rocks. Since it does, one of the necessary conditions for the truth (at this world) of ‘Rocky persists’ is not met in such circumstances, and so it follows that Rocky would not survive such heating. But this may be true even in a world without minds. Of course, by ‘Rocky’ we could have meant the mereological sum of these atoms, in which case the referent of that name would be capable of surviving, but then we would not be talking about Rocky, the actual rock before us (named by our actual term) at all, so this no more shows the truth of ‘Rocky couldn’t survive being heated to 5,000 degrees Celsius’ to be mind-dependent than the possibility of meaning by ‘gold’ what we mean by ‘fairy dust’ shows the truth of ‘There is gold’ to be mind-dependent. So thesis 3 does not follow from the basic (thesis 1) claim of modal conceptualism. The contribution minds and linguistic conventions make to determining the truth of modal statements on this view is no different from the contribution they make to other statements: in all cases, they establish the meaning and thereby help establish the truth-conditions of the statements, but don’t establish whether or not these are fulfilled; in fact, normally minds and conventions aren’t required for these truthconditions to be fulfilled. To move from the fact that (on the conceptualist view) minds and language are needed to establish the meanings of modal statements to the inference that on that view modal properties turn out to be mind-dependent is at bottom a use-mention mistake. Given that simple parallel analysis, one might wonder why there is such a prevalent tendency to think that the conceptualist must hold modal truths to be mind-dependent in a sense that nonmodal truths
identity, persistence, and modality
67
aren’t. The reason, I think, is this: we can normally distinguish the contributions of the meaning-establishing component from the factual component in determining the truth of a statement. But as we have seen, the most basic modal statements, for example, ‘Rocky (if he exists) could survive changes in location’ or ‘Rocks (if there are any) could not survive liquification’ are analytic. But analytic claims, traditionally understood, are those for which the factual component is null—nothing is required of the world to make it true, as this is ensured by the relations of meanings alone. There may then be a tendency to think that the truth of these claims is mind-dependent, since it is so directly (and exclusively) dependent on the meanings of terms, which obviously are established by our conventions. But it is as much a mistake here as in the case of nonanalytic statements to confuse the fact that minds are needed to establish the meaning of a sentence with the claim that the truth-conditions for the sentence include the existence of minds. So in the argument above, premise 2 must be rejected if it is interpreted as saying (premise 2a) that modal propositions require as truthmakers the (empirical) facts that we have adopted certain conventions. It is acceptable if it is interpreted as saying (premise 2b) that if modal conceptualism is true, then modal propositions are guaranteed to be true in virtue of their meanings (requiring no empirical facts). But from premises 1 and 2b, it clearly does not follow that (premise 3) modal properties are only exemplified in a region if matter in that region stands in certain contingent relations to human beings. The fundamental mistake of this and similar attacks on modal conceptualism seems to lie in assuming that modal truths require truthmakers, and concluding that these must either be intrinsic modal properties (considered as mind-independent and discoverable features of the world) or extrinsic properties whose existence depends on human minds and conventions (see Rea 2002, 77–8). But both of these options miss the point according to the conceptualist: since they are analytic, the most basic modal truths do not require truth-makers at all. A similar misunderstanding seems to underlie Crawford Elder’s recent arguments that ‘modal conventionalism’ (the term he uses, following Sidelle) is paradoxical in the literal sense of being ‘beyond belief’. For, he argues, if we accept that identity conditions are established by the linguistic conventions governing how our terms are to refer, then we must also accept that there being pluralities (i.e. two objects rather than one in a certain situation) is based on our conventions. (His argument for this is that judging that there are two Ks rather than one requires believing ‘‘that there exists at some time a K having some
68
ordinary objects
property p, and a K having some property p0 , such that no one K can simultaneously have p and p0 ’’ [2004, 15].) But then the conventionalist is caught in a paradox, for ‘‘the obtaining in the world of our conventions of individuation is [according to that view] logically prior to the existence of us as a plurality—and for that matter is logically prior to the conventions’ being conventions, plural. Yet surely it must also be true that our existing in the world is logically prior to our having any particular conventions’’ (17). In short, the conventionalist is said to run into paradox by being forced to hold both that conventions are logically prior to people (since this is a plurality, and so requires adopting conventions of individuation) and that people are logically prior to conventions (since people are presumably the ones who make up conventions). But we can unravel the supposed paradox by noting an ambiguity in ‘logical priority’. Claims that there are two people here (since Mary can’t be identical to Jane in virtue of their having different origins) may rely on the truth of a basic modal claim (about basic individuative criteria for people) and of empirical claims (about Mary’s and Jane’s particular origins), but neither of these requires the existence of conventions as truth-makers. So it is not the case that, on this view, there must be conventions for it to be a fact that there are two people here; our conventions are ‘logically prior’ to objects, including people, only in the sense that our conventions fix the meanings of our terms, thereby establishing what it would take for there to be objects of various sorts (including people), and what the most basic conditions of identity would be for the things (if any) they refer to. By contrast, people are ‘logically prior’ to conventions, in the sense that conventions cannot exist without people, and claims about the existence of conventions do require people among their truth-makers.
3.4 Analyticity and Truth-Makers I have argued that the fundamental mistake behind arguments that modal conceptualism leads to objectual antirealism is to assume that modal truths require truth-makers. But although it is a venerable view that analytic claims do not require truth-makers, as they place no demands on the world, this claim, too, has often been challenged: How, it might be asked, can there be any truths without truth-makers, and how can analytic claims fail to be saying something about the world?
identity, persistence, and modality
69
I would like to suggest least two ways in which we can make sense of the idea that analytic claims do not require truth-makers, the second of which will be more useful than the first—but I think the first is important enough to warrant mention. First, on the view I have defended above, analytic claims are illustrations of constitutive rules of language use. But rules are just disguised (and generalized) commands, so insofar as they are used as illustrations of some of the constitutive rules of language use, analytic claims should not be understood as reports of or assertions about anything, and thus as not expressions apt for truth or falsehood. Instead, with their rule-demonstrating force, they should be understood as something like a converted command, much as demonstrations of the proper way to dance the merengue or to set up a corner kick are. And indeed in conversation, analytic statements are often used as implicit correctives to someone we feel is misusing the term. Nonetheless, a lingering doubt may remain: surely (some may feel) analytic claims like ‘All bachelors are male’ do say something true about the world, namely that all the bachelors in it are male. So the view that analytic claims are simply illustrations of rules of use for our terms, and as such are not properly said to be true or false, may seem unsatisfying at best. As Sider puts the objection, how could ‘‘All bachelors are male’’ not say anything about the world? It contains a quantifier over bachelors, and says of them that they are male. So it says something about the properties of bachelors—as worldly entities as one could ask for. (2003a, 202)
I think there is a way to accommodate this, but we must tread carefully here. Many pieces of language may be used to perform different speech-acts, for example, ‘‘the door is open’’ may be used as a simple description, or as request to close it. So, similarly, even if (as I have argued) the forms of expression that occur in analytic statements are fundamentally used as illustrations of rules (in which case they are neither true nor false), they may also be used as descriptions, in which case the truth-conditions set by the rules of use are in force in establishing the truth-conditions for this (genuine) description. The sense in which analytic claims seem to be about the world is that they are stated in the object-language, using terms that at least purport to refer to nonlinguistic items of the world. But there is another perfectly good sense in which they ‘‘say nothing’’ about the world and are ‘‘entirely devoid of factual content’’ (Ayer 1952, 79). That sense is that—if we do treat them as true—it is clear that their
70
ordinary objects
truth does not depend on any empirical fact’s obtaining. Indeed, as Husserl often emphasized through his appeal to bracketing, the truth of such statements does not even depend on the real existence of an external world at all—or any part of it. So the sense in which the analytic claim ‘All bachelors are male’ is not ‘about’ bachelors or men is that, even if there were no bachelors, indeed no men whatsoever (and never had been), the sentence would be true. But if we treat analytic statements, descriptively used, as guaranteed to be true, another classic problem for conventionalism rears its head: How can a sentence be guaranteed to be true ‘by definition’ or ‘in virtue of meanings’? I think there is a way to understand this, on the above model—again without falling into the view that they are somehow made true ‘by pronouncement’ (a position criticized by both Boghossian, 1997, 336, and Sider, 2003a, 201). As I have described it, the rules of use for our terms set the application conditions for the terms they mention, which play a role in fixing the truth conditions for sentences in which those terms are used. So consider the analytic claim ‘All bachelors are male’ or, rendered more formally, 8x (Bx ! Mx). The relevant rule of use is: ‘apply ‘bachelor’ only where ‘male’ may be applied’ so the truth-conditions for ‘x is a bachelor’ include that x is male. This guarantees that if there is something that is a bachelor (i.e. to which ‘bachelor’ applies), then it is male (i.e. ‘male’ applies to it). This guarantees the truth of the conditional, for if the antecedent is true, the consequent is guaranteed to be true. The truth of the analytic claim, taken as a genuine description (of universally quantified conditional form), is guaranteed given the relations in the rules of use for the terms employed in the antecedent and the consequent— though the adoption of these rules is not a truth-maker for the claim (it only establishes the meaning of the terms involved and the truthconditions for each part). This also makes sense of the idea that the truth of analytic claims such as ‘All bachelors are men’ is independent of all empirical facts—even of there being bachelors or men, or indeed anything at all.
3.5 Modal Properties The central thesis of what I have called ‘modal conceptualism’—that all modal truths are either analytic truths or based on combining an analytic truth with an empirical truth—is often thought to commit one to thesis 2, denying a robust modal realism.22 If this is understood as the claim that
identity, persistence, and modality
71
modal properties are mind-dependent, it should be rejected for the same reasons we rejected thesis 3. So if thesis 2 is to retain any plausibility, we must understand it somehow as rejecting the view that modal properties are substantive features ‘out there’ in the world in much the same way that ordinary properties (such as being round or negatively charged) are, while allowing that claims about objects possessing modal properties may be mind-independently true.23 Thesis 1 also does not entail thesis 2, thus understood. For it is, for example, perfectly consistent with thesis 1 to hold a ‘modally plentitudinous’ ontology, according to which wherever there is an object at all, there are objects with intrinsic modal properties instantiating every consistent modal profile (see Bennett 2004, 354–63; Hawley 2001, 6–7; Rea 2002, 90–6; Sidelle 1992a, 284–5). On this view, our conventions are needed to disambiguate which of the many intrinsically modally rich objects we are describing, making claims about its basic identity and persistence conditions analytic (see Bennett, 361–2), yet we still posit modal properties as just as much a real and discoverable part of the world as any other properties, and capable of serving as truth-makers of modal claims. But while modal conceptualism as captured by thesis 1 doesn’t rule out the idea that there are such substantive modal properties and so doesn’t entail thesis 2, thus understood, it does render substantive modal features (whatever these are supposed to be) completely unnecessary to account for the truth of our modal claims.24 For the most basic modal truths are analytic claims that require no truth-makers—they place no conditions on the world—and others are derivable from these plus nonmodal empirical claims. So the minimal conceptualist need not accept anything other than nonmodal empirical features of the world as truth-makers for modal claims (and even that, just for the derived ones). Nonetheless, even the minimal conceptualist may, and I think should, still accept that there are modal properties in a sense, since we can move pleonastically from ‘Rocky couldn’t survive being liquefied’ to ‘Rocky exemplifies the modal property of not-possibly-surviving being liquefied’.25 If this is the case, the conceptualist approach to modality is independently appealing, since we have reason to think that it could help soften epistemic and ontological worries about modality—even without adopting the problematic claim of thesis 3. Talk about modal properties on this view may be understood as pleonastically derived from modal truths that have no need of truth-makers. And knowledge of modal properties comes with knowledge of the relevant analyticities— combined, as necessary, with knowledge of the empirical facts appealed
72
ordinary objects
to by the frame-level principles of individuation. While providing a story of how such knowledge may be acquired may be no simple matter, it clearly does not require positing some special form of intuition or any kind of knowledge acquisition beyond what is already required for knowing the meanings of claims in our language and knowing standard empirical facts. In any case, what is crucial for the work to follow is to note that the hybrid view of reference defended above and the form of modal conceptualism that follows from it do not lead us into any kind of antirealism, or to an ontology of mere stuff as opposed to things. They do, however, lead to the view that basic (frame-level) claims about identity and persistence conditions and modal features are analytic. This thesis will play a crucial role in enabling us to overcome the colocation problem and other difficulties supposed to arise if we accept ordinary objects. The time has now come to return to these problems and see how the fundamental theses argued for in this and the preceding chapter can help resolve them.
C c
four
problems of colocation
One prominent problem said to arise for those who accept an ontology of ordinary objects is that if we accept, say, the existence of statues as well as that of the lumps of clay of which they are made, and deny that they are identical (given their different identity conditions), we end up having to accept that (at least) two objects are colocated. The problem is sometimes expressed in claims that accepting the existence of artifacts, common sense natural objects, and the like would implausibly multiply the number of objects we must posit in a situation, forcing us to accept that, say over and above or in addition to a lump of clay in a certain corner, there is also a statue. But since it seems wrong to say that there is a statue over and above the clay, the tables are suddenly turned rhetorically, for then it is the realist about ordinary objects who seems to be the one defending an implausible position in tension with common sense.1 While such ‘nothing over and above’ claims function more as rhetorical appeals than as arguments, some of the most formidable arguments against ordinary objects are based on a similar worry about the realist’s need to accept that, since there is a lump of clay and also a statue not identical to the lump, we have ‘two’ colocated entities here and in many other cases that follow the same pattern (money and paper, persons and bodies, etc.). But colocation is often thought to be problematic, and if it is, then we must either (contrary to the view developed above) hold that the entities are identical or reject at least one of the entities supposedly colocated (Merricks 2001, 41). Such arguments are usually deployed 73
74
ordinary objects
against statues and similar constituted entities, but Merricks (41–2) argues that various attempts to eliminate one but not the other would be unacceptably arbitrary. Thus, to avoid those problems, worries about colocation can be turned into arguments for eliminating not just artifacts, but all inanimate composite objects (both statues and lumps alike). Accepting colocation is often thought to be one of the biggest problems facing those who accept the existence of ordinary objects: Merricks (2001), Zimmerman (1995, 90), and Heller (1990), among others, reject colocated objects, while Sidelle (2002, 121–2) treats colocation as a crucial stumbling block confronting realists about common sense objects. But what, exactly, is the problem with colocation supposed to be? The colocation problem used to be characterized as being based on violating the plausible principle that no two things may have exactly the same spatial location for exactly the same period of time ( Wiggins 1968, 90). But since some alleged cases of colocation (e.g. of a space-time region and a material object) might not be as problematic as the statue/lump case, the colocation problem is now more often put in mereological rather than spatial terms. That is, it is now most often posed as the problem that accepting such objects violates the principle that at a given time ‘‘no two physical objects could be composed of exactly the same parts at some level of decomposition’’ (Merricks, 38). The ‘‘at some level of decomposition’’ clause is meant to acknowledge that, while there might in some sense be parts of the statue such as hands and feet that are not parts of the lump, the problem of colocation arises as long as, at a finer grained level of division (e.g. into molecules), the statue and lump resolve into the same parts.2 Following Sidelle (119), we can call the principle that no two objects may be composed of exactly the same parts at the same time the ‘‘no coincidence’’ principle. Thus the colocation argument, like the causal redundancy argument, is based on the idea that we should reject ordinary objects, since accepting them would conflict with some independently plausible metaphysical principle. In this chapter I will begin (in x 4.1) by addressing the rhetorical appeals that surely there is ‘nothing over and above’ the constituting lump of stuff. While much mileage has been derived from these appeals, I will argue that the gain is ill gotten, and that the awkwardness of ‘something over and above’ claims does nothing to undermine the plausibility of accepting ordinary objects. Defusing such rhetorical moves can already make it evident that the colocation problem is not nearly as bad as its proponents make it sound. I then go on in sections 4.2–4.5 to address and respond more directly to various worries surrounding the colocation problem.
problems of colocation
75
4.1 ‘Nothing Over and Above’ Worries about so-called multiplication of objects are often initially expressed in intuitive rhetorical appeals to the effect that surely, once one recognizes the existence of the atoms arranged baseballwise, it is clear that there is nothing but the atoms there, or that there is no baseball over and above the atoms. Thus, for example, van Inwagen suggests that if we weave a long thin snake into a hammock, the snake (if intelligent) would aptly observe ‘‘there’s nothing here but me’’ (1990, 127). So, similarly, in all cases of ‘creating’ artifacts, van Inwagen insists, ‘‘we have not augmented the furniture of the world but only rearranged it’’ (127). Horgan and Potrcˇ similarly express the view that ‘‘it is not plausible that institutional entities like corporations and universities are denizens of the world-itself, over and above entities like persons, buildings, land masses, items of office equipment, and the like’’ (2000, 257, italics original). Merricks expresses the same idea in denying that there is anything additional to the atoms: ‘‘Sometimes ‘there are statues’ means that there is, in addition to various atoms in statuesque arrangement, some much bigger object—with a mass, centre of gravity, and so on—that has each of those atoms as a part. And when this is the meaning of ‘there are statues’, says the eliminativist, ‘there are statues’ is false’’ (2001, 15). Merricks draws out this ‘nothing but’ or ‘nothing over and above’ rhetoric by appealing to counting to express the idea that nothing more is there than the relevant simples, using this as the main way to distinguish his eliminativist position from positions that accept composite inanimate objects: ‘‘Suppose there are a million atoms arranged statuewise in a certain region of space. And suppose we ask how many (nonsubatomic) things there are in that region. My metaphysical opponents would say that there are at least one million and one (the atoms and the statue). I would say there are only one million’’ (2000, 48). Similarly, in objecting to the idea of constitution without identity, David Lewis writes: ‘‘It reeks of double counting to say that here we have a dishpan, and we also have a dishpan-shaped bit of plastic that is just where the dishpan is. . . . This multiplication of entities is absurd on its face’’ (1986, 252). While remarks of these sorts are not generally put forward as offering an argument for eliminativism, they are a central part of the rhetorical arsenal of the eliminativist, designed to make the realist about ordinary objects appear to be the one defending an incredible or,
76
ordinary objects
more weakly, implausible view. And such appeals can be powerful— there does seem to be something wrong with the idea that, over and above, or in addition to the particles properly arranged baseballwise, there is also a baseball distinct from those. But, as one might begin to suspect, given the results of chapter 1, the reason there is something wrong with these kinds of claims (so that readers are not inclined to ally themselves with them) is not that they are false (thus supposedly entailing the truth of eliminativism) but rather that they are based on false presuppositions. They are not things that a thoughtful realist about ordinary objects is likely to say or needs, for any reason, to assert—in fact, she can (and should) reject such statements not by joining the ranks of eliminativists in declaring their falsehood, but by pointing out the fact that they are based on the false presupposition that claims about the presence of the baseball (or statue) and its properly arranged atoms are independent. Consider, for example, the supposedly incredible claims that, in a situation with atoms arranged statuewise, there are atoms and there is (also) a statue (a snake and [also] a hammock; a lot of sand particles and [also] a fort). . . . These are indeed inappropriate things to say—not because really there is no statue there, but rather because sensibly asserting in a single conjunctive list that there are atoms arranged statuewise and (that there is) a statue presupposes that these claims are analytically independent, when they certainly are not. Such claims are inappropriate for just the same reason as it is inappropriate to say ‘He bought a lefthand glove, and a right-hand glove, and (also) a pair of gloves’ or ‘She watched the battalions, squadrons, and (also) a division march past.’3 Whether one takes such claims to be lacking in truth-value in virtue of this presupposition failure or to be merely misleading (though strictly true), one can easily explain their inappropriateness and thus our hesitancy to ally ourselves with them, while noting that the inappropriateness of these claims casts no doubt on the view that there are statues, pairs of gloves, and divisions. Claims that statues are nothing ‘over and above’ or ‘in addition to’ the atoms (relevantly arranged) are compelling for a similar reason: we begin with a list of component parts (the atoms) and then assert that there is nothing more there. The appropriate sorts of item to conjoin to the list of parts would be other component parts—and indeed, as the realist (like anyone) admits, no additional parts4 or ingredients (statue souls, essences, what have you) are ‘thrown in’ or needed to make a statue, beyond the component atoms. It would clearly be inappropriate (as we have seen above) to add the name for the whole to
problems of colocation
77
the list, as if this were an independent claim from the prior claims about the parts and their arrangements. So we readily assent to the claim that there is nothing ‘over and above’ the relevant atoms there. But the fact that we agree that the list is complete even though it does not include the whole statue (and that we would be hesitant to add the ‘statue’ to the list) is not in the least inconsistent with asserting (in a different ‘‘logical tone of voice’’)5 that there is a statue there. The ‘counting’ formulation of the claim, that it is implausible to think that if there are a million atoms arranged statuewise then there are a million and one things there (the atoms and the statue), encounters the same problem, for counting is simply a formal way of using a long conjunction (there’s this apple [1], and that one [2], and that one [3] . . .), so that where asserting a conjunction would be inappropriate due to analytic relations among the conjuncts, so is ‘counting’ them all up in a single tally.6 Thus, it would be equally inappropriate to say that Ryle’s shopper bought three things: the left glove, the right glove, and the pair of gloves. So, indeed, it would be inappropriate to say that a million and one things exist in a certain situation: a million atoms and the statue composed of them. But this has nothing to do with inherent problems with accepting statues.7 (This is not to say that, in some contexts, we cannot truly count entities that have such relations in a single tally, but rather that the apparent inappropriateness of counts like that Merricks attributes to the realist about ordinary objects is based on this kind of presupposition violation. I will have more to say about the presuppositions of counting and what sorts of counts can and cannot be sensibly undertaken in chapter 6.) ‘Nothing over and above’ appeals lose their rhetorical hold over us when we recognize that the apparently unpalatable phrases put into the mouth of the defender of ordinary objects are unpalatable not because they are false (or their negations are true) but because the attempt to assert them only makes sense if we presuppose independence between claims about the object and those about its constituting matter. Such appeals thus follow much the same pattern we saw behind part of the causal redundancy argument. The eliminativist takes a form of statement whose sensible use requires that we presuppose that the items referred to are separate and independent entities (e.g. ‘The xs and O both caused E’; ‘There are xs and, in addition, there is a y’). She then substitutes in terms for an ordinary object and its constituting matter, and attributes the assertion of the claim to the realist about ordinary objects, leaving the realist apparently saying such inappropriate things as ‘The atoms arranged baseballwise and the baseball both
78
ordinary objects
caused the window to shatter’ or ‘There are atoms arranged baseballwise and, in addition, there is a baseball.’ Then she uses the patent inappropriateness of the assertion as motivation for denying that claim and embracing eliminativism. But that cannot be the right move, since negations inherit the presuppositions of the statements negated (Soames 1989, 558; Stalnaker 1973, 448), and so accepting the negation of a statement with presupposition failure is as inappropriate as accepting the original. The true source of the problem is not accepting ordinary objects, but rather violating the presupposition for appropriate use of the phrase. With that assessment in hand, we can see why the colocation problem initially sounds bad: if you think of colocation as saying, for example, ‘‘In the centre of the market-place at a given time there is . . . the ship, and also the collection of planks’’ (Myro 1997, 153) or that there are ‘‘two numerically distinct objects . . . [sharing] all their parts at some level of decomposition’’ (Merricks 2001, 39), it does indeed sound like a crazy thing to say. But notice that at least part of the problem with what’s being said here is that such assertions are (like those in the causal redundancy argument and nothing-over-and-above claims) inappropriate because they violate presupposition requirements in adding or conjoining terms. If we reject that form of speech, we can already make colocation sound less crazy, and less problematic, by describing it as accepting (e.g.) that there is a statue, there is a lump, the statue is not identical with the lump (in virtue of different persistence conditions, etc.), and the atoms making up the statue are the same as the atoms making up the lump. In short, if we properly describe the situation in a way that avoids presupposition failures, much of the apparent implausibility of colocation begins to fade away.
4.2 The No Coincidence Principle The widespread wariness about colocation may be attributed at least in part to the sense that, when we are talking about material objects at least (like statues and lumps), they must (as Wiggins 1968, 94, put it) ‘‘compete for room in the world, and . . . tend to displace one another,’’ so that ‘‘intuitively, the problem with coincident entities is that of overcrowding. There just is not enough room for them’’ (Heller 1990, 14). Or similarly (for those who prefer to put the colocation problem in terms of sharing parts rather than space) the nervousness arises from the sense that material objects are rivals in their claims to be composed
problems of colocation
79
of the same parts at the same time, or in their claims to have mass or other physical properties (see Rea 1997a, 368; Zimmerman 1995, 87–8). Thus claims of colocation apparently violate another plausible metaphysical principle, the no coincidence principle, which holds that if A is an object and B is an object, and A is not identical to B, then (spatial version) A and B cannot occupy the same volume at the same time or (mereological version) A and B cannot be composed of exactly the same parts (at some level of decomposition) at the same time. But while these are natural assumptions when we are discussing separate and independent objects, the earlier discussions of causal redundancy should make us suspicious of claims that there is such a rivalry or competition between, for example, atoms arranged statuewise and the statue composed of them. The no coincidence principle itself, like the causal principle, derives its plausibility from considering separate and independent entities, whose claims to occupy a location at a time or to be composed of exactly the same parts at some level of decomposition (a clause I will henceforth omit for brevity) would genuinely be independent, rival claims. But it is not so obvious that this plausibility carries over when we discuss entities as closely interrelated as a statue and its constituting matter are, on the constitution view. For the defender of a constitution view holds that there is a statue, there is a lump, and the statue is not identical with the lump but is materially constituted by it. The fact that there is a statue analytically entails that there a physical lump constituting it, and the fact that the statue is in a particular location and made of particular parts analytically entails that its constitutive lump is in that particular location and made of these particular parts.8 (Competent speakers need know no more than that David is a statue and is here now and made of these parts to infer that its constitutive lump is here now and made of these parts.) But, as I argued in chapter 1, where there are such analytic entailments, it is clear that the truthmakers for the prior set of claims are also sufficient truth-makers for the latter claim, so that we are not making separate and independent claims that could aptly be considered rivals. This gives us reason to suspect that the no coincidence principle (like the causal principle) should be restricted in its application, not accepted as a ‘completely general’ metaphysical principle. The appropriate version of the no coincidence principle, perfectly in line with the analyses above and in chapter 1, it seems should be restricted to cases in which the two claims to occupy the same volume at a given
80
ordinary objects
time, or to be composed of the same parts at a time, are analytically independent.9 For where there are analytic interrelations between the claims to occupy a certain location or be composed of certain parts, there is no rivalry between these claims, and so, for example, there is no rivalry for space or parts between a lump and the statue it constitutes. Thus the defender of a constitution-without-identity view can consistently maintain her view without violating a version of the no coincidence principle restricted in this way. The plausibility of this restricted version of the no coincidence principle can explain our initial conviction that the principle is true, but the restricted principle clearly does not interfere with accepting the existence of constituted objects as well as their constituting bases.
4.3 Property Additivity A similar range of objections to the view that artifacts such as statues are constituted by, but not identical to, the material lumps of stuff that constitute them focuses not on a supposedly problematic proliferation of objects at a given location, but rather on an apparent proliferation of instances of properties shared by these objects. The problem is most often expressed in terms of mass (Rea 1997a; Zimmerman 1995): suppose, for example, that David has a mass of 500 kilograms. Lumpl as well then, it seems, will have a mass of 500 kilograms (since they share all the same parts at some level of decomposition). But if we accept a principle of additivity of weights—that, where A and B are not identical, if A weighs n kilos and B weighs m kilos, then A and B together weigh (n þ m) kilos—it seems that we should expect that when we put ‘both’ David and Lumpl on the scale, it will read ‘‘1,000 kilos.’’10 This supposed problem, of course, is parallel to the supposed problem of overdetermination, broadened to encompass not just the property of causing an event E, but other properties such as mass, height, and so on. The general direction of reply then is the same as that utilized in section 1.2, that (when a and b both have property P) there is no rivalry for possession of a single property instance, nor any ‘doubling up’ of properties where (given the constitution relations between them) a’s being P is an analytic entailment of b’s being P.11 Thus, just as the battalions, batteries, and squadrons marching past, and the division marching past, do not yield a double-supply of
problems of colocation
81
marching past, and the baseball shattering the window and the atoms arranged baseballwise shattering the window do not yield a ‘double causation’ overdetermination of the shattering, so, similarly, David’s having a mass of 500 kilos and Lumpl’s having that mass do not yield a total mass of 1,000 kilos. For, if one accepts that David (a statue) exists, that Lumpl (a lump) exists, and that Lumpl constitutes David, then the fact that Lumpl weighs 500 kilos (at a given time), and that Lumpl constitutes David at that time, analytically entails that David weighs 500 kilos at that time. The principle of additivity of weights is suitable when it is restricted to separate and independent objects, but not to objects (or collections of objects) whose claims to have that weight are analytically interrelated. (Much the same obviously would go for a range of other properties for which additive principles hold in cases of separate and independent objects, e.g. extension, volume, monetary value, and so on). Once again, it seems that accepting ordinary objects does not after all violate independently plausible completely general metaphysical principles. Instead, these apparent problems, like the causal redundancy problem, are based on misapplying principles that are suitable only when restricted in certain ways, and that do not apply where the claims in question (whether to causation, space, parts, or properties) are analytically interrelated.
4.4 The Grounding Problem But worries about rivalry for space or parts, or proliferation of property instances, are not the only worries that concern those who reject colocation. Most constitution-based colocation views allow that there may be, say, a statue and a lump composed of all the same parts at the same time and having all of their microstructural and relational properties in common. Nonetheless, they are said to differ in their modal properties since, for example, the statue has the modal property of notpossibly-surviving-a-crushing while the lump has the modal property of possibly-surviving-a-crushing. The worry here is expressed in the so-called grounding problem: given all that the statue and lump (or a tree and piece of wood) have in common, what could possibly ground their different modal properties?12 One obvious line of reply is to suggest that it is in virtue of their differences in sort (one being of the sort lump, the other of the
82
ordinary objects
sort statue) that the objects have different modal properties. But this alone won’t do since, as Sidelle (2002, 121) puts it, it is ‘‘equally mysterious how such objects can differ in sort. Whatever might make some tree sortally a tree—have the identity conditions for trees—will also be true of the wood colocated with the tree.’’ Dean Zimmerman describes the problem as follows: The friends of coincident objects will no doubt say that the difference here is one of sort, and that it is simply a ‘conceptual truth’ that objects of the one sort can do things objects of the other sort cannot. But the fact remains that the mass and living body [or the lump and the statue] are supposed to differ in the sorts of physical changes they can undergo without differing in their physical construction; explaining these differences by appeal to ungrounded sortal differences is merely to insist that the two do in fact differ in these ways. (1995, 90)
Similarly, Michael Burke writes: ‘‘Statue and Piece are qualitatively identical. Indeed, they consist of the very same atoms. What, then, could make them different in sort?’’ (1992, 14)13 Thus, more generally, we can distinguish what Karen Bennett (2004, 341) calls the ‘sortalish’ from the ‘nonsortalish’ features of objects, where the former include persistence conditions (especially modal properties like being essentially of a certain shape), sortal properties, and any others that an object possesses in virtue of possessing properties of the first two kinds. With that in hand, a more general way to express the worry behind the grounding problem is that the defender of colocation has violated another plausible metaphysical principle, call it the supervenience principle: that all an object’s features supervene on its nonsortalish microphysical parts, properties, and relations. Fortunately, our earlier work on identity and persistence conditions and modal conceptualism puts us in a position to see why we should not expect such a supervenience principle to hold.14 On this view, we need not think of modal properties as higher level features of objects to be explained as supervening on their more basic (nonmodal) intrinsic or relational properties. Instead, the differences in modal truths for statues versus lumps of clay (and so the differences in the modal properties each is said to have) reflect different analyticities for the terms ‘statue’ and ‘lump’ (or the names ‘David’ and ‘Lumpl’) used in stating the question (e.g. that it is analytic that, if ‘David’ refers at all, what it refers to could not survive squashing, though the same does not go for
problems of colocation
83
‘Lumpl’). Since these differ, it is true that David and Lumpl have different persistence conditions, despite the sameness of all of their nonsortalish parts, properties, and relations; the demand for a different sort of explanation of their differing persistence conditions (in terms of underlying differences in nonsortalish parts, properties, or relations) is misguided. Still, the objector may try to push the demand for explanation back, requiring an explanation of what it is in virtue of which these ‘two entities’ could be the proper referents of different sortals. As Burke writes (using the names ‘Statue’ and ‘Piece’ rather than ‘David’ and ‘Lumpl’): It would also be unresponsive to say that identification is always under a sortal and that Statue, unlike Piece, is identified under ‘statue’ while Piece, unlike Statue, is identified under ‘piece of copper’. An object can be identified under a sortal only if it already satisfies that sortal. And the questions we want answered are these: In virtue of what does the object identified under ‘statue’ satisfy ‘statue’? In virtue of what does the object identified under ‘piece of copper’ satisfy ‘piece of copper’? Given the qualitative identity of these objects, what explains their alleged difference in sort? (1992, 14–5)
This line of criticism, however, fails to isolate the two different kinds of criteria associated with a sortal term. As I have argued in section 2.3, sortal terms come associated both with application conditions—the conditions that determine whether the sortal term is successfully applied at all in a certain situation—and with coapplication conditions that determine when it can be applied again to refer to the same thing. Burke asks in virtue of what the object identified under ‘Statue’ satisfies ‘Statue’ (and correspondingly for ‘Piece’). As always, a sortal is ‘satisfied’—or rather, in my terms, applies in a situation—if the frame-level application conditions are fulfilled in that situation. So, for example, if the application conditions for ‘piece of copper’ are fulfilled by a large number of Cu atoms tightly bonded together, then the sortal ‘piece of copper’ is satisfied by the presence of such tightly bonded Cu atoms. Similarly, if the application conditions for ‘statue’ (as seems plausible) appeal to the existence of a lump of stuff that has been manipulated intentionally by an artist intending to make a statue,15 then the fulfillment of those conditions is sufficient to ensure that the sortal ‘statue’ applies. A given world situation (e.g. there being a large number of Cu atoms tightly bonded that are manipulated by an artist intending to
84
ordinary objects
make a statue) may of course be sufficient for both sortals to apply, provided both sets of application conditions are met. But to say that these sortals both apply in the situation is not to say that there is some (one) thing to which both sortals apply. Which thing each sortal refers to is only disambiguated with the addition of coapplication conditions. As I discussed in sections 2.3 and 3.1, application conditions for a sortal must be distinguished from the coapplication conditions that determine the identity and persistence conditions for anything falling under the sortal. As Dummett’s ‘book (copy)/book (work)’ example illustrates, even if two sortals have exactly the same application conditions (which is not the case for ‘statue’ and ‘piece of copper’), still they may come with different coapplication conditions. So while we can explain the satisfaction of both sortals in terms of the world’s meeting the application conditions associated with each, we can nonetheless explain the difference in sort of the things referred to by the two names ‘Statue’ and ‘Piece’ by noting that these names are associated with sortals that (while both have application conditions satisfied in that situation) have distinct coapplication conditions that yield different identity and persistence conditions for the things (if any) that they refer to. Given these coapplication conditions for the terms in question, if the terms refer at all, ‘Statue’ must refer to something with the identity conditions of a statue, and ‘Piece’ to something with the identity conditions of a piece. But there remains another worry about this reply. Karen Bennett (2004) argues that that an appeal to analyticity does not relieve us of the need for an explanation of the different modal truths. Just because a claim is analytic doesn’t mean we can’t explain (or ask for an explanation of ) why it is true in terms of features of the world. Thus, for example, although ‘All bachelors are unmarried’ is analytic, we can still ask what it is that makes it the case that they are unmarried. Similarly, as Bennett (2004, 362) argues, the analytic claim ‘The actual man on my left [if there is one] is the man on my left’ nonetheless admits of explanation, for example, if the actual man on my left is Bob, we can say ‘‘He is a man in virtue of having a Y chromosome (etc.), and is on my left in virtue of straightforward facts about his spatio-temporal relation to me’’ (2004, 363). It is true, as these examples show, that an appeal to analyticity alone does not eliminate the possibility of explaining a claim’s truth in terms of features of the world. Nonetheless, these cases of analytic truths that seem to admit of such world-oriented explanations are importantly different from those above. In the cases above, I think there is a sense
problems of colocation
85
that the original analytic statement does not require any world-oriented explanation of its truth; the sense in which it seems we can rightly demand an explanation comes from our ability to rephrase such claims in other terms where the truth is not analytic, and can require explanation. That is, it seems that we can ask for an explanation of why all bachelors are unmarried in the sense that we could ask, of each of the men who is a bachelor: ‘Why is he (that person) unmarried?’ Similarly, it seems we can ask for an explanation of why it’s true that the actual man on my left is on my left in the sense that we can ask, of Bob, why he’s a man, and why he’s just there. In these cases, the analytic claim involves an inessential description of an independently identifiable entity, so there is the possibility of referring to that entity by other means (e.g. as ‘that person’) and asking why it fits the relevant descriptions. But the same options are not available for rephrasing the basic modal claims in question. If ‘Rocky’ is a rock-name or ‘Statue’ a statue-name, and we can only pick out the same individual by using a name for an entity of that category (which would come with the same identity and persistence conditions), we can’t rephrase questions like ‘Why couldn’t Rocky survive liquification?’ or ‘Why couldn’t David survive squashing?’ in terms that make the relevant description accidental and the demand for explanation appropriate. No further explanation is needed or legitimately required of why these claims are true than the explanation that ‘David’ simply is a name for a statue, so that if it refers to anything, it is guaranteed to refer to something with the persistence conditions analytically associated with statues. In short, then, the demand of the grounding problem to provide a ‘bottom-up’ explanation of why the referents have different modal properties is inappropriate. The supervenience principle that suggests that such explanations should be available derives its plausibility by presuming an analogy between modal (and other sortalish) properties, and macrophysical nonsortalish properties of objects (such as, say, having a certain macroscopic shape). But even if the latter do supervene on possession of micro-physical properties and relations, the work of chapter 3 makes it clear why we should not expect the same of modal and other sortalish properties. Indeed if we accept the hybrid approach to reference developed in chapter 2, we should also (I have argued) accept that the sort of entity referred to (if reference should succeed), and with it the basic identity and existence conditions and modal features of the thing (if any) referred to, are fixed analytically in fixing reference. But if sortalish properties are indeed fixed in
86
ordinary objects
determining what (if anything) our terms refer to, the demand for a deeper bottom-up explanation of these differences in sortalish properties is out of place, and the sense that it is a problem to lack such an explanation is borne of a false analogy with other—nonsortalish— properties.
C c
five
problems of vagueness
Ordinary or common sense concepts are susceptible to sorites-style arguments because of the notorious vagueness of the associated terms. That is, terms such as ‘table’ or ‘stone’ have clear cases in which they apply, and clear cases in which they do not apply, but there are also borderline cases in which it’s unclear whether or not one should accept that there is a table or a stone. Moreover, there are cases in which it is unclear whether or not the case is a borderline case or a clear case of application (or nonapplication). Such terms are often said to be ‘tolerant’ in the sense that small changes (e.g. in how many atoms are present) cannot make a difference to whether or not the term applies (e.g. in whether or not there is a table here), and it is this tolerance that apparently precludes drawing any sharp distinctions between cases of clear application and those of nonapplication or borderline application (or borderline-borderline application, etc.). The vagueness of terms for ordinary objects has figured into arguments for rejecting these objects in two sorts of ways. The most encompassing and potentially damning arguments against the existence of ordinary objects are those based on the sorites paradox.1 They are more encompassing than many of the arguments addressed above, since they are often taken to apply not only to artifacts and common sense natural objects (such as sticks and stones) but also to organisms and persons. They are also more potentially damning than many of the others, since they allege that our thought about ordinary objects is internally inconsistent, not just that accepting such objects conflicts with important metaphysical principles.2 87
88
ordinary objects
To defend the existence of ordinary objects one need only show that there is some plausible way around these problematic conclusions, and indeed a great many solutions have been offered to the problems of vagueness that can save us from the conclusions of sorites-style arguments. Nonetheless, defenders of such arguments object that those are all just ad hoc technical remedies based on altering rather than explicating our ordinary concepts. I will argue, however, that the understanding I have been developing above of the ways in which application conditions and coapplication conditions are established for our customary terms provides a way of understanding the source of vagueness that fits naturally with certain technical solutions to the problems of vagueness. This gives us reason to think that such solutions are not at all ad hoc, but rather provide promising explications of the ways our vague terms function, so that we are justified in adopting some such solution and rejecting the claim that sorites arguments should lead us to deny the existence of ordinary objects. A different form of argument based on the vagueness of ordinary terms is based not in a supposed contradiction in our beliefs about ordinary objects, but rather in the idea that accepting that our vague terms for ordinary objects refer would require us to accept that there are vague objects. Some have argued that the very idea of ontic vagueness is simply unintelligible—the only intelligible sense of vagueness being that which arises from our representations. Others have argued that vague objects would be incoherent, involve us in the problem of the many, or in accepting indeterminate identity, which is said to be impossible. I will close by discussing whether or not the solutions considered above commit us to a serious sort of vagueness in the world, and whether or not ontic vagueness is as problematic as its critics suppose.
5.1 Sorites-Style Arguments Sorites-style arguments attempt to show that our thought about things like sticks and stones, tables and chairs, involve what Unger calls ‘‘a rather blatant inconsistency’’ (1979b, 121). As Unger puts it, the inconsistency is supposed to occur among these three propositions (120): 1. There is at least one stone. 2. For anything there may be, if it is a stone, then it consists of many atoms but a finite number.
problems of vagueness
89
3. For anything there may be, if it is a stone (which consists of many atoms but a finite number), then the net removal of one atom, or only a few, in a way which is most innocuous and favorable, will not mean the difference as to whether there is a stone in the situation.
The inconsistency, of course, arises from following the allowance of proposition 3 a multitude of times, each time removing one or a few atoms, until there are no atoms left. By proposition 3 there should still be a stone in that situation, but this contradicts proposition 2, since ‘it’ consists of no atoms at all. Unger argues that the only reasonable response to this inconsistency is to deny proposition 1, and thus embrace eliminativism about stones and other ordinary objects (but perhaps not about physical objects generally, since in that case we can say that there is a physical object down to the last atom, but not thereafter). Any other response to the problem, Unger claims, requires us to believe in a ‘‘miracle.’’ This could be a ‘‘miracle of metaphysical illusion,’’ making it physically impossible to remove another atom at a certain stage, and so preventing the problematic process of removal from continuing. Or it could be a ‘‘miracle of conceptual comprehension’’ that would enable our rather coarse everyday concepts to draw a fine line determining precisely how many atoms may be removed before there ceases to be a stone: We must suppose that with, say a trillion trillion atoms there, in a certain case, there really is a stone, whether anyone can ever tell or not. But, with one or a few, say fifty, gingerly removed from the outside, the situation suddenly changes, even if no one can ever tell. . . . To believe in this is, I say, to believe in a miracle of conceptual comprehension. Thinking of our everyday thought as relevantly imprecise and unrefined, this alternative response also has little appeal for me. (1979b, 126)
In short, sorites arguments such as these rely on exploiting the vagueness of ordinary terms, for it is that vagueness that lends credence to proposition 3 and means that any attempt to deny it would require a ‘‘miracle of conceptual comprehension’’ by providing a sharp dividing line where our concept seems to provide none.3 The only plausible alternative, then, is said to lie in denying proposition 1: that there are stones, tables, or other ordinary objects. Before examining how to reply to this argument, it is worth pausing to examine the nature and source of this vagueness.
90
ordinary objects
5.2 The Source of Vagueness What is it about certain terms that makes them vague, and thus susceptible to such sorites-style arguments? Paradigm vague expressions are observational predicates such as ‘red’ or ‘noisy’. But clearly these are not the only vague expressions. For many vague expressions, including ‘valuable’, ‘mature’, and those nouns of particular concern to us here, including ‘stone’, ‘table’, and ‘statue’, are not observational at all in the sense of being applied or refused directly on the basis of human perceptual experience. Indeed most if not all expressions of natural language are vague, and the vagueness of natural language expressions also threatens to infect scientific or logical terms insofar as they are defined even in part on that basis.4 We can extract some clues about the source of vagueness by noting certain essential features in the behavior of vague predicates. First, as noted above, vague predicates are alike in that there seem to be cases in which there is no determinate answer as to whether or not they apply, and where this lack of an answer seems not a mere function of our lack of information, but rather seems to be ineliminable no matter how much other information we have about the case or how perfect our epistemic situation is (Hawley 2001, 100; Schiffer 1998, 202). This, of course, is vexing to those inclined to think of these vague predicates as referring to properties, and to accept the principle that (at least at the macro level) the world itself is fully determinate, so that for any property P, and any object x, either x has or lacks P. Second, vague predicates have other strange logical features, for example, since there are clear cases in which someone is tall, and clear cases in which someone is not tall, and a full range of cases in between, it might seem there must be some stage at which a transition occurs between its being true that the individual is tall and its not being true, so that (given a complete range) there must be some height that is the first tall one. Yet—since vague predicates lack any sharp boundaries—there is no particular height of which it is true that it is the transitional ‘first tall’ (FT) height. This is again vexing since it seems to violate the legitimate inference from: There is some x, such that FT(x), to it being true, de re of some particular height a in the domain, that a is the first that is tall. These two initial features alone are striking, since they are both features that famously arise in intensional contexts. Although it is plausible that the world is fully determinate, so that for any property P and object a, a either has or lacks P, it is certainly not the case that our
problems of vagueness
91
belief systems and concepts are complete, so it is not the case that, for any object a and property P, Jones either believes Pa or Jones believes not-Pa. Much the same, of course, goes for other intentional concepts: it also is not the case that, for any a and P, Jones accepts that Pa or accepts that not-Pa, intends that P be a or that P not be a, and so on. Similarly, it is plausible that, for any domain of objects, if it is true that there is some x such that Px, then of some nameable object in the domain, a, it is true that Pa. But this again notoriously fails for intensional contexts: Jones may believe that someone killed Smith without there being anyone of whom Jones believes that he or she killed Smith, and Jones may intend to eat a banana for breakfast without there being any banana such that Jones intends to eat it. Given these similarities, it seems worth investigating whether the failures of determinacy observed in the case of vague terms may be attributable to these terms’ applicability being in some sense dependent on human intentionality. But great care is needed in spelling out exactly how this dependence goes. Social and institutional terms, such as ‘professor’ and ‘convict’, depend on human concepts for their applicability, in the sense that they apply only if people collectively accept certain sufficient conditions for their application, and those conditions hold ( Thomasson 2003b). But ‘mountain’ and ‘stone’ are not like this— it is no part of their application conditions that people have any concepts regarding that kind or any other; indeed it seems no part of the application conditions for these terms that there be any people or intelligent life whatsoever. So in what sense can the terms’ applicability depend on human intentionality? Not in the sense that the conditions appealed to include humans having certain beliefs and practices, but rather in the sense that human intentionality sets up at least the most basic conditions for the terms’ application (and, in the case of sortal terms, coapplication). I argued in chapter 2 above that, in virtue of the qua problem, it seems that for the reference of a singular term to be established, it must come associated with a sort of entity to be named, where sortal terms come with at least some basic, frame-level application conditions establishing what it would take for the grounding of the term to succeed, and coapplication conditions establishing under what conditions the term may be used to refer to the same thing again. Although these requirements may include deference to the world in various ways, at the most basic level what it takes for the term to apply (or coapply) is determined by what grounders and other competent speakers accept as being required.
92
ordinary objects
As a result, there is a basic level at which the most basic application and coapplication conditions for our terms are established by human intentionality. These conditions for our terms, as I have argued in chapter 3, set up the truth-conditions for all metaphysical claims about the existence, identity, and persistence of the entities (if any) referred to by these terms, and so also fix the most basic conditions of existence, identity, and persistence governing the things (if any) these terms refer to (as well as determining the spatial and temporal boundaries of the individual—if any—referred to). While I have focused on names and sortal terms in the work above (given the focus here on ordinary objects), it is reasonable to suppose that although nonsortal predicates such as ‘tall’, ‘mature’, ‘red’, and so on lack coapplication conditions, they, too, come associated with basic, frame-level application conditions that enable competent speakers to evaluate various hypothetical situations as ones in which ‘red’ would or would not be properly applied (thus distinguishing it from ‘shiny’, ‘round’, ‘sweet’, and the endless other predicates that might be grounded in the same situation). These, like other application conditions, would play a crucial role in establishing the truth-conditions for sentences using such terms. But the fact that such conditions are determined by human definition or stipulation brings risks of indeterminacies. Kit Fine (1975/ 1997, 120) provides a useful artificial example of a stipulated technical definition5 of ‘nice1’ as: (a) n is nice1 if n > 15 (b) n is not nice1 if n < 13
Like ‘nice1’, ‘stone’ clearly applies to certain cases, thus it would seem wrong to conclude there are no stones or nice1 numbers. But it would equally be wrong to label ‘14 is nice1’ true or false—the term ‘nice1’ is, in Fine’s terms (120), simply deficient in meaning, making it impossible to determine whether or not 14 falls under that predicate, where this impossibility seems not to be a consequence of being in a subpar epistemic situation. Now this may seem an unduly artificial example, but such risks are rampant, and their effects frequently observed, in real cases where it is explicit that human concepts stipulate the application conditions for a term. Thus, for example, suppose (counterfactually) that the rules of baseball specified that, if first base is tagged before the batter reaches the base, the batter is out; if the batter reaches first base before it is tagged, the batter is safe (not-out). In that case, ties would be left completely indeterminate, so that no further investigation could possibly show
problems of vagueness
93
whether, in cases of a tie, the batter was out.6 Similarly, suppose that according to election rules, if a ballot in a presidential election in Florida contains a hole punched by a candidate’s name, that is a vote, and if no hole is punched, then that is not a vote. In such cases, it clearly becomes a matter of stipulation, not discovery, whether or not to ‘count’ hanging chads and their like as cases of votes. Since these are explicitly institutional terms, it should be no surprise that their application conditions are determined by the conditions we accept as relevant to their application, so if there are cases that (e.g.) the stipulators have not thought of, and not considered whether or not the term applies, there is no fact of the matter about whether or not the term applies in those conditions. Thus even someone in ideal epistemic circumstances with knowledge of all the relevant precise facts could not say whether the term applies, and so such questions, if they are to be resolved at all, must be resolved by decisions, not discoveries.7 But what about noninstitutional ordinary terms? It is plausible that at least many of these (‘tall’, ‘red’, ‘table’, etc.) are defined at least in part ostensively rather than (like the rules of baseball or democracy) through explicit stipulation. Wherever applicability of concepts is determined by ostension of paradigms (things colored like this are red, things colored like that are not red), we will have a parallel situation, as there may be many things not closely resembling either paradigm, for which the applicability of the predicate ‘red’ is as undetermined as the applicability of the artificial term ‘nice1’ to the number 14. On this view, although the application conditions for our natural language terms may take many different forms, they are all determined at least in part, at the most fundamental level, by what application conditions we accept. And since the conditions we accept as relevant for application of a predicate and/or its negation may be very incomplete, there may be many cases in which the applicability is simply indeterminate, and no further facts of the world could possibly decide the issue. If we must provide a determinate answer for some practical reason (e.g. in the case of elections), such answers are explicitly a matter of decision and stipulation, not of discovery.8 If we accept that application conditions for our terms are determined, at least in part, by human intentions, concepts, and so on in this way, then we can explain the indeterminacies in their application in terms of the incompleteness of human intentions. But one critical feature of vagueness has not yet been explained. There is a crucial way in which Fine’s example differs from true cases of vagueness: Although there are cases of indeterminate application for his predicate ‘nice’,
94
ordinary objects
there is a sharp boundary (at 15) between cases in which it is indeterminate whether or not the predicate applies and cases in which it applies determinately. But this is not the case for truly vague predicates, for (as we observed above) they suffer from a more general boundarylessness—there are no sharp boundaries anywhere; not between cases of applicability and nonapplicability, applicability and borderline applicability, applicability and borderline-borderline applicability, and so on all the way up. Cases such as Fine’s are thus sometimes labeled (e.g. Schiffer 1998, 199) as terms that are merely incompletely defined, not vague.9 Of course in some cases, the new stipulated definition may involve such vague terms that complete boundarylessness is readily apparent, for example, if I define ‘Thomasson green’ as any color closely enough resembling this one. But in others another explanation of full boundarylessness (as opposed to mere indeterminacy) is needed. Moreover, even in the stipulated cases in which full boundarylessness is apparent, this apparently relies on the presence of other vague terms used in the stipulation, and so that cannot provide any bottom-up account of the general source of higher-order vagueness. How can we explain this higher-order vagueness? Unlike the technical term ‘nice’ above, which has (incomplete) application conditions explicitly stipulated by its official definer, on the view I have argued for above, the central cases of notoriously vague terms are those whose definitions are not explicitly and officially stipulated by some authoritative source, but rather are defined through common practices in applying, reapplying, and refusing the terms. Call terms defined through practice ‘customary’ and those defined through explicit stipulation ‘technical.’10 As I have argued in chapter 2 above, those who ground the reference of a singular or sortal term must do so by associating the term with some frame-level application conditions (which may also appeal to various further real-world conditions to be filled out). These set up the conditions that must be satisfied if the term is to be successfully grounded at all—and establish under what conditions it may be successfully applied, leaving open possibilities for areas of indeterminacy in application. But for customary terms it is also a vague matter who counts as a ‘grounder’, which attempts at reference count as grounding or regrounding the term, how many would-be grounders must associate the term with the same frame-level application conditions for these to be successfully established as the application conditions for the customary term, and so on. Even if (counterfactually) each speaker associated perfectly precise application conditions with the term ‘stone’ (one
problems of vagueness
95
counting anything above 5 millimeters a stone, another counting only those things above a centimeter . . .), as long as these individuals vary among themselves, as seems inevitable with such predicates, it will be vague where the borders of application conditions for ‘stone’ are. Of course, competent speakers do not associate precise application conditions for the use of terms such as ‘stone’; in fact Tye (1994/1997, 289) suggests that even an individual speaker might (confronted with a gradual array of cases) utilize different boundaries on different occasions.11 The failure even of individual speakers to use the term precisely is a reflection of the fact that successful induction into the practice of using the term ‘stone’ also includes recognizing it as a customary term that it is inappropriate to treat as having sharp boundaries. So while the cases of indeterminate application can be explained by the expressions’ conditions for application being determined by conditions accepted as relevant to their application, the full boundarylessness characteristic of vagueness can be explained by the fact that the base-level application conditions for customary terms are determined not by a single definitive stipulation but rather by common use, where it is in turn vague exactly what counts as ‘common use’. This explains why vagueness seems an inevitable part of a natural language composed of customary terms, and why, indeed, it is exactly what we should expect.
5.3 Solutions to the Problems of Vagueness At least two prominent lines of solution to the problems of vagueness— supervaluational solutions and indeterminist solutions—fit well with the idea that the phenomena of vagueness result from deficiencies in meaning based on the fact that the basic conceptual content of terms is established by human intentionality through common usage. Supervaluational treatments of vagueness take seriously the idea that vague predicates are genuinely imprecise, where this imprecision is blamed on the fact that, in these cases, we simply haven’t bothered to make these terms precise through our concepts or usage (Lewis 1986, 244), although we could make them more precise in a variety of ways if we so chose. To evaluate the truth of a claim using a vague predicate, we must evaluate the truthvalue the claim would have on any of the acceptable precisifications. If ‘x is a stone’ is true on every such precisification, it is supertrue (true simpliciter); if it is false on every such precisification, it is superfalse; and if it is true on some precisifications and not on others, then it is neither
96
ordinary objects
true nor false. On any legitimate precisification, there will be some boundary between stones and nonstones, thus it is supertrue that there is some point at which the net removal of one atom, or only a few, will make a difference to whether there is a stone in the situation. Thus supervaluational views avoid sorites-style arguments by treating proposition 3 as false. Nonetheless, since there is no particular boundary that is accepted on all precisifications, it is not supertrue of any number of atoms that removing one more will turn a stone into a nonstone, and so no particular boundary is endorsed. Such supervaluational approaches do well at blocking sorites arguments while respecting the incompleteness of our concepts associated with vague terms. They also nicely map the odd quantificational structure according to which it is true that there is some changeover between cases of application and nonapplication, but not true of any point that it is the boundary. Thus they seem to provide suitable formal ways of handling the phenomena of vagueness as described above, and have the virtue of preserving the idea that the law of the excluded middle is a logical truth, since it remains true on any precisification. There is some reason for worry, however, about whether simple supervaluational approaches preserve the full boundarylessness that seems to inevitably characterize vague terms. For on a simple supervaluational proposal, if one assumes that there is a fixed and precise set of legitimate precisifications of the concept, sharp boundaries still arise between cases in which it is supertrue that n atoms make a table (since all legitimate precisifications include it), and not-supertrue (since one legitimate precisification does not). To avoid sharp boundaries here, one must also accept that it is a vague matter what counts as a legitimate precisification. If we think of legitimate precisifications as those that entail judgments about the predicate’s applicability and nonapplicability that are commonly accepted, then we can see the vagueness of what counts as a legitimate precisification, in turn, as having its source in the vagueness of what counts as being commonly accepted by grounders of the term’s reference. Indeterminist solutions developed, for example, by Michael Tye (1994/1997), also provide responses to sorites arguments that fit naturally with the understanding developed in section 5.2 of the source of vagueness. These solutions allow that there are truth-value gaps, so that while some statements are true and others false, others are indefinite or lacking in truth-value. This thus enables us to handle the indeterminacies in application that may result whenever application conditions are determined by conditions we accept, by treating such cases as simply
problems of vagueness
97
indefinite—so it is indefinite whether or not the number 14 is nice1, and whether or not partially-punched cards are votes. Similarly, there will be many cases in which ‘x is a table’ or ‘x is a stone’ are indefinite in truthvalue. According to the natural formalization Tye offers, universally quantified statements are true if they are true for all substitution instances, false if they are false for some substitution instance, and otherwise indefinite. So if one allows that there are sentences of the form ‘x is a stone’ that are indefinite, one should allow that statements of the form of proposition 3 (‘‘For anything there may be, if it is a stone (which consists of many atoms but a finite number), then the net removal of one atom, or only a few, in a way which is most innocuous and favorable, will not mean the difference as to whether there is a stone in the situation’’) are indefinite. For, putting it more formally, proposition 3 asserts: For all situations x, if x is a situation containing a stone of n atoms, then a situation otherwise identical but with n-1 atoms contains a stone. But there are situations in which it is indefinite whether or not there is a stone, and where a situation differing only in having one atom fewer is likewise indefinite with regard to whether or not there is a stone. Thus there are instances in which the conditional is indefinite, and so the universally quantified statement is indefinite.12 As a result, Tye’s indeterminist solution renders proposition 3 not true but indefinite, since (although it is never false), some substitution instances will make both the antecedent and consequent indefinite, making the whole conditional indefinite (288). Thus we avoid the contradiction of the sorites without accepting sharp boundaries or attributing miraculous precision to ordinary concepts. Critics, however, might try to force the issue by moving up to the metalanguage. So consider a sorites sequence of sentences M1,000,000,000 (‘There is a stone of 1,000,000,000 atoms here’) to M0 (‘There is a stone of 0 atoms here’). Some sentences in the sequence are true, some are false, and some are indeterminate in truth-value. But then it might seem that sharp boundaries reemerge, for there must be some pair of statements, Mk and Mk-1, such that Mk is true and Mk-1 is not true. Tye avoids this problem by noting that boundarylessness must not only infect predicates of the object language, but go all the way up through levels of semantic ascent. So the claim ‘There is some pair of statements, Mk and Mk-1, such that Mk is true and Mk-1 is not true’ is itself not true (it is indefinite, as is its negation) (1994/1997, 288–9). One might think it can’t be indefinite, since that would require some sentence of the form ‘Mk is true’ to be indefinite, but (it might seem) that cannot be, since, if ‘Mk’ is true, ‘Mk is true’ is true, and if ‘Mk’ is false or indefinite, ‘Mk is true’ is false—so that either way, it’s not
98
ordinary objects
indefinite. That, however, relies on the assumption that any sentence has one of exactly three truth-values (true, false, or indefinite)—but on Tye’s view there is no determinate fact of the matter about whether that is so, and so no determinate fact of the matter about whether there are statements of the form ‘Mk is true’ that are indefinite (290). The truth-value predicates themselves are vaguely vague, that is, ‘‘there simply is no determinate fact of the matter about whether the properties they express have or could have any borderline instances. So, it is indefinite whether there are any sentences that are neither true nor false nor indefinite’’ (290). Similarly, in the higher level metalanguages, it must be never true that transitions from one truth-value to another are sharp. This again fits naturally with the view that, if the frame-level application conditions of ‘stone’ are determined by what those who ground and reground the reference of the term commonly accept as its application conditions, then (given the vagueness of ‘commonly accept’), it may also be indefinite what exactly the associated conditions for stonehood are (and thus what the meaning of ‘stone’ is). As a result of the fact that it is indefinite precisely what the associated conditions are, it may be indefinite whether or not it is indefinite whether or not a given situation meets the associated conditions—yielding the characteristic borderline-borderline cases. Given the ways in which metalanguage statements involving semantic predicates like ‘true’ and ‘false’ are linked to object-language statements, if there is genuine boundarylessness at the object level for the reasons we have outlined, we must accept that there is boundarylessness all the way up through layers of metalanguage. For otherwise, as we have seen, accepting sharp transitions in truth-values at a higher level (i.e. between sentences of the form ‘Mk is true’ which is true, and ‘Mk-1 is true,’ which is not true) will entail that there are sharp divisions below (making Mk true and Mk1 not true, thus introducing a sharp boundary). Indeterminist views and supervaluational views alike are sometimes criticized for attempting to avoid sorites arguments by way of mere technical solutions that are insufficiently motivated (e.g. Keefe and Smith 1997, 55)—especially considering that they require us to abandon bivalence. Even Tye himself notes that his ‘‘logical apparatus needs to be associated with a view about the locus of vagueness’’ (1994, 16). But if the arguments of chapters 2 and 3 were correct, the defender of ordinary objects who takes the terms for things of these kinds to be customary terms has a rather natural explanation of the vagueness of the terms for ordinary objects that can show why it is
problems of vagueness
99
perfectly natural, not merely an ad hoc technical maneuver, to allow that deficiencies in meaning may lead to truth-value gaps and require us to abandon bivalence. Bivalence may be a very natural and useful presupposition for carefully defined artificial languages, and might be suitable if the boundaries between when terms do and do not apply could be established not at all by human beliefs and practices but rather purely by mind-independent features of the world. But I have argued above that reference cannot be established without human concepts playing some role at the basic level; and when application conditions are determined, even in part, by the conditions people accept as relevant to their application, risks of indeterminacy result. More important, when we note that, for customary terms, this assignment of application conditions is done through common practice rather than explicit authoritative stipulation, we can see why our customary terms suffer not just from simple indeterminacies in application, but from the complete boundarylessness characteristic of vagueness. The truth-value gaps posited by supervaluational and indeterminist theories of vagueness are precisely what we should expect, since they follow from the role of human beliefs and practices in establishing the application conditions of customary terms (along with identity and persistence conditions for their referents). (Insofar as customary terms must be appealed to when we define expressly stipulated technical terms, vagueness will also infect technical terms, however carefully one attempts to stipulate their definition.) Given the availability of natural and suitable responses to sorites arguments provided by supervaluational and indeterminist solutions, those who accept the work of chapters 2 and 3 above need not fear that such arguments should lead us to deny the existence of ordinary objects. But which of these responses should we adopt? Each has familiar advantages and disadvantages, and we need not decide the issue here.13 But one apparent disadvantage of Tye’s treatment does need to be faced here: that apparently it (unlike supervaluational treatments) requires us to accept that there are vague objects in the world. Tye makes it explicit that his handling of vagueness (unlike most of its competitors) involves positing ontic vagueness, since it requires accepting the existence of vague sets as the extensions of vague predicates (1994/1997, 283), where a set is vague, on Tye’s view, just in case ‘‘(a) it has borderline members and (b) there is no determinate fact of the matter about whether there are objects that are neither members, borderline members, nor non-members’’ (284). Supervaluational treatments of vague statements provide a way of understanding how
100
ordinary objects
claims like ‘‘there is a table in the next room’’ could be true without positing vague objects. For one can hold that there are, in the world, only precise objects (say, precise collections of particles), and the vagueness of ‘table’ is just a matter of semantic indecision about which of those our term ‘table’ is to refer to. Even if there are ways of avoiding sorites arguments that do not involve positing vague objects, those who accept that our ordinary vague terms such as ‘table’, ‘chair’, ‘stick’, and ‘stone’ refer to objects seem compelled to accept that they refer to vague objects. For if we accept only that there are many different precise objects (e.g. collections of particles) wherever we would say there is a table, treating one of these (but not the others) as the referent of our standard English word ‘table’ would seem quite arbitrary, while treating all as tables would land us firmly in the grip of Unger’s (1981) problem of the many. The correct response seems to be to hold that if we accept that there are tables, stones, mountains, and other ordinary objects, we should accept that they are vague objects (see Johnston 1987). It has often been argued, however, that vagueness in the world is impossible or even unintelligible; that the only intelligible treatment of vagueness locates it in our ways of representing the world, not the world itself (Evans 1978/1997; Horgan 1994, 1998; Horgan and Potrcˇ 2000; Lewis 1986; Russell 1923/1997, 62). So the would-be realist about ordinary objects not only has to block sorites arguments, she also must confront a variety of arguments that there are, or can be, no vague objects.
5.4 Vagueness and Contextual Semantics Terence Horgan (1994, 1998) and Matjazˇ Potrcˇ (Horgan and Potrcˇ, 2000) argue that strictly speaking there cannot be vague objects, based on the view that vagueness involves us in a sort of weak logical incoherence that cannot have an ontological counterpart.14 So, while Horgan allows that under some (normal) standards, vague statements may be true, nonetheless, he argues, when engaged in serious ontological discussion we should deny that there are vague objects. On Horgan’s ‘transvaluationist’ analysis of vagueness, the proper use of vague terms imposes on us an inconsistent set of semantic constraints. On the one hand, the boundarylessness characteristic of vague terms imposes an individualistic requirement that any two adjacent statements in a sorites sequence have the same semantic status (i.e. both
problems of vagueness
101
are true, or both are false, or both are indefinite) (Horgan and Potrcˇ 2000, 261). On the other hand, there is the collectivist requirement that taken as a whole, the statements in a sorites sequence cannot all have the same truth-value, since there are clear cases of application and clear cases of nonapplication for vague predicates. These demands cannot be satisfied simultaneously, nor can either be defeated. Nonetheless, the collectivist requirement ‘dominates’ the individualist requirement, leading us to accept that there are some clear cases of application, and some of nonapplication, and to reject the universally quantified conditional (proposition 3) as not-true. Various ways of handling the logic of vagueness may respect this semantic situation (e.g. Tye’s, 1994; Horgan’s own similar proposal, 1994; and iterated supervaluationism, Horgan, 1998) and enable us to forestall sorites arguments. Nonetheless, the tension between the competing semantic requirements still comes to the fore if we are taken on a ‘forced march’ through a sorites sequence, asked one at a time such questions as: Is a situation with n atoms a situation in which there is a stone? Is a situation with n-1 atoms a situation in which there is a stone? and so on. While the proper reaction to such a forced march is simply to refuse to answer (Horgan 1998), they remain as meaningful questions, and so the need to refuse them is just another manifestation of the tension between semantic requirements that lingers even when contradiction is avoided. Since the incoherencies of vagueness can be quarantined in this way, Horgan argues, we needn’t embrace nihilism or reject all vague discourse. Statements uttered in ordinary contexts involving vague terms such as ‘stone’, ‘table’, ‘bald’, and the rest may even be true although strictly speaking there are no vague objects, for Horgan maintains that truth is a sort of ‘correct assertability’ under whatever standards are operative in the relevant context. In fact, Horgan and Potrcˇ (2000, forthcoming) maintain that strictly speaking, there is only one concrete particular, the ‘blobject’ that is the whole universe (with all of its complexity and local variations), so that strictly speaking there are no medium-sized material objects, social objects, or other ordinary objects at all. Nonetheless, given their contextual semantics, they hold that statements made about such things in everyday contexts may be true. So when, for example, we make claims like ‘‘In summer of 1999, NATO was conducting a massive bombing campaign against targets in Serbia and Kosovo’’ (2000, 214), the contextually relevant standards require only indirect correspondence to the world. Thus such sentences may still be ‘semantically correct’ (true) as
102
ordinary objects
long as they track ‘‘real local spatiotemporal variation in the blobject’’ (2000, 214). The diversity of variations in the blobject may contribute holistically to making statements like those above true, even though they involve ‘‘posits (e.g. NATO, Serbia) that do not designate genuine denizens of reality’’ (254). According to Horgan, we have independent reason for thinking that, under normal conditions, truth is a matter of indirect correspondence— for statements involving such terms as ‘university’, ‘concerto’, ‘government agency’, and ‘book’ (used to refer to a type rather than token book) also seem capable of truth or falsehood, although ‘‘it is plausible that the mind-independent, discourse-independent, world does not include among its real constituents such putative entities’’ as universities, concertos, agencies, and book types (1998, 320), nor are the sentences paraphrasable into statements that plausibly do refer to real objects and properties. While indirect correspondence may be good enough to make the assertions of our ordinary discourse true, Horgan holds that higher standards come into force when we engage in serious ontological discussion. In the latter context, the operative standards require that there be direct correspondence between, for example, the nouns and predicates of our language and the objects and properties in the ‘‘mindindependent, discourse-independent, world’’ (1994, 99). But, given the weak logical incoherence that is characteristic of vague discourse, it cannot directly correspond to the world, ‘‘for, this would mean that the world itself would have to exhibit the same logical incoherence that is present in vagueness; and this is impossible’’ (2000, 265). Horgan’s argument thus provides a different sort of vagueness-based reason for denying (in serious ontological contexts) the existence of ordinary objects: it is not that accepting such objects inevitably brings us into contradiction, but rather that there can be no objects (in the mind-independent world) that exhibit the ontological correlates of the weak logical incoherence of semantic discourse (1998). So, given that strict ontological discourse demands that there be direct correspondence between our noun terms and ‘‘denizens of the mind-independent, discourse-independent, world’’ (1994, 99), according to such strict discourse there can be no such objects. This raises two important questions, which I will treat in turn: first, Do such higher semantic standards make sense? and second, Can there be vague objects? What are we to make of the requirement that, when doing ontology, the proper standards demand direct correspondence between our terms and the ‘real’ objects and properties that are part of ‘mind-
problems of vagueness
103
independent, discourse-independent reality’—standards that claims, for example, about Serbia, NATO, or even tables and rocks do not meet? As the work of chapter 3 suggests, there are at least two senses in which we can talk about the referents of our terms being mind- and discourse-independent. First, they may be independent in the sense that the application conditions for the terms make no appeal to minds, mental states of any kind, or discourse of any kind, so that it is possible that the terms apply even in worlds in which there are no minds or discourse. This sort of independence condition is clearly not met by such social and cultural terms as ‘university’, ‘concerto’, or ‘government agency’, but is met by such (nonetheless vague) terms as ‘mountain’ and ‘stone’, so entities of the latter sort cannot be ruled out from consideration as ‘real’ constituents of the world on that basis. A second sense in which one could (somewhat misleadingly) talk about the referents of our terms being mind- and discourse-independent is a sense in which—independently of all minds, discourse, and so on— there are facts of the matter about the identity, persistence, and boundaries of the putative referents of the terms. This seems to be more the sense that Horgan has in mind in using the supposed failure of independence to rule out mountains and stones. Indeed talk about the need for direct correspondence between objects and properties in the world seems to suggest that the requirement is for our nouns (e.g.) to pick out objects that are individuated independently of any of our decisions about how terms are to be used, and are (in that sense) not just ‘‘artifacts of our conceptual scheme’’ (Horgan, 1994, 103). But this again could be interpreted in two ways. If the requirement is that there may be claims about the identity, persistence, and distinctness of various objects that are true independently of all minds, then (as I have argued in chapter 3) this requirement can perfectly well be met by our discourse about ordinary objects. So claims about many ordinary objects (e.g. sticks and stones) may meet the ‘mind-independence’ condition for direct correspondence both in the sense that the objects they refer to could exist even in the absence of all minds and language and in the sense that claims about the identity, distinctness, and persistence of such objects may be true (or false) regardless of what anyone thinks or believes about them. On the other hand, we might take the mind-independence requirement to demand that even the most basic conditions of identity and persistence for the objects we refer to are purely discoverable features of the world (rather than being fixed in fixing what sort of entity our terms are to refer to). But if that is what Horgan has in mind as a
104
ordinary objects
requirement for objects to not be ‘artifacts of our conceptual scheme’, then—if the arguments of chapters 2 and 3 are correct—the supposed ‘maximally strict’ standards demand the impossible, and the ideal of ‘direct correspondence’ cannot be a standard that is suitably applied to ontological discourse or any other sort of discourse. For, as I have argued in chapter 3, at the most basic, frame level, identity and persistence conditions for the objects we refer to are established in disambiguating reference, and may be read off of the determination of reference. So we cannot legitimately require that all identity and persistence conditions for the objects we refer to be independently discoverable features of the world (rather than features discoverable through conceptual analysis) for our statements to be true in the most rigorous (ontological) contexts. We are left then with the weaker standards of ordinary discourse, and as long as we are using these, Horgan and I agree: there are mountains, tables, universities, and other ordinary objects.
5.5 Is There Vagueness in the World? But although the work of the prior chapters gives us tools with which to resist Horgan’s arguments, other worries remain about accepting that our vague terms refer to vague objects. The above view clearly does seem to posit vagueness in the world, in the sense that it seems that if we accept that ordinary predicates such as ‘red’ and ‘tall’ sometimes apply, and that ordinary terms such as ‘that table’, ‘Mount Everest’, and so on sometimes refer, then we may also have to accept vague properties and objects, since vague predicates would seem to pick out (if anything), vague properties, and vague singular terms (‘that table’ and ‘Mount Everest’) seem suited to pick out (if anything) vague objects (see Tye 1994, 2). After all, we can easily transform statements about claims with indefinite truth-values, for example, ‘It is indefinite whether this apple is red’ to statements apparently about objects and properties, for example, ‘This apple is such that it has the property of being-indefinite-with-respect-to-redness’ (see Schiffer 1998, 212). Such transformations are formally conveyed by lambda abstractions, so that from ‘r(Pa)’ (it is indefinite whether a is P) we may derive ‘lx(r(Px)(a))’ (a has the property of being-an-x-such-that-it-is-indefinitewhether-x-is-P), thus leading to an apparent de re commitment to vagueness—at least with regard to the property of being P.
problems of vagueness
105
Some would say, however, that the trivial sense in which we can derive claims that there are objects such that they are indeterminate with respect to some properties is not in itself enough to establish commitment to vague objects (but at most to vague properties). Tye, for example, requires compositional vagueness for true ontic vagueness, noting that for there to be ontic vagueness it must be the case that ‘‘there is at least one object, x, that satisfies ly 9 z(r(z is a part of y)x),’’ that is, there must be some object that has the property of beingindeterminate-with-respect-to-whether-some-object-is-part-of-it (2000, 198).15 More commonly, it has been held that a thesis of genuine vagueness in the world requires positing objects that are indeterminate with respect to identity. Nonetheless, we can perform similar transformations to get ontic vagueness claims of both of these sorts, moving, for example, from ‘It is indeterminate whether or not this rock is part of Mount Everest’ to ‘Mount Everest is such that it is indeterminate whether or not this rock is part of it’ (making Mount Everest compositionally indeterminate); or from ‘It is indeterminate whether ship S1 is identical to ship S2’ (in a case of gradual plank replacement) to ‘Ship S1 is such that it is indeterminate with respect to the property of being-identical-to-ship-S2.’ Mark Sainsbury (1994) argues, however, that accepting vague objects merely in the sense above—as satisfiers of vague predicates (whether any predicates will do, or one restricts them to predicates regarding composition or identity) that can be formed by property abstraction from claims of indeterminacy—is not enough to secure a ‘substantive’ thesis of ontic vagueness. For ‘‘theorists of all persuasions, if they believe that vagueness exists at all, should accept that there are . . . objects which, when a vague expression is applied to them, are implicated in the phenomena of vagueness’’ (63). Vagueness of this sort, Sainsbury argues, is a mere ‘anodyne’ sort of vagueness in the world, since it can still be explained as a consequence of the vagueness of our linguistic (or other) representations. On this view, ‘‘every property of representations induces a property of nonlinguistic things: for every linguistic property, an ontic one’’ (66), and so the supposed vagueness in objects (arrived at after the transformations) is just a trivial consequence of the initial vagueness in our representations (as described before the transformations), and so even someone who (like Russell) believes that vagueness resides in our representations, not the world, can accept vagueness in this sense (67). Tye rejects Sainsbury’s claim that vagueness in the above sense can be explained in terms of vagueness in our representations. In particular,
106
ordinary objects
Tye takes compositional vagueness to establish a genuine thesis of ontic vagueness—that is, there is ontic vagueness if ‘‘there is at least one object, x, that satisfies ly 9 z(r(z is a part of y)x)’’ (2000, 198). But compositional vagueness, Tye argues, cannot be explained by the vagueness in the term ‘part of’, since ‘part of’ may apply to objects without their being vague. In a world with perfectly precise objects, it seems ‘part of ’ would apply precisely, and so ‘‘we cannot explain the fact that Everest is compositionally vague . . . simply by noting that ‘part of ’ is vague’’ (199). But we can acknowledge that the vagueness in the application of ‘part of’ may be a result of the vagueness in the relata (not in the potentially precise part of relation) while still noting that the vagueness in the relata may be a trivial consequence of the vagueness in our ways of representing the relata. That is to say, as I have argued in chapter 2, whether we attempt to refer to individuals by way of names, demonstratives, or descriptions, attempts at establishing reference to an individual must involve frame-level application conditions and coapplication conditions that determine the broad (ontological) category of thing that the expression will refer to, if it succeeds in referring at all. Above I have been discussing the indeterminacies in application conditions accepted that lead to the boundarylessness of vague general terms. But the most basic criteria of identity and persistence for the things (if any) referred to (as I have argued in chapter 3) are also fixed in grounding the reference of singular terms, and may be indeterminate for the same sorts of reason that application conditions and coapplication conditions may be. It is the indeterminacies in the criteria of identity and persistence associated with these categorial terms that can lead to certain claims about the identity or composition of the objects referred to being indeterminate in truth-value. And claims that it is indeterminate whether these are parts of the object, or whether the object still exists, or whether these objects are identical . . . may then be trivially transformed into claims about objects having vague boundaries, and so on. So if it is indeterminate whether or not a certain rock is part of Everest, this can be traced to the indeterminacy in the individuative conditions associated with the categorial concept of ‘mountain’ (or ‘geological feature’) associated with the name ‘Everest’, and thus can still be traced to our ways of representing the world. As a result, it seems to me that the kind of vagueness in the world that we must posit to accept ordinary objects on the view I have been developing is indeed a kind of anodyne vagueness that still explains the vagueness in our world in terms of vagueness in our representations,
problems of vagueness
107
rather than the other way around (pace Tye 2000, 208). Thus this thesis is not in substantial, deep, conflict, with Russell’s claim that ‘‘apart from representation . . . there can be no such thing as vagueness or precision’’ (1923/1997). There is, I think, vagueness in the world in the only sense that we should have ever expected there to be, but this vagueness is still traceable to indeterminacies in our language and ways of representing the world.16
5.6 Can There Be Vague Objects? Nonetheless, as with nuclear power plant safeguards, so in philosophical argumentation, redundancy is often advisable. So, just in case there are readers unconvinced by the above arguments that the ontological vagueness described above is anodyne, we might ask: Supposing that the above solution did require us to accept a ‘serious’ kind of vagueness in the world, would that present any serious problems? Perhaps the most discussed problem with accepting the existence of vague objects comes from Evans’s (1978/1997) argument that there cannot be indeterminacy of identity. His proof runs as follows, using r for ‘‘it is indeterminate whether,’’ D for the determinacy operator, and substituting (for Evans’s original notation) the notation of lambda abstraction, so that lx(r(x ¼ a)(b)) means ‘b has the property of beingan-x-such-that-it-is-indeterminate-whether-x-is-identical-to-a’: 1. Suppose that r(a ¼ b) then 2. lx(r(x ¼ a)(b)) but 3. r(a ¼ a) so 4. lx(r(x ¼ a)(a))
so, by Leibniz’s law we may conclude: 5. (a ¼ b)
From which, he suggests, we may conclude: 50 . D (a ¼ b)
Thus, from the assumption that it is indeterminate whether a ¼ b, we are brought to the conclusion that, determinately, it’s not the case that a ¼ b. Many questions have arisen regarding how, exactly, Evans’s argument should be interpreted (including how the determinacy and indeterminacy operators function) and whether or not it is valid.17 Johnsen (1989, 106) provides a charitable and detailed reconstruction
108
ordinary objects
of Evans’s argument using a three-valued logic much like Tye’s (at least at the level of the object language), and defining the operator ‘r’ (‘it is indeterminate whether P’) such that ‘rP’ is false if P is true, false if P is false, and true if P is indefinite, and defining the operator ‘D’ such that ‘DP’ is true if P is true, or if P is false, but false if P is indefinite:18 (1) (2) (3) (4) (5)
r(a ¼ b) lx(r(x ¼ a))(b) r(a ¼ a) lx(r(x ¼ a)(a)) (a ¼ b)
(6) D (a ¼ b)
assumption from 1 by abstraction by logic of ¼ and r from 3 by abstraction from 2, 4, and Leibniz’s law (LL) by modus tollens by definition of determinacy operator
This interpretation makes the Evans-style argument apparently valid, provided that the relevant application of Leibniz’s law is acceptable. But Johnsen ably points out that we have independent reason to think that, in the context of a logic like Tye’s, we cannot accept an unrestricted version of Leibniz’s law any more than we can accept unrestricted versions of the law of the excluded middle or other tautologies of classical logic. The law of the excluded middle, for example, can only be accepted in a restricted version, since ‘P v P’ is indefinite if P is indefinite (see Tye 1994/1997, 282). We can, at least, say that the classical law of the excluded middle is never false, and we can accept an altered version that respects its holding in classical twovalued logics by noting that it still holds provided P is determinate in truth-value: ‘DP (P v P)’ ( Johnsen 1989, 108), and in general we can rely on the principles of classical logic in many-valued systems provided their application is restricted to those involving only statements determinate in truth-value. Similarly, we have independent reason to think that Leibniz’s law does not hold if we use a logical system like Tye’s. For consider a case of some wavelength of light that is indeterminate in color between being blue and green (G). Leibniz’s law should guarantee that the following is determinately true: (a ¼ a) ½lx(Gx)(a) lx(Gx)(a) But in fact, the latter biconditional is indeterminate, since both sides are indeterminate, making the conditional as a whole indeterminate in truth-value ( Johnsen 1989, 107–8). So in the context of such manyvalued logics, we have independent reason to restrict Leibniz’s law,
problems of vagueness
109
though we can still legitimately hold that it (like the law of the excluded middle) is never false, and that it is always true in those cases in which all the claims involved are determinate in truth-value. That is, we can accept the restricted version: LLv: 8y8z([D (y ¼ z) & Dlx(Fx)(y) & Dlx(Fx)(z)] [(y ¼ z) lx(Fx)(y) : Dlx(Fx)(z)]. But this restricted version of Leibniz’s law clearly won’t do the job needed in Evans’s proof, since we do not have it that D(a ¼ b) as the antecedent requires. As a result, it seems that even if (contrary to the arguments above) the position I have been arguing for commits us to the existence of vagueness in the world in a ‘serious’ way, we need not worry unduly about accepting that there are ordinary objects that are vague in the above sense. So whether we are concerned about the vagueness of ordinary terms and their susceptibility to sorites arguments, or about the vagueness of their apparent referents—the ordinary objects—the problems with vagueness do not give us reason to reject the existence of ordinary objects.
C c
six
handling existence questions
The arguments we have examined so far against accepting ordinary objects, including the causal redundancy argument, anti-colocation arguments, and arguments from vagueness, all concern what might be called ‘internal’ problems with an ontology of ordinary objects. That is to say, all involve problems that are supposed to arise for anyone who accepts ordinary objects, since doing so is supposed to lead either to internal conflict (at worst, self-contradiction) or to a conflict with what are supposed to be independently plausible general metaphysical principles. The arguments we have yet to consider take a somewhat different form. The arguments from composition, rivalry with science, and parsimony each are based on taking a global approach, and arguing that we can provide a better ontology overall if we omit ordinary objects. Thus, for example, those who use the special composition question to argue against ordinary objects urge that we can offer a principled, nonarbitrary, non–ad hoc ontology only if we accept a compositional principle that leaves out most ordinary objects. Those who defend the idea that there is a rivalry between the ontologies of common sense and of science argue that we cannot accept both of these ontologies, and that, since we can get an ontology with superior epistemic credentials by choosing the ontology of the scientific image, we should choose the latter. Finally, arguments from parsimony are obviously grounded in the view that we can get a better, because sparer, ontological view overall by denying the existence of ordinary objects.
110
handling existence questions
111
Before I turn to address those specific arguments in detail, we need to investigate what is involved in asking the general ontological questions used in these arguments—questions such as ‘What exists?’ and ‘How many things are there?’ The background developed in chapters 2 and 3 has implications regarding which existence questions are wellformed, answerable questions, and what method is to be undertaken in answering them. I will begin here by laying out these consequences, and then will go on in chapters 7, 8, and 9 to show how they help us unifiedly defuse the superficially diverse arguments against ordinary objects based on the special composition question, alleged rivalry with science, and comparisons of parsimony.
6.1 Specific Existence Questions In section 2.5 I argued that evaluating the truth of existence and nonexistence claims requires determining what category of entity was the intended referent in a presupposed range of prior statements. For the nonexistence claim ‘N doesn’t exist’ (where ‘N’ is a name) or ‘Ks don’t exist’ (where ‘K’ is a kind term) is true just in case the history of those uses does not lead back to a grounding in which the relevant application conditions for terms of the presupposed category are met. Thus, on this view, there are two steps to answering well-formed existence questions: first, there is the task of undertaking a kind of conceptual analysis, determining what category of entity is presupposed in standard uses of the term, and so what it would take (according to the frame-level application conditions for that categorial term) for there to be entities of the relevant kind, be they artifacts, fictional characters, numbers, or persons. Second, there is the empirical task of discovering whether or not these conditions are in fact fulfilled. As long as the existence claim averts to a range of prior uses of the term as purporting to refer to something of a given category, the existence claim is truthevaluable in this way. Call these ‘specific existence claims’. Similarly, to be complete and truth-evaluable, counting claims presuppose a category or categories of entities to be counted. For counting claims rely on identity claims, the truth-conditions for which are category-relative (see chapter 3). It is true, for example, that there are two Ks here (where ‘K’ is a categorial term) only if there is some x here that is a K, and some y that is a K, and x = y, where the latter is the case only if x and y fail to meet the identity conditions governing Ks.1
112
ordinary objects
In certain contexts, even such highly general terms as ‘object’ and ‘thing’ may figure in specific existence and counting claims, provided they are used as sortals, with the speaker associating them with at least high-level application and coapplication conditions outlining what it would take for there to be an object or thing in a given situation, and under what conditions we would have the same object or thing again. There are, I think, some (perhaps quite vague) conditions of application and coapplication often associated with these terms (perhaps including medium-sized lumps of stuff well bonded together but independently mobile from surrounding stuff . . .), enabling us to use them in standard English situations to ask, for example, ‘‘How many things did you get for Christmas?’’ or, in the party game of memory, ‘‘How many objects were on the tray?’’ and to be understood when we so ask and often to reach agreement about the suitable answer. So to the extent that the relevant terms ‘object’ and ‘thing’ are associated with application conditions, the claim ‘there is some thing here’ may be truth-evaluable, and the question ‘Is there some thing here?’ may be answerable by the straightforward method described above. Similarly, to the extent that these terms are associated with coapplication conditions, claims that there are, for example, seven objects on the tray may be truth-evaluable, and questions such as ‘How many things/ objects are there?’ may be answerable. But even where terms like ‘thing’ and ‘object’ are used as sortals, there seems to be a great deal of indeterminacy and inconsistency in assumptions about what the associated application and identity conditions are (and thus about what category of term ‘object’ and ‘thing’ are). This is easy to see, in normal English, in disagreements about how many ‘things’ one got for Christmas (and which child got more) when one was given, say, a bicycle, a chess set, and a trip to the Ice Capades; or in fights among party guests about how many objects were on the tray when a capped pen or boxed disassembled puzzle2 was among them.
6.2 Generic Existence Questions As I will argue below, however, most distinctively metaphysical debates that rely on claims about whether there is some thing in a certain situation, or how many objects there are, presuppose a use of ‘thing’ or ‘object’ that can not involve treating it as a sortal. Call this the ‘(category-)neutral’ use. Given the arguments of chapters 2 and 3, we
handling existence questions
113
have reason to suspect that such generic existence and counting claims involving an allegedly neutral use of ‘thing’ are not complete, truthevaluable claims, and that the parallel generic existence and counting questions are unanswerable questions. Some questions are unanswerable for practical reasons (e.g. ‘How many fish are in the sea?’) or principled reasons (e.g. What is the exact position and momentum of that particle now?). Others are unanswerable owing not to failings of the world, difficulties in our inquiries, or problems with the relation between the two, but rather, because although they have the superficial form of proper questions, on closer examination they are incomplete pseudoquestions, for example, ‘How long is a piece of string?’ and ‘Do Dell computers help you get more out of now?’ Attempts to answer such questions would be misguided, and differing answers would not really be expressions of conflicting judgments about some fact in the world. The proper response is to ask for a more specific or better stated question—not to provide an answer put in the same terms as the question. More precisely, I will say that a question is ‘unanswerable’ if no straightforward answer to it, stated in the same terms as the original question, is truth-evaluable (where this failing is in principle, not a reflection of mere epistemic shortcomings but of deficiencies in meaning). Given what has come before, we have serious reason to worry that the neutral use of ‘thing’, ‘object’, and the like makes generic existence questions involving these terms unanswerable. Claims such as ‘There is some thing here’ have the form of normal sort-existence claims (e.g. ‘There is some elephant here’), and so by the arguments of section 2.5, it seems should be truth-evaluable only by way of determining the category of entity prior speakers intended to refer to with the use of the term ‘thing’ or ‘object’, and then establishing whether or not the relevant chain of use leads back to grounding situations in which the associated application conditions are met. But on the neutral use, ‘thing’ and ‘object’ are not category-specifying terms; indeed it has often been observed that although ‘thing’ and ‘object’ have the superficial grammatical status of count nouns (in virtue of being pluralizable), such terms as ‘thing’ and ‘object’ (on their most general use) are not genuine sortal terms (Hirsch 1982, 38; Lowe 1989, 11–2, 24–5; compare Wiggins 2001, chap. 2). If these terms do not come associated with at least frame-level application conditions, then existence claims involving them cannot be evaluated for truth by seeing whether or not the application conditions for some presupposed category are met.3 But if generic existence claims are not truth-evaluable, then generic
114
ordinary objects
existence questions, such as ‘Is there some thing here?’ turn out to be underspecified, unanswerable questions. As Alan Sidelle writes, ‘‘to get a definite answer—or ask a definite question—we need something more specific than ‘thing’ ’’ (1992b, 423). Similarly, if terms such as ‘thing’ and ‘object’ are not associated with frame-level coapplication conditions, we cannot evaluate the truth of claims about how many things or objects there are, and the parallel generic counting questions such as ‘How many things are there?’ fall under suspicion. As E. J. Lowe writes: There are ways of counting the number of men or tables or books in a given room, but no way of counting the number of red things there are: and this is not because there is such a number but one beyond our powers of determining . . . but because it apparently does not even make sense to speak of such a number until the sorts of red thing one is to count have been specified. (1989, 10).
Nor is it because of the vagueness of ‘red’ that we can undertake no such count; even if we were able to specify an exact shade of red, perhaps in objective terms of reflectance, and ask how many red things there were, the question would remain unanswerable. As Lowe continues: Suppose, for example, that the room contained a red table: then that, it might be urged, is clearly one red thing. But what about its red top and its red legs, or the red knob on one of its red drawers? . . . And what about, say, the red paint covering one of the table’s legs: is that also to count as a distinct ‘red thing’ in its own right? It rapidly becomes apparent that there is no principled way of deciding these matters, until we are told what sorts of red thing we are supposed to be counting. (10)
If counting claims involving ‘thing’ and the like in a neutral use are not truth-evaluable, then the corresponding generic counting questions (‘How many objects are there?’ etc.) turn out likewise to be incomplete, unanswerable questions. If generic existence and counting questions are unanswerable, apparently differing answers to such generic questions should not be taken as expressing genuine conflicts, any more than we would take differing responses to ‘How long is a piece of string?’ to express genuine conflicts. This makes an enormous difference to our understanding of various metaphysical debates (including those regarding the special composition question, rival ontologies, and parsimony), as I will argue in chapters 7–9.
handling existence questions
115
These questions would clearly become answerable if we replaced them with a series of better questions, each involving a sortal term, for example, ‘Do witches exist?’ ‘Does phlogiston exist?’ ‘Do tables exist?’ ‘How many cows are there?’ and so on, or use ‘thing’ or ‘object’ in some specified sortal use of their own. Some might charge that approaching existence and counting questions via these several categoryspecific questions makes metaphysical inquiry too ad hoc, lacking in a unified principle. Yet all existence questions can be answered by means of a uniform method: For any sortal term ‘K’, to find out if there are Ks, one must first determine what category of entity competent speakers intended to refer to with ‘K’, and then determine whether or not the chain of term-use leads back to a grounding situation in which the application conditions associated with that category are met.
6.3 Bare Quantification Since Quine, it has become commonplace to state existence claims quantificationally, so that ‘there exists something such that it . . .’ is taken as interchangeable with the claim that ‘Ax(. . . x . . .)’. Similarly, English existence questions about what things exist may be phrased quantificationally, asking whether it’s true that ‘Ax(. . . x . . .)’. Peter van Inwagen (1998) moreover has argued in some detail that the ‘‘source of the meaning’’ of quantifier-variable phrases is ‘‘given by the phrases of English—or of some other natural language—that they abbreviate.’’ For example, the source of the meaning of ‘8x (. . . x . . .)’ is the English phrase ‘It is true of everything that it is such that . . .’, and the source of the meaning of ‘Ax(. . . x . . .)’ is ‘It is true of at least one thing that it is such that . . .’ (238). But if that is the case then whatever suspicions we have about generic existence claims and questions involving a purely neutral use of ‘thing’ carry over to quantified statements (or questions). If they are to be fully meaningful, such quantified statements must likewise presuppose implicit or explicit reference to a certain category or categories of entity in a domain over which we are quantifying; call this the ‘categorial quantification’ view. On this view, simple quantificational statements about whether or not ‘Ax (. . . x . . .)’ are only truth-evaluable if such categories of entities are at least tacitly assumed in presupposing a domain of quantification.4 In standard English, the categories used in specifying a domain are normally supplied unreflectively by our
116
ordinary objects
practices and the context of discussion, enabling the quantified claims to be truth-evaluable without further ado. But the point becomes essential to note in metaphysical arguments that presuppose a completely category-neutral form of quantification. Nonetheless, since many contemporary metaphysicians are accustomed to assuming that quantificational questions and claims bare of any (even implicit) categorial disambiguation make sense, and to phrasing their debates and claims in these terms, this is a point that meets substantial resistance. Fans of bare quantification may thus wish to rethink the argument that got us here. I argued in section 2.5 that, given the need to account for the variability in truth-conditions for existence claims, we should accept that existence claims are only truth-evaluable relative to some category of entity the term used in the existence claim was to refer to, and are true just in case the chain of term-use leads back to a grounding situation in which the relevant application conditions (associated with that categorial term) are met. But fans of bare quantification might suggest that we can solve the problem in another way:5 not by asking whether the application conditions for the relevant categorial term are met (the point that apparently implies that category-neutral bare existence claims are not truth-evaluable), but rather by treating existence claims as quantifying in completely category-neutral fashion over ‘everything whatsoever’, and as true if (among those things) there is some thing that is of the relevant category. On this view, we can successfully distinguish between the truthconditions for existence questions about whether there is (some thing that is) a nightclub versus (some thing that is) a city, but we do not give up the idea that bare quantificational claims and questions make sense. On the contrary, the truth-conditions for all specific existence claims depend on those for the bare quantificational claims, since the specific existence claim that there is a C is true if and only if there is some thing, such that it is C. The bare quantification view and my categorial quantification view treat the priorities between generic and specific existence questions as running in opposite directions. On the bare quantification view, the most fundamental form of quantification is completely categoryneutral quantification over everything whatsoever; other forms involve various ways of restricting this completely general form of quantification, and specific existence questions (e.g. about whether there are tables, books, or fictional characters) are based on generic existence questions (about whether there is some thing).
handling existence questions
117
On the categorial quantification view, the most fundamental form of quantification involves presupposing some category or categories of entity we are quantifying over; more generic quantificational claims supposedly over ‘everything whatsoever’ can best be understood as employing what I will call a ‘covering use’ of ‘thing’, generalizing over at least a wide range of category-specific quantificational claims (for further discussion of the ‘covering’ use and of the idea of absolutely universal quantification, see section 6.6). And generic existence questions (about whether there is some thing here) can be best understood as either employing ‘thing’ sortally or as based on evaluating at least a wide range of specific existence questions (is there some animal, artifact, lump of stuff . . .). So the bare quantification approach holds that completely categoryneutral existential and quantificational claims are truth-evaluable, and that the truth-evaluability of all specific existence claims depends on that of these generic existence claims. But are category-neutral existential and quantificational claims truth-evaluable? Here the arguments of section 2.3 regarding the qua problem come into play. For the truthevaluability of quantified claims ‘(Ax (Fx))’ is parasitic on the truthevaluability of substitution instances involving names (or other forms of singular reference) to pick out items in the domain, and thus relies on the truth-evaluability of ‘(Fn1),’ ‘(Fn2)’ . . . where ‘n1,’ ‘n2,’ and so on are names. But I have argued in section 2.3 that consideration of the qua problem gives us reason to hold that singular reference can only be disambiguated to the extent that names, demonstratives, and other singular terms are associated with a category (or categories) of entity to be referred to, where this involves presupposing certain frame-level application and coapplication conditions for the terms. Without such disambiguation of reference, claims involving these would-be singular terms are not truth-evaluable. So if we take the arguments of section 2.3 seriously, it seems that substitution instances involving singular terms are not truth-evaluable without (at least tacit) categorial disambiguation, and if they are not, then neither are the quantified claims whose truth-evaluability presupposes that of the various substitution instances, nor are the English generic existence claims that are supposed to be interchangeable with these quantified claims. We do have reason, then, to be suspicious that the bare quantification approach makes generic existence claims not truth-evaluable. Moreover, since on this view the truth-conditions for specific existence claims depend on those for generic existence claims, it gives us reason to suspect that it also interferes with the truth-evaluability of
118
ordinary objects
specific existence claims. This gives us prima facie reason for taking the categorial quantification view seriously, and adopting it as a working hypothesis. Its chief virtues, like those of the views of reference and modality in chapters 2 and 3, will become evident in the chapters that follow, as we see its wide-ranging ability to diagnose and avoid a great number of the central problems of metaphysics—especially those supposed to arise for those who accept the existence of ordinary objects.
6.4 Quantifier Variance Those who reject questions like ‘How many objects are there?’ or ‘What exists?’ often do so by accepting instead what Hirsch (2002b, 51–2) has called ‘‘the doctrine of quantifier variance’’—that is, the view that the notion of ‘existence’, and with it the quantifier, have ‘‘a multitude of different uses rather than one absolute ‘meaning’’’ (Putnam 1987, 19).6 While Putnam is apparently committed to the quantifier actually having a number of different meanings, Hirsch claims only that it has many different possible meanings, which could each be invoked to make true the claims of different ontologists about how many objects there are or what exists (Hirsch 2002a, 60–1), though in ordinary English, he thinks, the meaning is more stable. Others have denied that there could be multiple meanings for the quantifier (Sider 2001b, xx; van Inwagen 1998, 187–8). But quantifier variance is not the proposal I am defending. While the fan of quantifier variance holds that there are too many potential meanings for the quantifier for a sentence like ‘There exists something’ to ground genuine debate, I hold the view that such claims—at least when interpreted in their most generic sense—are deficient in meaning rather than overly abundant in possible meanings. And while I agree that the truth-conditions for the phrase ‘There exists some thing’ may vary in the mouths of different speakers, I trace the variation to differences in the meaning or use of ‘thing’ in the phrase (according as it is used in a covering use or in any of many possible sortal uses), not to variations in the meaning of ‘There exists’. It is important to notice that we can accept the above approach to existence questions without holding that ‘existence’ or the quantifier has a number of different meanings.7 Just as there may be a unified understanding of truth (e.g. the minimalist’s ‘‘‘P’’ is true iff P’), although different sentences have different truth-conditions, so similarly,
handling existence questions
119
we can retain a unified understanding of existence although the existence conditions for things of different categories vary. On the view I’ve argued for above, existence questions, to be complete, must implicitly presuppose a certain category or categories of entities enquired about. But for any sortal term K, general existence claims of the form ‘Ks exist’ are true just in case the history of uses of ‘K’ leads back to a grounding situation in which the application conditions for terms of the category associated with ‘K’ are fulfilled. Similarly, singular existence claims of the form ‘N exists’ are (tenselessly) true if and only if the application conditions for the categorial term associated with the name ‘N’ were fulfilled at an initial baptism.8 And this is just the sort of meaning we should expect for a logical term—one that describes the role the term plays in determining the truth-conditions of sentences (Hirsch 2002b, 54). As a result, while there are different criteria for existing for things of different categories, that does not imply that there are different meanings of ‘existence’ (see Lowe 1989, 22) any more than the presence of different truth conditions for different sentences means that each is ‘true’ in a different sense.9 Nor must we decide among ‘competing’ notions of ‘existence’ or quantification which is ‘right’, or which captures the standard English meaning of ‘existence’.10
6.5 The Number of Objects The conclusion that the generic question ‘‘How many objects are there?’’—on the alleged ‘neutral’ use of ‘object’—is ill formed bears a superficial resemblance to apparently antirealist claims such as Putnam’s (1987) position that there is no number of ‘things’ or ‘objects’, which is often taken to be an expression of antirealism. But to think of the earlier claims as committed to antirealism would be a misunderstanding: the point is not the antirealist claim that what there is depends on our ways of representing, understanding, or interpreting the world, but merely that purely generic existence questions (engaged in the ‘neutral’ use of ‘thing’) such as ‘What things are there?’ are ill-formed, unanswerable questions, so that different responses to them need not be seen as competing candidates for the correct answer. Putnam writes: ‘‘the idea that there is an Archimedean point, or a use of ‘exist’ inherent in the world itself, from which the question ‘How many objects really exist?’ makes sense, is an illusion’’ (1987, 20). For, according to Putnam, the question ‘‘How many objects really
120
ordinary objects
exist?’’ can only be reasonably asked from within a conceptual system, or ‘version’. As he points out, even a simple question, such as ‘How many objects are in a world consisting solely of x1, x2, and x3?’ will receive different answers from an ordinary person (3) than from certain mereologists (7), who (using a different conceptual system) would also include the sum of any combination of x1, x2, and x3 in the final count. As a result, Putnam argues, we should reject traditional realism in favor of his ‘Internal Realism,’ according to which ‘‘how many objects there are in the world . . . is relative to the choice of a conceptual scheme’’ (32). As Risto Hilpinen (1996) and Eli Hirsch (2002b) have argued, however, Putnam’s observations above do not at all threaten traditional realism, understood as a thesis about the independence of (at least much of ) the world from language, thought, and experience. In fact, as Hilpinen maintains, Putnam’s central point can be reduced to the simple observation that ‘object’ is not a sortal term; different answers to the question ‘How many objects are there?’ are the products of supplying ‘object’ with different existence and identity criteria—that is, turning the term into a sortal in a different way. (I will discuss other ways of explaining different answers to the question ‘How many objects are there?’ below.) But this would make apparent metaphysical disputes about how many objects there are resolve into semantic differences in the use of the term ‘object’—it does not imply anything about the minddependence (or independence) of the world. As soon as one substitutes the term ‘object’ with a sortal term such as ‘alligator’, the apparent difficulty disappears: replaced with a sortal, the idea that, for example, how many alligators there are depends on one’s choice of a conceptual scheme rapidly loses plausibility (Hilpinen 5), for the question ‘How many alligators are there?’ does seem to have a unique answer entirely independent of our conceptual scheme. If the generic question is replaced with a multitude of questions involving sortal terms (‘How many F-objects are there?’ for various individuating concepts F), the apparent difficulty for realism disappears. Similarly, if we address the question using ‘object’ in a covering sense, in which we specify a range of categorial terms to be covered (e.g. ‘organism’, ‘artifact’, ‘fundamental particle’) and agree on their associated application and coapplication conditions, we have the route to a univocal answer, the truth of which may obtain independently of anyone’s conceptual scheme. In short, the trouble in settling on a unique answer to the question ‘How many objects are there?’ does not in the least tell against realism (understood
handling existence questions
121
as a thesis that, for at least many individuating concepts11 F, ‘‘the world contains a fixed totality of mind-independent F-objects’’ [6]). Instead, it points to problems in our understanding of generic existence and counting claims.
6.6 Can We Revive the General Question? Even if we accept that existence and counting questions must always be (at least tacitly) categorial, so that if they are to be completely meaningful they must presuppose some category or categories of entity we are enquiring about, one might wonder whether we couldn’t still revive a way of understanding completely general existence claims such as ‘There is some thing here.’ For there is another common use of ‘object’ or ‘thing’ in normal English and in philosophy, which I have called a ‘covering’ use, on which, if any sortal term applies—that is, if there is a fork or an elephant or a movie or a protest—that analytically entails that ‘thing’ applies (that there is some thing in the garbage disposal or standing on one leg or showing at the theater or happening on Main Street). On this use, then, ‘object’ and ‘thing’ are not themselves used as sortal terms, but rather are covering terms guaranteed to apply given the application of any genuine (first-order) sortal term (or at least most such terms). (Clearly there may also be more or less restricted covering uses that, for example, license the inference that ‘there is some thing’ only from the application of certain sortals—say those for substances rather than events or processes). Whether or not such covering terms apply must then be determined by way of determining whether or not the particular sortals apply: if ‘table’ or ‘rock’ or ‘plant’ does, then ‘thing’ does. And since the rules for applying ‘thing’ and ‘object’ on this model are based on those for applying individual sortals, the individual sortals must be supposed to have application conditions that don’t themselves appeal to the existence of some thing in the relevant situation. But what about claims of nonexistence? One might say that ‘There is no thing here’ is true if no actual sortal or, more broadly, categorial term applies. Thus we could try to rephrase the generic existence question ‘Is there some thing here?’ using second-order quantification over categorial terms, asking, of all categorial terms C, whether there is some C here. Similarly, we might hope to revive a general version of the counting question capable of providing a total ‘number of things’ by first generating a list of all categorial terms, using each to count the
122
ordinary objects
number of instances of things of that category, and summing them up.12 Can one revive general versions of the questions in this way? Let me emphasize at the start that—from the perspective of the arguments about ordinary objects to come—nothing hangs on whether or not one can, and so there is no need to take a stand on this issue (which, of course, is just another version of the contested question of whether absolutely unrestricted quantification is possible). Nonetheless, it is worth briefly examining the prospects for reviving the general form of the question in this way so that we can see what would be involved in that whenever the suggestion arises. Tim Williamson (forthcoming, x 8) argues that even if one accepts that first-order quantification must presuppose a sort of entities quantified over, a form of universal quantification can be revived by quantifying first over all expressions, and then over their ‘compliants’— things to which the expressions apply. While this may be perfectly coherent, and give quite a broad sense of generality, note that if there is any hope of generating complete lists or counts in this way, one must not just quantify over all actual expressions (in English or any actual language or languages) that serve as categorial terms, since there may, of course, be categories of things (and things of categories) for which any given language (or all) lacks terms. Consider a culture of people who use tools and live in dwellings constructed according to certain traditional designs, but speak a language, Lemish, that lacks any artifactual kind terms, employing only simple terms for lumps of matter. If speakers of that language first attempted yesterday to count how many ‘things’ there are using only their existing categorial terms, and then today coin the term ‘artifact’ (with the same meaning as our term ‘artifact’), their ‘count’ of ‘things’ will suddenly go higher, since the application conditions for that term are already met. But the change in ‘number of things’ in their count between yesterday and today clearly only gives the illusion of reflecting a change in the world, for the count goes higher even if no manufacturing or birthing (indeed no other change in the physical facts of the world) has occurred. The compliants of artifactual kind terms were already in place in their world yesterday; all that they lacked was the terms themselves. So in short, if we quantify only over the actual terms in any (or all) actual languages, the resulting ‘counts’ will be as much a reflection of which categorial terms are in the language(s) (and what criteria are set up) as of the facts of the world that establish how many times the relevant criteria are each fulfilled.
handling existence questions
123
‘‘But that’s just the problem with making counting depend on sortals,’’ a critic might say, ‘‘it allows you to simply ‘define things into existence.’’’ Properly understood, however (and as Eli Hirsch, 2002b, 52–4, has argued at greater length), the view that we cannot answer the question of what or how many ‘things’ there are without substitution of genuine sortal or categorial terms does not entail that we can ‘define things into existence’. To think that something new (other than a word) has come into existence would be a use-mention confusion: (according to the very criteria set up for the word ‘artifact’) the things described existed before and independently of the use of the term—in the Lemish world, yesterday there were artifacts, there just weren’t things called ‘artifact.’ The illusion that some ‘thing’ has been ‘defined into existence’ arises only from combining the view that counting questions must be asked using categorial terms with the view that we can meaningfully undertake a total count of ‘things’ in the world using actual categorial terms available. But if you ask instead, of yesterday or today, how many Cs were in the world, for (almost) any categorial term C13, then (assuming no changes in the world apart from the introduction of the term), the answer would be the same for each categorial term. So if we hope to get a kind of universal quantification, instead of quantifying over (actual) expressions, it seems we must attempt to quantify over all possible categorial concepts. What, then, is a possible categorial concept? In line with section 2.3, we might say that there is a possible sortal concept wherever there is a set of consistent and nonempty application and coapplication conditions, and any possible sortal concept entails that there is a possible categorial concept, under which would fall things of that (and perhaps other) sorts. Can we then get a complete list of what exists (playing the role of completely unrestricted quantification) by asking, for all possible categorial concepts C, whether there is a C-object, or get a complete ‘count’ of all the ‘things’ there are by counting the number of C-objects for each possible categorial concept C and summing them up? There are at least two worries about this. One is that it seems reasonable to suppose that there is no determinate number of ways in which categorial concepts may be distinguished. We can reconceive the world in an endless variety of ways as different needs or interests arise, and application and coapplication conditions may be drawn out in apparently endlessly different ways to mark these distinctions. The fact that the relevant domain would thus have to be infinite is not itself
124
ordinary objects
a barrier—we must do the same to quantify over, for example, the natural or real numbers. But in those cases the domain can at least be delineated by way of a procedure for generating members (e.g. use of the ‘successor’ function for natural numbers). But there is no comparable procedure for generating all possible categorial concepts, since they may divide things up in inventive ways we cannot yet imagine. As a result, it is not clear that we have any grasp of how the supposed domain could be generated (even if not in a finite period of time), and so it is not clear that we can genuinely quantify over all possible categorial concepts, and if we cannot, it seems that we can ‘‘form no definite conception of the totality of all objects which could be spoken of ’’ (see Dummett 1973/1981, 566–7, 582–3). The second worry is that, even if universal quantification over all possible (first-order) categorial concepts is possible, and also enables us to quantify over all instances of those categories (thus regaining quite a broad sense of generality), set-theoretic-style paradoxes threaten the idea that this should count as an unrestrictedly universal quantification over ‘everything whatsoever’ (see 567–9).14 Call a possible categorial concept ‘not self-compliant’ if and only if it does not apply to itself (e.g., animal is not itself an animal) and ‘self-compliant’ otherwise. Here a Russell-style paradox quickly arises when we consider the category non-self-compliant categorial concept: is it self-compliant or is it not? If it is, then it is not; if it is not, then it is. This kind of looming paradox gives reason for thinking that talk of possible categorial concepts must be carefully typed, so that we can (in the metalanguage) quantify over all possible (first-order) categorial concepts (animal, artifact . . .), and all of their compliants, but then there are possible (second-order) categorial concepts over which we have not quantified (e.g. first-order category), so there is a sense in which we have not yet quantified absolutely universally. We could pick these up at the next level, but only by using a still higher order categorial concept, and so on, so that it seems there is no level of discourse from which one could rightly claim to be quantifying absolutely universally. So it is not clear that one can revive a sense of absolutely unrestricted quantification (such as could perhaps give us a complete inventory of what exists or a total number of things) by quantifying first over all possible categorial concepts—and the issue of whether one can will ultimately hang together with issues about the need for type theories generally, issues that cannot be resolved here. Fortunately, for the arguments that follow below, nothing hangs on the issue of whether or not completely general versions of existence
handling existence questions
125
and identity questions can be revived by way of such a formalization of the standard English covering use of terms like ‘thing’. What we can do is to regain a kind of generality, if not an absolute generality, by conjoining as many categorial terms as we wish, and asking if there is something of any of those kinds. Similarly, we can revive somewhat general ‘counts’ by employing as many categorial terms as we like, numbering the instances falling under each, and adding them up. The results may be misleading, since counts of objects (like long conjunctions) normally presuppose that the things being counted are of the same category, so if we conjoin a multitude of categories in a count, we may get a misleading number in reply. Nonetheless, with that caveat, it can presumably be done (however misleading or useless the result may be)—though again the number must be noted to be a reflection jointly of the terms considered and the world’s fulfillment of them.
C c
seven
the special composition problem
The special composition question is (roughly) the question of when a plurality of objects adds up to, or composes some object (van Inwagen 1990, 30–31).1 More formally, using the device of a plural referring expression (‘the xs’), we can ask it as: ‘When is it true that there is some y, such that the Xs compose y?’ Peter van Inwagen argues at length that the only nonarbitrary, plausible answer to this question is that there is some y such that the xs compose y ‘‘if and only if the activity of the xs constitutes a life’’ (90), thus leading him to deny the existence of inanimate macroscopic material objects (including such ordinary objects as tables and chairs, sticks and stones), holding that simples and organisms are the only physical objects (98). In fact, consideration of the special composition question has served as the basis for arguments against ordinary objects somewhat independently of the particular answer to the special composition question offered. Thus, for example, Terence Horgan and Matjazˇ Potrcˇ argue for a different answer to the special composition question—that composition never occurs ‘‘because there is only one real object, viz. the blobject’’ (2000, 266)—but similarly use the special composition question as the basis for repudiating ordinary objects. I will argue, however, that if we accept the understanding of existence questions proposed in chapter 6 we can diagnose the problems behind the special composition question. For the arguments of chapter 6 suggest that the generic version of the special composition question used in arguing against ordinary objects is ill formed and unanswerable, while answerable 126
the special composition problem
127
versions of the question do not tell against the existence of ordinary objects. Although its adherents do not generally explicitly lay out the underlying argument form, the basic form of the argument from composition used to eliminate composite inanimate objects (and maybe more besides) seems to be the following. 1. There is no acceptable uniform answer to the special composition question that gives us composite inanimate objects. 2. No nonuniform answer to the special composition question is acceptable. 3. If no acceptable answer to the special composition question says that there are composite inanimate objects, then there are no composite inanimate objects 4. Therefore, there are no composite inanimate objects.
By ‘uniform’ answer, I mean an answer to the special composition question that provides a single (nondisjunctive) answer to the question of when there is some y such that the xs compose y, which can be stated without restriction to xs and/or ys of particular kinds, and so offers the same answer for any xs and ys whatsoever. I will consider premises 1 and 2 in turn.
7.1 Uniform Answers to the Special Composition Question It is easy to see the difficulties in finding an acceptable uniform answer to the special composition question that will yield an ontology of ordinary composite objects. Consider some of the proposed answers to the special composition question that would yield (at least many) composite physical inanimate objects. The first series of answers van Inwagen canvasses involve putting objects together in increasingly strong ways—beginning from mere ‘contact’ and moving through fastening, cohesion, and ultimately fusion (conjoining them so that there is no discernible boundary).2 While each of these seem, pretheoretically, to be what makes the difference to creating an object in some cases (cairns, furniture, collages, and metalwork, respectively), van Inwagen argues that it is absurd to think that they always do. The strongest examples against their sufficiency come from cases of conjoining people in any of the relevant ways, for even in the strongest case (if one surgically fuses the hands of two people and allows them to grow back together so that
128
ordinary objects
there’s no discernible boundary), it is just implausible to think there is now some larger object of which both are a part: ‘‘despite their fusion, nothing is such that Alice and Beatrice [the fused people] compose it’’ (1990, 59). (It is worth noting that here and in other cases, answers to the special composition question are rejected at least in part on the basis of their conflicting with our pretheoretic sense of what objects there are and are not, e.g., that there is just no object composed by two fused people. Yet the conflict between van Inwagen’s proposed answer and the common sense ontology that claims there are tables, chairs, rocks, and the like is not taken as a serious problem with that answer, since such common sense judgments—although not literally true—can be understood as saying something true, and capturable in ‘language of refuge’ claims about simples arranged in certain ways. Of course, however, the common sense intuition—if there is one—that fused people form no new ‘object’ could equally well be captured by acknowledging that—although not literally true—it is true that there is no new ingredient added to the world, no new organism or person created, etc. So the question remains of what the justification is for taking common sense ontological judgments as counter-evidence to answers to the special composition question in some but not other cases.) Attempts to preserve a common sense ontology by offering other sorts of answers to the special composition question (not directly or merely involved in conjoining objects) fare no better. One might think, for example, that what makes a difference to simples composing (or not composing) a statue is their being arranged in that particular way (statuewise). Yet, van Inwagen argues, if it is shape or arrangement that is supposed to be relevant to the composition of a ‘new’ object here, then consistency requires that one allow that a ‘new’ object comes into existence every time (e.g.) the clay is squished in a slightly different way, whether accidentally or intentionally. Then ‘‘you must, as you idly work the clay in your fingers, be causing the generation and corruption of the members of a compact series of objects of infinitesimal duration. That is what seems to me to be incredible’’ (1990, 126).3 One might hope to do better by focusing not on intrinsic factors such as shape, but rather on the objects’ relation to external features like intentions, since, for example, it seems to be intentions that are essential to determining that artifacts are created and maintained over time. In fact, van Inwagen paraphrases claims about the persistence of artifacts over time by averting to the continuation of a ‘‘history of maintenance’’ of certain arrangements (1990, 133), where that history
the special composition problem
129
of maintenance involves reference to the ‘‘activities and intentions of intelligent beings’’ with respect to that arrangement of materials, but allows for the component materials to undergo gradual replacement. This, one might think, could be transformed into a suitable principle of composition that would include artifacts as well as organisms, in both cases by allowing that it is the ongoing maintenance of a certain organized system (whether biologically in the case of a life, or by a history of maintenance in the case of an artifact) that makes the difference as to whether or not individual xs compose some larger object. Thus, one could propose as an answer to the special composition question: there is some y such that the xs compose y if and only if ‘‘the activity of the xs constitutes a life or the xs are the current objects of a history of maintenance’’ (138). Yet van Inwagen rejects this amendment since it violates what he considers to be an independently plausible principle: that whatever it is that distinguishes cases in which composition occurs from cases in which it does not, it must be a matter of an internal multigrade relation among the simples, not an external relation between those simples and, say, people’s intentions and activities: My deepest instincts tell me that composition is an internal relation and that, therefore, a proper answer to the Special Composition Question must take the form of a statement that asserts a necessary extensional equivalence between the relation expressed by ‘the xs compose something’ and some internal multigrade relation. (1990, 138)
Whether or not one shares van Inwagen’s reasons for rejecting this answer to the special composition question, clearly even such an answer would be of only limited help in preserving a common sense ontology, for it would clearly not help for natural objects such as sticks and stones. (Nor, as stated, would it help for artifacts created and quickly destroyed or left to decay—which seem no less artifacts than those carefully maintained across the centuries.) At best, it would provide a way of preserving only some inanimate macroscopic objects. While van Inwagen’s objections to these proposals are typically objections to their sufficiency (since accepting them as general principles would entail accepting far too many, wildly implausible, composite ‘objects’), what is perhaps more telling from the point of view of defenders of ordinary objects is that none of these proposals can even be taken as a necessary condition for the existence of a composite object without eliminating many of the ordinary objects we commonly think we refer to. Bonding answers would still omit unassembled or
130
ordinary objects
scattered common sense things like puzzles or the state of Michigan; simples-in-a-given-shape would have the wrong persistence conditions to be identified with statues or forks, which admit of certain shifts in shape and composition; and common sense natural objects such as sticks and stones are not intentionally created or maintained. So even if one is willing to countenance much else along with common sense objects, none of the uniform answers suggested by van Inwagen seems to provide hope of yielding our common sense ontology of tables and chairs, sticks and stones, statues and signs.
7.2 Nonuniform Answers to the Special Composition Question But why should we need a uniform, completely general answer to the special composition question? Even if there is no such answer that tells us that there are objects of a certain common sense (or any) kind K, that doesn’t give us reason to conclude that there are no Ks unless we also accept premises 2 and 3. David Sanford urges us not to take the failure to provide a uniform answer to the special composition question that yields ordinary objects as evidence against their existence ‘‘I council [sic] naı¨ve mereologists everywhere not to rise to this challenge [of answering the special composition question]. Do not assume that composite beings cannot exist unless there is a true answer to the special composition question’’ (1993, 223). Sanford suggests that, taken in its general form, this is an impossible question, which we would be better off replacing with a multitude of possible questions—asking not in general ‘When is it the case that there is some y such that the xs compose y?’ but rather such manageable substitution instances as ‘When is there some ship such that the planks compose the ship?’ or ‘When is there some fort such that the rocks compose the fort?’ and so on (224). Even if there is no completely general, uniform answer to the special composition question that will yield an ontology of tables, chairs, sticks, and stones, it certainly seems that a highly disjunctive, nonuniform answer based on answering a sequence of questions like those above could easily provide us an ontology that includes ordinary composite objects by providing diverse principles of composition for the diverse sorts of things we seem to be surrounded by.4 Thus, although most discussion of the special composition question has focused on attempts to provide a uniform answer to the question, it is crucial to the success of the argument from composition as an
the special composition problem
131
argument against the existence of composite inanimate objects that nonuniform answers be unacceptable. So what is the argument for premise 2, the claim that no nonuniform answer to the special composition question is acceptable? The nonuniform answers van Inwagen considers are what he calls ‘‘series’’ answers—answers of the form ‘‘(Ay the xs compose y) if and only if the xs are F1 and stand in R1, or the xs are F2 and stand in R2, or . . . or the xs are Fn and stand in Rn’’ (1990, 63). While the sort of thing (y) to be composed is still left completely unspecified, answers of this form at least leave open the option of offering different sorts of relation for different sorts of thing, so that we could allow, for example, that if the xs are planks, and stand in the relation of being nailed together in a certain form for a certain purpose . . . there is some ship that they compose, without having to allow that nailing guinea pigs together in like form would compose anything. Van Inwagen offers two major objections to any such series-style answers, as well as a number of other ‘‘remarks’’ offering considerations against them. His major objections concern (1) circularity and (2) transitivity, and are deployed against a particular example of a series-style answer (designed to work in a simplified world): ‘‘(Ay the xs compose y) if and only if the xs are particles and are maximally Pbonded or the xs are atoms and are maximally A-bonded (or there is only one of the xs)’’ (1990, 64). The circularity objection runs as follows: either the terms on the right-hand side of a series-style answer (F1, F2, . . . Fn) contain mereological concepts or they don’t. If they do, then the answer cannot provide a proper answer to the special composition question. If they do not, then the series-style answer can be reduced to a nondisjunctive answer (64–5). The sample answer, he argues, falls on the first horn of the dilemma, since ‘atom’ is implicitly a mereological concept, for ‘‘what does ‘atom’ mean (in respect of our imaginary universe) if not ‘object composed of particles’?’’ (64) If we are lucky enough to be able to remove the term ‘atom’ from the righthand side—if, say, it is impossible for anything but atoms to be Abonded—then the answer will reduce to a uniform answer, since we can simply restate the right-hand side as ‘‘the xs are maximally bonded (or there is only one of the xs)’’ (64). But however plausible this assessment may be of the sample series answer, is it really a threat against highly disjunctive answers to compositional questions involving concepts of ordinary inanimate objects? The second horn of the dilemma does not seem relevant, for prospects for reducing all of these diverse answers to a single sort (as we have seen
132
ordinary objects
above) seem hopeless. But is it really the case that any such disjunctive answer must contain mereological terms on the right side? It does not seem to be the case that, for example, ‘plank’ means ‘object composed of molecules’ (or anything else)—the ordinary concept of planks (and bricks, clay, etc.) seems to be quite neutral on what simple (or simpler) sorts of entities, if any, compose planks.5 It is certainly far from obvious that all of the ordinary concepts we might plug into a specific compositional principle are really mereological concepts. The second objection van Inwagen offers is that series-style answers interfere with the transitivity of parthood, a ‘‘nonnegotiable feature of parthood’’ (1990, 65), since ‘‘for A to be a proper part of x is for there to be ys other than A such that A and the ys compose x’’ (65). But, for example, in the sample series answer, a particle could be part of an atom, and that atom be part of a molecule, without the particle being part of a molecule (since, according to the sample answer, the particle does not combine with other things in an A-bonding relation to compose the molecule). But surely one can respond that the problem here is not with series-style answers, but with the proposed understanding of ‘proper part’. If we instead say, for example, ‘for A to be a proper part of x is for there to be ys other than A such that A joins with other ys to compose x, or, for some z, for A to be a proper part of z and z to be a proper part of x,’ then we can retain the ideas that (in the relevant example) the particle is part of the molecule and that parthood is transitive, while retaining a series-style answer.6 So neither of the major arguments against series-style answers seems decisive against highly disjunctive answers that might provide different principles of composition for different sorts of inanimate composite objects. Beyond these direct arguments, van Inwagen offers a number of ‘‘remarks’’ against the acceptability of series-style answers, which I will divide into two major categories. The first is that such answers, if they are to take account of all the ways common sense allows that some sorts of object may compose others, would be ‘‘disgracefully messy’’ (1990, 66), and since they would involve allowing, for example, that if planks are fastened together, they may compose an object, whereas if persons are so fastened they do not, they must also be unacceptably arbitrary, for ‘‘what could justify such discrimination?’’ (69). While van Inwagen only mentions these briefly as background concerns with disjunctive answers, Horgan and Potrcˇ (2000) take the demand for nonarbitrariness as the main reason for seeking a uniform answer to the special composition question, and offer a lengthier defense of this demand.7 As they aptly point out, van Inwagen’s treatment
the special composition problem
133
of the special composition question suggests that there is a tension between ‘‘(1) finding a systematic, general answer to the SCQ [special composition question], and (2) adopting an ontology that conforms reasonably well to our pre-theoretic beliefs, and our scientifically informed beliefs, about what kinds of physical objects there are’’ (259). Given that tension, Horgan and Potrcˇ agree that (1) satisfying what they call the ‘‘principle of the nonarbitrariness of composition’’ is far more important than preserving (2) ‘‘the posits of common sense and science.’’ (Indeed Horgan, 1993, argues that satisfying (1) itself requires that we reject organisms as well as artifacts, and thus accept an ontology still sparer than van Inwagen’s.) They justify the claim that satisfying (1) is more important as follows: An adequate metaphysical theory, like an adequate scientific theory, should be systematic and general, and should keep to a minimum the unexplained facts that it posits. In particular, a good metaphysical or scientific theory should avoid positing a plethora of quite specific, disconnected, sui generis, compositional facts. Such facts would be ontological surds; they would be metaphysically queer. Even though explanation presumably must bottom out somewhere, it should bottom out with the kinds of ‘unexplained explainers’ we expect to find in physics—viz., highly general, highly systematic, theoretical laws. It is just not credible—or even intelligible—that it would bottom out with specific compositional facts which themselves are utterly unexplainable and which do not conform to any systematic general principles. Rather, if one bunch of physical simples compose a genuine physical object, but another bunch of simples do not compose any genuine object, then there must be some reason why; it couldn’t be that these two facts are themselves at the explanatory bedrock of being. (259)
Later (in chapter 11) I will argue that the analogy between the work of metaphysical theories and that of scientific theories is importantly misleading.8 But at least initially, one might think of unity, simplicity, nonadhocness, and nonarbitrariness as quite general theoretic virtues that apply as much to metaphysical as to scientific theories, and so the call for a uniform principle of composition might still seem legitimate on those grounds. Van Inwagen’s second set of remarks highlights the fact that highly disjunctive answers to the special composition question (unlike uniform answers) could not help us solve a number of central metaphysical problems. First, they could not help us answer the fundamental ontological question ‘What exists?’—for to take into account all of the relations that seem, common-sensically, to be relevant to whether
134
ordinary objects
some sorts of objects compose some other sort, it seems that ‘‘we must first decide what objects we think there are and then try to devise an answer that will generate them’’ (1990, 66). Answering the special composition question in this way presupposes an ontology rather than yielding one. Similarly, van Inwagen argues that series-style answers cannot help us resolve many of the philosophical questions about the proper identity and persistence conditions for objects of various types, whereas uniform answers such as his own can—and ‘‘it would be nice if we had an answer to the special composition question that suggested ways of adjudicating’’ hard cases about identity and persistence (70). The desire to find a nonarbitrary, non–ad hoc answer that could help us to solve other metaphysical problems (first and foremost, providing an answer to the question ‘What exists?’) seems fair enough at first glance. But are these really legitimate demands that should incline us to reject all nonuniform answers to the special composition question (and with them, reject an ontology of common sense objects)? To determine whether they are, we need to step back to investigate some of the presuppositions behind the way the special composition question is posed, and what it would take for it to be answerable in its completely general form.
7.3 Behind the Special Composition Question The special composition question asks ‘‘In what circumstances is a thing a (proper) part of something?’’ (van Inwagen 1990, 21) or, more precisely ‘‘When is it true that Ay (the xs compose y)?’’ (30) Thus, it is a certain kind of existence question posed in quantificational terms, requiring us to say, of various situations (e.g. when atoms are arranged baseballwise, when two people’s hands are glued together, etc.), whether or not there is ‘some y’ in that situation—that is, some one y— or one ‘thing’—composed by those xs. But how are we to interpret this generic existence question, about whether or not (in various situations) there is any thing composed by the xs? As I have argued in chapter 6, there seem to be three ways to try to understand ‘thing’: sortally, in a covering use, or in an allegedly neutral use. I’ll consider each of these possible interpretations in turn. If ‘thing’ is being used sortally, then the speaker must associate it with application and coapplication conditions outlining what it would take for there to be an object or thing in a given situation, and under what conditions we would have the same object or thing again. But
the special composition problem
135
then the fact that a particular answer to the special composition question denies the existence of tables would just be a matter of denying that so-called tables fit criteria the eliminativist presupposes for application of the term ‘object’. But this would be a relatively uninteresting semantic point about the eliminativist’s own (proposed or presupposed) use of the language, not a surprising claim about the world, and the special composition question could not be used to seriously deny the existence of ordinary objects. Moreover, all disputes about composition would turn out to either involve participants talking past each other (by using ‘thing’ differently) or at best be conceivable as shallow disputes about the best way of turning the term into a sortal (not as deep disputes about what there is) (compare Hirsch 2002a, 106; Sidelle 2002, 141–2). Thus those who take the special composition question seriously generally reject this interpretation of it in no uncertain terms. As van Inwagen writes: Many philosophers . . . have insisted, despite repeated protests on my part, on describing my position in words like these: ‘‘Van Inwagen says that tables are not real’’; ‘‘. . . not true objects’’; ‘‘. . . not actually things.’’. . . These are words that darken counsel. They are, in fact, perfectly meaningless. My position vis-a`-vis tables and other inanimate objects is simply that there are none. (1990, 99)
So suppose instead that we understand ‘thing’ here as being employed in a covering use. Then we can rephrase the question as asking: when it is the case that (given certain basic entities), for some possible sortal term ‘S’, the domain of Ss is nonempty? While it is not clear that it is possible to regain unrestricted quantification by way of quantifying over all possible sortal or categorial terms (see x 6.6), here it is enough to note two things. One, if we approach the question in that way, we only have reason to expect a nonuniform (disjunctive) answer to the question, giving different conditions for things of different sorts being composed. And so if the question is to be interpreted in that way, we should reject premise 2 and allow that nonuniform answers may be acceptable. Two, if we understand the special composition question as asking whether, given certain basic entities, there is some possible sortal—or even just some actual sortal in English—that has a nonempty domain, the outcome of the question certainly does not favor the eliminativist. For then each substitution instance is true if the application conditions for the sortal in question are fulfilled. But in just those controversial cases in which the eliminativist wanted to say there was ‘no thing’ there (e.g. where there are many particles arranged
136
ordinary objects
tablewise) there is, for example, a possible (indeed actual) sortal term ‘table’ such that the application conditions for that sortal are fulfilled in that situation (e.g. one in which particles are arranged tablewise), giving us the result that there are ordinary objects. (I will return to this point in x 9.3). The third option is to attempt to restate the special composition question using ‘thing’ not as a sortal or in a covering use, but completely neutrally, asking whether there is ‘anything whatsoever’ there, composed by the xs. This is apparently the interpretation van Inwagen has in mind (1990, 31), and given the problems with the other two interpretations, it seems that this must be the interpretation eliminativists rely on in making what are supposed to be substantive arguments against ordinary objects. So all hopes for using the special composition question in arguments against ordinary objects rely on the idea that we can make sense of this neutral use of ‘thing’. But, as I have argued in chapter 6, if we interpret generic existence questions as using ‘thing’ or ‘object’ in an entirely ‘neutral’ sense, we have reason to think that they are incomplete and unanswerable—the special composition question, on this hypothesis, would turn out (as Sanford suggested) to be an ‘impossible’, unanswerable question, and the failure to find a uniform answer that yields the objects of common sense would be no reflection on the issue of whether or not those objects exist, but only reflect the fact that the question is ill formed. In short, if we accept the results of chapter 6, then in whichever of these ways we interpret it, the special composition question cannot be used as the basis for arguing against the existence of ordinary objects. As a result, if the above is correct, the demand for a uniform, nonarbitrary, non–ad hoc answer to the special composition question that yields an ontology of ordinary objects is indeed inappropriate. If we address the issue of composition at all, we can expect only highly disjunctive answers, giving different principles for different sorts—an answer that should not be surprising, given the earlier arguments that existence questions must be addressed categorically, since questions about composition are at bottom questions about the existence of composite objects. And such highly disjunctive answers are certainly compatible with an ontology of ordinary objects.
C c
eight
problems of rivalry with science
The world as described by physical science has come to resemble less and less the world we seem acquainted with in our everyday experience. Thus the problem of the relation between the common sense world and scientific world—a problem that has been around at least since the days of Descartes and Newton—has continued to intensify. The divergence between the world-descriptions provided by physical science and common sense has led to some of the oldest and most persistent arguments for eliminating ordinary objects. For if, as some have thought, the descriptions of science compete with those of common sense, given their superior epistemic credentials (it is said), the scientific descriptions surely win out, and we must accept that common sense descriptions of the world as containing solid rocks, green grass, or sweet strawberries apply to nothing. There are, in fact, two distinct forms that such arguments can take. The stronger form—inspired but apparently not endorsed by the astronomer Sir Arthur Stanley Eddington—alleges that there is a conflict between the descriptions or claims of common sense and those of physical science, a conflict that physical science wins. A slightly weaker form of argument, popularized by Wilfrid Sellars, holds that while the two images may not strictly be said to conflict, there is nonetheless a rivalry between them, as each purports to offer the true and complete description of the world. Thus again, if the two are rivals, surely (it is said) the scientific view must win out at the expense of the common sense view, and we must deny the existence of ordinary objects in favor 137
138
ordinary objects
of an ontology sanctioned by physical science. I will treat these two arguments in turn.
8.1 The Case for a Conflict The idea that the descriptions of the world furnished by physical science conflict with those of common sense was popularized by Eddington’s famous discussion of the ‘‘two tables’’: Yes; there are duplicates of every object about me—two tables, two chairs, two pens. . . . One of them has been familiar to me from earliest years. It is a commonplace object of that environment which I call the world. . . . It has extension; it is comparatively permanent; it is coloured; above all it is substantial. . . . Table No. 2 is my scientific table. . . . My scientific table is mostly emptiness. Sparsely scattered in that emptiness are numerous electric charges rushing about with great speed; but their combined bulk amounts to less than a billionth of the bulk of the table itself. (1928, ix–x)
The descriptions of the ‘table of science’, Eddington emphasizes, do not merely differ from the descriptions of the ‘table of common sense’, they conflict with it in various ways. The central conflict Eddington points out is that the common sense table is ‘substantial’ and solid, while the scientific table is ‘‘nearly all empty space’’ (1928, x) and so neither substantial nor solid, but he could just as well add that the common sense table is (while the scientific table is not) colored, fragrant, or sonorous. The apparent conflict between these two descriptions suggests that we cannot accept that both are true descriptions of a single table (hence the appeal to ‘two tables’)—one of them must be discarded. For most scientists, it is clear which description wins this conflict; as Eddington writes, ‘‘I need not tell you that modern physics has by delicate test and remorseless logic assured me that my second scientific table is the only one which is really there’’ (1928, xii). And indeed it is not just the scientist who will find this a reasonable choice. The fact that the scientific description is arrived at through careful theory construction, testing, and observation—rather than through the rather hazy credentials of ‘common sense’—gives reason to declare it the winner, if indeed there is a conflict between the two, just as Copernican scientific theory was declared the winner over the ‘common sense’ geocentric view of the cosmos.
problems of rivalry with science
139
In fact, however, insofar as it is possible to extract a single position from Eddington’s text, this seems not to be the conclusion Eddington himself ultimately endorses.1 Ultimately, he seems to endorse the view that the world as described by physics is ‘‘purely symbolic,’’ a ‘‘world of shadows’’ (1928, xiv–xv), concerned only with making predictions connecting some ‘pointer readings’ to others, and not at all concerned with intrinsic descriptions either of its own named phenomena or of the entities familiar from everyday life (247–55). This view (which Eddington claims to be ‘‘essentially the current scientific doctrine,’’ 254) seems in places to be a kind of operationalism, that takes the terms of physics not to have any metaphysical descriptive content (and thus none that could conflict with the descriptions of common sense), but merely to be defined in terms of experimental methods used to confirm or disconfirm their application. Steven French (2003) interprets Eddington rather as a structural realist (along the lines of Russell’s views in The Analysis of Matter, 1927), which would also ensure a lack of conflict between the merely structural properties physics imputes to the world and the qualitative content involved in ordinary world descriptions. Either way, so interpreted, the results of the natural sciences would after all provide no threat to ordinary objects, since they are not providing any descriptions that could in principle conflict or compete with those of common sense. For present purposes, the exact status of Eddington’s historical views is in any case not at issue. For whether or not he ultimately endorsed the view that there is a genuine conflict between scientific and common sense descriptions, clearly many have taken observations like those with which he begins his book as reasons for thinking that there is a conflict between the two, adding that, given that conflict, the scientific view must win out and ordinary objects be dispensed with. The first issue to address here, then, is whether the argument can be successfully made that there is a conflict between scientific and common sense descriptions that requires us to choose at most one of them as describing the ‘real’ objects. The central conflict Eddington emphasizes is that, according to common sense, tables and planks are solid, while according to physics, such entities are ‘‘nearly all empty space’’ (1928, x). Conceive [of electrons] as substantially as you will, there is a vast difference between my scientific table with its substance (if any) thinly scattered in specks in a region mostly empty and the table of everyday conception which we regard as the type of solid reality. (xi)
140
ordinary objects
Other apparent conflicts could be mentioned: The brownness of the one, the colorlessness of the other; the woody smell of the first but not the second, and so on. Is there a genuine conflict here that should lead us to reject the common sense descriptions of the world as containing things like tables, which are solid, colored, fragrant, and so on? In order to demonstrate a conflict, one must show that the two descriptions are talking about the same thing (name it ‘t’), and utilizing the same (meaningful) predicate (‘P’), with one asserting that t is P, and the other denying that t is P. There are, however, difficulties every step of the way with demonstrating that there is such a conflict. The traditional response to these claims of a conflict has been to deny that the common sense and scientific frameworks are utilizing the same meaningful predicate ‘P’ (e.g. ‘solid’), with the former claiming it applies to t, while the latter denies it. Thus, Susan Stebbing argues that one cannot at the same time accept that the ordinary predicate ‘solid’ is meaningful, and deny that it applies to ordinary objects such as wooden planks. Unless we understand what ‘solidity’ means, we cannot understand what the denial of solidity to the plank amounts to. But we can understand ‘solidity’ only if we can truly say that the plank is solid. For ‘solid’ just is the word we use to describe a certain respect in which a plank of wood resembles a block of marble, a piece of paper, and a cricket ball, and in which each of these differs from a sponge, from the interior of a soap-bubble, and from the holes in a net. (1958, 51–2)
So the attempt to give precedence to the scientific description, and deny that the plank is really solid, leads us into nonsense, since it robs the term ‘solid’, and thus also ‘nonsolid’, of its meaning: ‘‘If the plank appears to be solid, but is really nonsolid, what does ‘solid’ mean? If ‘solid’ has no assignable meaning, then ‘nonsolid’ is also without sense’’ (53). What view of the semantics of terms like ‘solid’ is Stebbing presupposing here? Surely not that ‘solid’ is meaningless unless a certain plank can truly be said to be solid, since it is surely part of the ordinary use of ‘solid’ that it can be mistakenly applied in individual cases (e.g. where the plank turns out to be rotten to the point of sponginess). Instead, it seems we should understand it as the view that terms like ‘solid’, in their ordinary use to describe planks and tables, cannot have meaning unless they truly apply to all or most of the paradigm exemplars to which the word is commonly applied—perhaps because such terms are defined ostensively by application to exemplars. On that
problems of rivalry with science
141
view, if it is denied that ‘solid’ holds of all of its paradigm exemplars, ‘solid’ and ‘nonsolid’ must both be meaningless, and we do not have a case in which, for some t and some meaningful predicate ‘P’, common sense asserts while science denies that t is P. Other views of the semantics of terms like ‘solid’ can also be used to deny that there is any genuine conflict between common sense descriptions of tables and planks as solid, and scientific descriptions of the same spatio-temporal regions of the world as containing only fundamental particles at some distance from one another. Instead of taking it to be defined ostensively, one could take ‘solid’ to be defined in terms of other functional or experiential properties (clearly in any case its meaning is interrelated with that of other terms for functional and experiential properties)—so that possessing such features as being able to support coffee cups, having a surface without visible gaps, excluding other medium-sized objects separate from it, and so on, is enough to warrant attributing it solidity. In such a case, one could perhaps meaningfully deny that planks and tables are solid, but the scientific view that all physical objects are ultimately made up of subatomic particles separated from one another by distances far greater than their own extent does not give us any reason to deny that tables that tables, planks, and so on, are solid in this sense. For the common sense attribution of solidity to a table is simply noncommittal about the fundamental microscopic nature of its constitutive basis.2 Scientific and ordinary terms typically function differently, with a different range of application and coapplication conditions, and the application conditions for common sense terms are typically quite open-ended, so that they may (in principle) be satisfied by any of a great variety of microscopic conditions that may underlie their observable (or other macro-level) features.3 At most, the scientific view gives us reason to deny solidity in the sense of involving something like an underlying mathematical continuum of matter (see Stebbing 1958, 54). It is hard to see this, however, as the ‘ordinary use’ of the term ‘solid’ when it is being used to distinguish the respect in which planks resemble tables and cricket balls and differ from sponges, soap bubbles, and water (pace Martin 1997, 200–201). Here the application conditions for the term ‘solid’ seem to have been implicitly shifted from the common sense application conditions to those appropriate to a mathematical physics, and only then is it denied that the familiar objects of the ordinary world satisfy them. If there are substantially different application conditions attached to different uses of ‘solid’, then the fact that the ordinary person would assert, and (some) physicists deny, that the table is solid does not show that common sense
142
ordinary objects
and scientific claims about solidity conflict, for although they might use the same sounds and typography to express their claims, they are not both using the same predicate (with the same sense) ‘P’. But if there is no conflict, we are not forced to choose one or the other as providing the sole true description of the world and its inhabitants. But while problems have traditionally been raised for showing that the same predicate is being used in scientific and common sense claims, even if those could be overcome, a crucial problem would remain: showing that they are saying conflicting things about the same object. What is the alleged t that common sense asserts, and science allegedly denies, is solid? The natural answer (and the answer Eddington gives) is ‘a table.’ But Stebbing herself suggests that it is absurd to speak of the object of scientific description (supposedly in competition with the familiar table) as a ‘table’ at all: ‘‘I venture to suggest that it is as absurd to say that there is a scientific table as to say that there is a familiar electron or a familiar quantum’’ (1958, 58). The position argued for in chapters 2 and 3 gives us a principled basis to substantiate this suggestion: scientific theories certainly do not use sortals such as ‘table’, and if science and common sense are using sortals of different categories, the ‘things’ picked out by the two descriptions cannot be identical. Of course, one could try to speak more carefully, and simply speak of t as a ‘thing’ that common sense asserts and physics denies is solid. But here the views defended in chapters 2 and 6 come into play: If ‘thing’ alone, in its neutral use, is not a sortal term, it cannot enable us to establish reference to something. But then there is no way of specifying which ‘thing’ science and common sense are both referring to, and supposedly disagreeing about, and the claim that there is a conflict between them cannot be made. One could try to find a common term utilized in both scientific and common sense descriptions, for example, ‘physical object’. Yet even if that term is used as a sortal, the idea that there is a single, unified physical object (in anything like the common sense understanding of ‘physical object’) seems to be denied by the framework of physics. As Sellars writes: Many years ago it used to be confidently said that science has shown, for example, that physical objects aren’t really colored. Later it was pointed out that if this is interpreted as the claim that the sentence ‘Physical objects have colors’ expresses an empirical proposition which, though widely believed by common sense, has been shown by science to be false, then, of course, this claim is absurd. The idea that physical objects aren’t colored can make sense only as the (misleading)
problems of rivalry with science
143
expression of one aspect of a philosophical critique of the very framework of physical objects located in Space and enduring through Time. (1956/1997, 82)
Instead of speaking of a ‘thing’ or ‘physical object’ here at all, physics would speak of a great many subatomic particles in fields of force. But we cannot switch the discussion to these terms, for about such subatomic particles common sense says nothing, and thus cannot be held to assert that they are ( jointly) solid, colored, and so on. Much the same would go for attempts to switch the discussion to the ‘neutral’ territory of talking about the occupant(s) of certain spatio-temporal regions, for exactly designated spatio-temporal regions are no part of the framework of common sense, and it is not clear that common sense directly asserts anything about their inhabitants. In short, the conceptual frameworks and ontologies of common sense and physical science are so different that it is hard to find a common conceptual ground enabling them to pick out the same individuals and ascribe them conflicting properties. This fits with Ryle’s assessment that there is no rivalry between the scientific and ordinary accounts of the world, since in fact scientific theory (properly understood) is ‘‘constitutionally speechless’’ about the matters about which the ordinary conception makes pronouncements (and no doubt vice versa) (1954, 78).4 Of course there is a sense in which some people might wish to claim that physics tells us the real truth about tables and chairs; nonetheless, it seems that, given the above, the more perspicuous way of putting this point is to say that physics tells us the real truth about the stuff of which tables and chairs are constituted. This does not, of course, rule out the idea that in some limited contexts the results of something like scientific inquiry can conflict with (and overrule) claims that might count as common sense—provided that it is clear that both are talking about the same things and using a predicate with the same sense. Thus, to use the familiar example, it is plausible that the common sense claim that the earth is flat—if understood in the sense of claiming that the extended mass of land and water we live on has a continuous roughly planar surface and edges off of which one could fall, if one ventured too far—was contradicted by the discovery that the earth in fact lacks edges, but is an oblate spheroid. So similarly, the ‘common sense’ claim that eating nonfat foods helps one avoid getting fat5 is contradicted by apparent scientific claims in recent diet books. But in both of these cases, we pretty clearly have common sense and scientific discovery speaking of the
144
ordinary objects
same things, in the same terms (and if they are not, the case for a conflict evaporates). This is precisely not the case regarding common sense claims about there being tables, chairs, and tennis balls, and the claims of contemporary physics couched in terms of waves and particles. So the existence of some conflicts, and even cases in which supposedly ‘common sense’ claims have been overturned by scientific ones, does nothing to show that modern physics conflicts with accepting the existence of ordinary objects.
8.2 The Case for a Rivalry Wilfrid Sellars acknowledges the difficulties of trying to find a direct conflict between the two views, and develops a different form of argument that the precedence of what he calls the ‘‘scientific image’’ might require us to reject the ontological claims of a refined common sense. The problems above arise from the fact that the critic begins by accepting the common sense picture and use of terms such as ‘solid’ and ‘red’, and then tries to deny its truth in its own terms, for example, by claiming that physics shows that planks are not really solid nor apples really red. But the attempt to do so results in absurdity by mixing the conceptual frameworks of common sense and physics: In short, ‘Physical objects aren’t really colored’ makes sense only as a clumsy expression of the idea that there are no such things as the colored physical objects of the common sense world, where this is interpreted, not as an empirical proposition . . . within the common sense frame, but as the expression of a rejection (in some sense) of this very framework itself. (1956/1997, 82)
Physical science cannot sensibly be said to have falsified the claims of common sense, taken individually, because it is not speaking in those terms at all—not utilizing the definitions or acknowledging the ontology of common sense. But if physical science is not in the business of utilizing the framework of common sense at all, and thus not in the business of refuting it directly, it is, in Sellars’s view, in the business of offering us a superior alternative to the entire framework of common sense (or its refinements); there is ‘‘a sense in which the scientific picture of the world replaces the common sense picture; a sense in which the scientific account of ‘what there is’ supersedes the descriptive ontology of everyday life’’ (1956/1997, 82). Thus:
problems of rivalry with science
145
as long as the existing [common sense] framework is used, it will be incorrect to say—otherwise than to make a philosophical point about the framework—that no object is really colored, or is located in Space, or endures through Time. But, speaking as a philosopher, I am quite prepared to say that the commonsense world of physical objects in Space and Time is unreal—that is, that there are no such things. Or, to put it less paradoxically, that in the dimension of describing and explaining the world, science is the measure of all things, of what is that it is, and of what is not that it is not. (82–3)
Such a proposal can clearly avoid the difficulties above: one needn’t get the frameworks to speak to each other and conflict; instead, one simply grants the primacy of the scientific framework over the common sense one. But what is the argument for replacing the common sense framework with the scientific one if there is no conflict between them? Sellars returns to this issue in Science, Perception, and Reality, where he shifts from speaking of the ‘‘common sense picture’’ and the ‘‘scientific picture’’ of the world to speaking in terms of the ‘‘manifest image’’ and the ‘‘scientific image,’’ making more precise what is involved in each. The manifest image, as he describes it there, is not conceived pejoratively as an unscientific, uncritical, or unrefined picture. Instead, it is derived from refinements of the original image—the image constructed in which people first had a conception of themselves as people-in-the-world, and thus also became persons, not mere biological Homo sapiens.6 The original image of ourselves and our place in the world, however, has been subject to a great many changes and refinements, including both empirical refinements (as we get better at making predictions about correlations of events) and conceptual refinements. Clearly much of social science could be considered as providing empirical refinements of the manifest image, and Sellars holds that much of traditional philosophy— including both continental philosophy and the Anglo-American traditions of common sense and ordinary language philosophy—can be considered as attempts to construct adequate conceptual accounts and/or refinements of the manifest image (1963/1991, 8). The manifest image is distinguished from the scientific image insofar as the manifest image ‘‘by stipulation, does not include . . . the postulation of imperceptible entities, and principles pertaining to them, to explain the behaviour of perceptible things’’ (7). The scientific image, then, by contrast, is ‘‘the image derived from the fruits of postulational theory construction’’ (19), which ‘‘postulates imperceptible objects and events for the purpose of explaining correlations among perceptibles’’ (19).
146
ordinary objects
Having so defined the relevant images, Sellars goes on to argue that, although they cannot meaningfully be said to conflict (nor the former be replaced piecemeal with the latter), they are rival schemes between which we must choose—and given the choice, we must choose the scientific image. We would have no need to consider them rival images if they could be seen to be providing different images (perhaps from different ‘perspectives’ or at different levels of analysis) of what are ultimately the same things, but there is a crucial barrier to identifying the objects of the two images. The manifest objects described in the manifest image include the ordinary perceptible objects of our surrounding world—solid tables, fragrant flowers, noisy bells, and pink ice cubes (maybe not many of those). Why cannot these be identified with the imperceptible particles postulated by the scientific image—or rather, with large collections and systems of these? In general, there is no reason why large, complex perceptible objects cannot be identified with systems of imperceptible particles, where the whole may have properties that each of its parts lacks (being composite, having a certain shape and size . . .). But for such identifications to succeed, Sellars argues, the properties of the whole must be a matter of the intrinsic properties of and relations among the parts. Yet the homogeneous perceptible properties that pervade the objects of the manifest image—like the pinkness of the ice cube—cannot be construed as simply a matter of the intrinsic properties of the relevant fundamental particles and their interrelations: It does not seem plausible to say that for a system of particles to be a pink ice cube is for them to have such and such imperceptible qualities, and to be so related to one another as to make up an approximate cube. Pink does not seem to be made up of imperceptible qualities in the way in which being a ladder is made up of being cylindrical (the rungs), rectangular (the frame), wooden, etc. The manifest ice cube presents itself to us as something which is pink through and through, as a pink continuum, all the regions of which, however small, are pink. It presents itself to us as ultimately homogeneous. (1963/1991, 26)
Thus the two images cannot be taken as providing merely alternative views of the same entities; nor can they be taken as complementary, insofar as each is describing some subset of the objects of the world, for, as Sellars describes it, both images purport to provide ‘‘the true and, in principle, complete, account of man-in-the-world’’ (1963/1991, 25). The inability to identify the objects of the images, combined with the claim of each to completeness, jointly make the case for a rivalry
problems of rivalry with science
147
between the images. Moreover, since both purport to be true and complete, any account that attempted to incorporate both the manifest and scientific images ‘‘would contain a redundancy’’ (31).7 If the two images are indeed rivals then we must choose between either (1) giving primacy to the manifest image (while considering the scientific image to provide merely ‘symbolic’ representations of manifest objects, as Eddington seems to do), or (2) giving primacy to the scientific image, and considering the objects and properties depicted in the manifest image as mere ‘‘ ‘appearances’ to human minds of a reality which is constituted by systems of imperceptible particles’’ (1963/1991, 26).8 Given this choice, Sellars argues that we should choose the scientific image—not because (as the sorites arguments discussed in chapter 5 allege) there is some inconsistency within the manifest image, but simply because the scientific image provides a ‘‘more intelligible,’’ and more explanatory, account of what there is (29).
8.3 Is There Really a Rivalry? The question to address, then, is whether there really is a rivalry between the two images. The main case for a rivalry was as follows. (1) The objects and properties of the two images cannot be identified, but (2) each purports to be a true and (in principle) complete account of the world. But it is hard to see what it could mean for any ‘image’ or representation to purport ‘‘to constitute the true and, in principle, complete, account of man-in-the-world’’ (Sellars 1963/1991, 25)—or why, if any image did make such claims for itself, we should be at all inclined to accept them. Even from the way Sellars describes it, it does not seem plausible that the manifest image purports to be complete. For, as he describes it, the manifest image may identify correlations, appeal to the existence of causal explanations, and provide the framework on which scientific explanations may be built, but it cannot itself make good on any such causal claims or provide any explanations (1963/1991, 17). For that, the scientific image is necessary. So the manifest image itself seems to include an appeal to the need for a further, scientific image to posit the underlying entities that will explain the observed correlations, perhaps even including placeholders for whatever entities are included in the appended scientific image. This is consistent with the fact that the motivation to develop a scientific image must arise out of an existing manifest image, which (even in its own terms) is understood to be
148
ordinary objects
incomplete in some way, and to require supplementation with an image that postulates imperceptible objects and events to help explain the manifest correlations observed. But if the manifest image does not purport to be complete in any way that excludes the scientific image, does the scientific image nonetheless claim to be complete in a way that squeezes out room for the manifest image? Again, according to Sellars’s initial descriptions, the scientific image postulates ‘‘imperceptible objects and events’’ but does so ‘‘for the purpose of explaining correlations among perceptibles’’ (1963/1991, 19). From this description, it seems again as if it is built into the original idea of the scientific image that it is to be conjoined with rather than replace the manifest image it was originally designed to explain. In any case, as we have seen in chapter 6, we have independent reason to doubt that either could legitimately purport to provide a complete account of what there is. For I have argued there that any account of what there is presupposes a certain category or categories of entities under discussion, and so such accounts can only purport to be complete inventories of entities of those sorts or categories. But (if purporting to be true and complete) they only even purport to rule out other accounts that are done in the same terms, purporting to provide a complete account of things of the same sort or sorts as those of other accounts. Do the manifest image and the scientific image compete for completeness in this way? It seems not. For the manifest image and scientific image are each concerned with different categories of entities, and employ different characteristic sortal terms. So even if each purports to be complete in some sense (i.e. offering a complete account of things of those sorts), they still do not purport to be complete in any way that would make them compete. ‘‘But,’’ one might respond, ‘‘surely they do compete in the sense that both purport to offer a complete description of what things there are in the world.’’ The discussions of chapter 6 again should give us pause here, however. How is ‘thing’ being used here? Presumably not in a sortal use jointly accepted by those who speak in terms of the manifest and scientific images (if so, what is this use? What are the shared criteria of application and identity?)—the terms each uses are different, and purport to pick out different sorts of thing—that is why there is an appearance of a rivalry between them. One could attempt to form a more complete account of what there is by quantifying over all categories, and considering whether each of them has compliants—using ‘thing’ in a covering sense. We have seen
problems of rivalry with science
149
(x 6.6) reason to doubt whether this move can give us a genuinely complete inventory of everything there is, but even if it is possible, it will not help revive a rivalry here. For the manifest image and the scientific image do not employ all possible categorial terms—or even a large portion of actual such terms. Such images, in fact, are distinguished from each other in terms of the sortal and categorial terms each employs, with the manifest image omitting terms for imperceptible fundamental particles and the like, and the scientific image omitting terms for artifacts, social objects, and the like. Since each explicitly limits its range of discussion to a small portion of the actual categorial terms available, there is no question of either purporting to offer a complete account by first surveying all categorial terms and then enumerating all instances of each. And if each uses ‘thing’ in a covering sense that presupposes a different range of sortals, their resulting accounts of what things there are cannot be true rivals. In sum, the supposed rivalry between scientific and manifest image accounts of what there is can only arise based on the assumption that each ‘image’ purports to offer (at least in principle) a true and complete account of what there is (Sellars 1963/1991, 20). But properly understood, neither such image (with its own characteristic sortal and categorial terms) can really purport to offer a complete account of what there is. So there is at least no obvious sense in which the scientific image or the manifest image may legitimately purport to be complete in a way that would rule out the other. Nor has the case been made that there is any outright conflict between them. As a result, however impressed one is by the scientific image, and however jarring it may seem to those accustomed to the manifest image, accepting the scientific image does not require us to reject the idea that the ordinary objects of the manifest image exist. One final attempt to find a rivalry would be to claim that we can make sense of a neutral use of ‘thing’ (that neither involves treating that term as a sortal nor as a covering term) in terms of which we can say that the two images provide rivals to the claim to offer complete accounts of what ‘things’ there are. But it is hard to see how such claims to offer a complete account of what ‘things’ there are, using ‘thing’ neutrally, could possibly be made sense of. For if, as the neutral use requires, ‘thing’ is not being used as a sortal term, it does not come associated with the coapplication conditions that yield the identity conditions needed for counting, and so we cannot in principle answer the question whether a given list covers all of the things there are or if there might be more. So we have serious reason to doubt that such
150
ordinary objects
neutral uses of ‘thing’ could be used in claims of either image to offer a complete account of what things there are that could be held to rival that of the other. Thus both of these prominent arguments from competition with a scientific ontology at bottom rely on the idea that such a generic, category-neutral use of ‘thing’ can be made sense of. Arguments that there is a conflict between common sense and scientific ontologies relied on the idea that we could make category-neutral reference to some ‘thing’ about which they give conflicting reports, and arguments for a rivalry rely on the idea that both images may provide categoryneutral yet complete inventories of what ‘things’ exist. Of course, as I argued in chapter 6, this same purely neutral use of ‘thing’ (on which we can ask whether some ‘thing’ is composed in various situations) is required to make sense of arguments against ordinary objects based on the special composition question. I think it is very interesting, and very suspicious, that these apparently quite different arguments against the existence of ordinary objects all turn out to rely on this same presupposition. At the least, the fact that they do so should raise hopes that we are nearing bedrock, and close to seeing the true basis of debates about the existence of ordinary objects.
C c
nine
parsimony and ontological commitment
The final battle in arguments against ordinary objects generally takes place on the grounds of parsimony. Thus one might say, even if we are convinced (from chapter 1) that atoms and the baseball they compose are not causal rivals, so that we need not declare the baseball to be epiphenomenal, still, all the causal work done by the baseball can be accounted for by the work of its atoms, and so we do not need to posit the baseball to have a complete causal story. As Merricks puts it, the claim, for example, that there is a statue that is colocated with a lump ‘‘seems to imply—as far as causal explanations are concerned—a needless multiplication of physical objects’’ (2001, 40), and it might be said, if we do not need to posit these entities, then Occam’s razor, ‘‘It is vain to do with many what can be done with fewer,’’ enjoins us to eliminate them. Arguments from parsimony should be more carefully examined, however. In fact, both of the strategies employed above can be used to fight a two-front battle against the idea that parsimony requires us to deny ordinary objects. First, as with the causal principle (chapter 1) and various resemblance principles (chapter 4) we should be careful to consider whence the principle enshrined in Occam’s razor derives its plausibility, and whether it can really be generalized in a way that carries over to show that we should eliminate, for example, baseballs in favor of atoms arranged baseballwise. Second, we can utilize the discussion of counting and methods of answering existence questions in chapter 6 to reexamine whether 151
152
ordinary objects
eliminativists (who, like Merricks and van Inwagen, deny composite material objects and paraphrase talk apparently about them) genuinely offer an ontology of ‘fewer things’ than the realist does. Handling this issue will require some background work on the idea of ontological commitment and the uses of paraphrase—work that will finally contribute to an argument for the existence of ordinary objects. Unlike many who argue for accepting various kinds of entities, I will not be arguing that there are phenomena that cannot be explained (or not as well explained) by these eliminativists. Thus I will not claim that ordinary objects make a causal contribution that could not otherwise be accounted for on such eliminativist views. Nor will I argue that, unless we accept such objects, we cannot properly account for elements of our speech, thought, practices, or the apparent truth-values of sentences apparently about them.1 In short, I will not be arguing that ordinary objects are indispensable, but rather that they are (nearly) unavoidable. For, I will argue, they are ‘minimal’ relative to the sorts of entities with which eliminativists (of these stripes) try to replace them. As a result, I will argue, theories that try to do without them almost always turn out to be theories that are implicitly committed to there being tables and chairs, sticks and stones. So whether or not you wanted ordinary objects in your ontology, it’s rather likely that you’ve had them all along. Nonetheless, there remains a more severe form of eliminativism that would not contain implicit commitment to ordinary objects and so genuinely be more parsimonious. The costs of this severe eliminativism, however, are much greater than those associated with the more conciliatory eliminativisms of Merricks and van Inwagen. In section 9.7 I will address this alternative and assess its price.
9.1 Parsimony’s Plausibility In its paradigm plausible applications, Occam’s razor enjoins us to eliminate entities that are superfluous to causal explanations and, among empirical theories, (other things being equal) to choose that which requires positing the fewest entities. So like the causal principle, Occam’s razor derives its plausibility from cases in which its application is implicitly restricted to separate and independent entities, as it is when it is employed to help us choose among scientific theories or the hypotheses of detectives. But it is far from clear that the principle retains its plausibility when it is applied ‘neutrally’ substituting any terms
parsimony and ontological commitment
153
whatsoever, or that it applies as happily to choosing among philosophical theories that—if they may be thought of as providing ‘explanations’ at all—certainly are not in the business of providing competing causal explanations of anything. In fact, the plausibility of Occam’s razor seems to come into question in precisely those cases in which there are analytic entailments between the existence claims of the entities posited. While it may be obvious that a detective who supposes one murderer (rather than two) was responsible for the crime, or a scientist who posits one microbe rather than two as responsible for a disease, offers a more parsimonious theory than her rival, it is not at all obvious that someone who says that a drawer contains a (matching) left-hand glove and right-hand glove but not a pair of gloves offers a more parsimonious theory than his common sense rivals or can in any legitimate sense (any sense that truly gives us reason for preferring it) be said to posit fewer ‘things’. I have argued above that there is no ‘doubling up’ of causal work, entities, or properties where there are analytic interrelations between the claims to cause a certain effect, be composed of certain parts, or possess certain properties. The relevant principles that suggest there is illicit doubling up (including the causal principle, no-coincidence principles, and resemblance principles), I have argued, each require for their sensible application that they be restricted to cases in which the supposedly rival claims are (analytically) independent. It thus seems entirely plausible that the principle of parsimony enshrined in Occam’s razor, like these other principles, derives its plausibility from considering its use in empirical contexts, in which the analytic independence of the existence claims (e.g. about particles or perpetrators) is normally assured, and should only be accepted in a suitably restricted version. So if there are, as I have argued, analytic entailments between, for example, claims about atoms arranged baseballwise and claims about baseballs, it is not at all obvious that those who accept the existence of atoms arranged baseballwise but reject that of baseballs are offering a theory that is genuinely more parsimonious in a laudatory sense, or that those who accept both violate a legitimate and completely general metaphysical principle. In the sections that follow I will try to make a more precise case that, owing to the relations of meaning between (e.g.) claims about baseballs and claims about simples arranged baseballwise, eliminativists who seek to paraphrase claims about the former in terms of claims about the latter do not genuinely offer a sparer ontology.
154
ordinary objects
9.2 Parsimony and Counting The very idea of parsimony is the idea of positing fewer things to do the same (explanatory) work elsewhere done with many. Thus the idea of parsimony relies on the idea of counting, and questions of parsimony can only arise where entities can be counted up (so that the numbers of entities posited by different theories can be compared). This is fine when we can ask how many murderers or how many subatomic particles a certain theory requires to account for the evidence. But on the view I argued for in chapter 6, questions involving counting, for example, ‘How many things are there?’ presuppose a category or categories of entity to be counted. For counting claims rely on identity claims, the truth-conditions for which are category-relative (compare Dummett 1973/1981, 74; Geach 1962/1980, 63). If we accept this view of counting claims, then comparisons of how many ‘things’ the realist and eliminativist about ordinary objects each countenance cannot be made if ‘thing’ here is being used neutrally (not as a sortal term). Indeed, as I have argued in chapter 6, we have good reason to think that on the alleged neutral use of ‘thing’, ‘How many things are there?’ is an ill-formed, unanswerable question. On the other hand, if ‘thing’ is being used as a sortal term, the different ‘counts’ provided by the realist and eliminativist seem to be just the products of turning ‘thing’ into a sortal in different ways (by associating it with different application conditions); they are not comparable counts done in the same terms, and the different sums merely reflect differences in the semantics presupposed for the term ‘thing’. Once again, the results of chapter 6 about how existence and counting questions should be understood give us the tools to immediately diagnose the problem with this argument against ordinary objects. One might say that, instead of attempting a sortal-less counting, to make such comparisons of parsimony we may count things up using multiple sortals such as ‘atom’, ‘lump’, and ‘statue’—so that we count ‘things’ using ‘thing’ in a covering sense. We have seen in section 6.6 that it is unclear whether one could hope to get a total count of ‘things’ by substituting in all possible categorial terms (if one could, the counts in either case would pretty clearly be uncountably infinite, marking no distinction in quantity between them). But counts using a finite list of relevant categorial terms seem doable (if potentially misleading) and might be enough to make the needed comparisons of parsimony. Note, however, that in order for there to be a true comparison between countings of any kind, the same sortals must be used in one
parsimony and ontological commitment
155
count as another. So suppose we give both the eliminativist and the realist about ordinary objects the same sortals (these, just being terms, the eliminativist can hardly refuse—or if she does refuse the terms, the theories cannot be compared in terms of parsimony and the argument from parsimony cannot be made), with the same application and identity conditions. Does the eliminativist’s ‘count’ of things using a covering sense of ‘thing’ that includes these sortals now really come out lower, providing a genuinely more parsimonious theory? I will argue that it does not, so that on none of these possible readings can an eliminativist position like van Inwagen’s legitimately claim to be more parsimonious than a view that accepts ordinary objects. Some might suggest that the crucial point of parsimony, in metaphysics at least, is not to posit the smallest number of entities per se but the smallest number of kinds of entities. Thus, despite the points I made above about counting (it might be said), the eliminativist can offer a more parsimonious theory by denying the existence of kinds of things (e.g. chairs, tables) that the realist accepts. I will also argue, however, that (despite explicit protests) the usual eliminativist’s view is not really one on which there are no chairs, tables, or other kinds of things the common sense ontologist accepts, and so does not offer us a view more parsimonious on this measure either. The arguments that even the would-be eliminativists are (generally) tacitly committed to an ontology of ordinary objects provide the beginning of the case for ordinary objects.
9.3 The Case for Ordinary Objects Consider an apparent disagreement about existence, for example, between the common sense ontologist and the eliminativist (organicist, nihilist, or what have you) about whether or not chairs exist. I have argued in section 2.5 and chapter 6 that the question of whether, for example, ‘Chairs exist’ is true is to be addressed by first determining what category of entity ‘chair’ was to refer to in a presupposed range of prior uses, and seeing what the associated application conditions are. We must then examine whether or not these application conditions are fulfilled; if they are, then something of the relevant kind exists, since the application conditions for the sortal term establish the existence conditions for a member of the sort. So if we give the eliminativist and the realist the same sortal term ‘chair’, with the same application conditions standardly associated by competent speakers, is it really true that
156
ordinary objects
the world-situation the eliminativist accepts is different from that the realist accepts, as in the former there are no chairs? The eliminativist certainly accepts that there are situations in which well-bonded particles, ‘arranged woodwise,’ are then assembled by an artisan with the intention of creating a seating device for one person with back support and in which these particles, so arranged (now ‘chairwise’) can indeed jointly perform the intended function of comfortably supporting a seated person. It is precisely the eliminativist’s ability to accept this that makes her theory palatable, distinct from the ‘madman’s’ belief that there are no chairs, and able to account for some sense in which claims like ‘There are two chairs in the next room’ are true (or, as Merricks, 2001, 171–85, puts it, ‘‘nearly as good as true’’). But it seems the application conditions ordinarily associated with the term ‘chair’ are perfectly well satisfied in that situation. Merricks considers something close to this line of objection (15–6) and mentions the following as extra conditions (beyond there being particles arranged statuewise) that he takes as required for there to be statues: (1) that ‘‘the atoms arranged statuewise stand in the relation of composing something to one another’’ (15), and (2) that (in a situation with n particles arranged statuewise) ‘‘there are at least n þ 1 things in that room.’’ But if the earlier work of chapters 7 and 6 (respectively) was correct, one cannot use arguments from composition or counting of things as evidence that the conditions it takes for there to be ordinary objects are not met. So putting aside suggestions like these (which have been dealt with), what more could it be supposed to take? It seems the world the eliminativist describes is one in which there are chairs, according to the application conditions associated with the sortal term ‘chair.’2 As long as they are provided with the same sortals with the same application conditions as those used by the friend of ordinary objects, it is hard to see how realists and eliminativists could genuinely differ in their answers, and so it looks as if eliminativists do not really provide a more parsimonious ontology after all.3 Indeed, given the above method of addressing existence questions, it seems that unless you accept some sort of massive error or conspiracy theory according to which there are never particles so arranged (which clearly eliminativists do not, Merricks, 2001, 2), you should accept that there are members of most of the kinds named by our ordinary terms— providing, at last, the beginnings of an argument for accepting ordinary objects. For, as long as one accepts that there are, for example, artisans who arrange particles into certain shapes with certain intentions, and that those particles arranged chairwise are collectively capable of
parsimony and ontological commitment
157
supporting a seated person, and so on, one accepts a situation in which (according to the application conditions ordinarily associated with the sortal ‘‘chair’’) there are chairs. One can only then deny the existence of chairs by supposing that really there are some further conditions for there being a chair that the above situation does not meet (compare Johnston 1997, 57–9). But the eliminativist cannot claim to have ‘discovered’ some real existence conditions for chairs beyond those entailed by the semantic rules associated with ordinary use of the term ‘chair,’ for as we have seen, the basic application and identity conditions associated with sortal terms are established by these rules. So whatever these additional conditions are supposed to be, they would have to be enshrined in the ordinary meaning and use of the sortal ‘chair.’4 But it is hard to think of any candidate ‘extra’ conditions that are remotely plausible as part of the usual requirements associated with the word ‘chair’ (and the lack of which would require us to revise all of our previous paradigm applications of the term), ordinary language being rather noncommittal on issues to do with microstructure and the like, and easily satisfied with perceptible, functional, and intentional features about which even the eliminativist does not suppose us to be mistaken. There is thus a presumption against radically revisionary views that would take chairs, organisms, and other familiar objects not to exist, provided that (as I have argued) claims of existence are to be resolved by determining whether the application conditions are fulfilled, and that the application conditions for those ordinary terms are established by ordinary competent speakers. If the serious ontologist disregards the application conditions standardly accepted by competent speakers in favor of higher metaphysical conditions, then her denial that these conditions are met tells us nothing about whether or not there are any chairs, for if she shifts the application conditions she shifts the terms of discourse and is not denying the existence of our familiar chairs. The standard reply of the serious ontologist is to say that in fact, an essential condition for there being a chair is not met: for there to be a chair, there must be some (one) thing there composed by that wood (van Inwagen 1990, 100), but there is no such thing, and thus there is no chair. The eliminativist and realist ontologist in each case disagree on whether or not there is any thing at all in the disputed situations, and so they also disagree about whether or not there is any chair, and a genuine difference between them is revived. The idea that specific ontological debates (e.g. about chairs) can be revived in this way relies on the idea that claims about whether or not
158
ordinary objects
there is a thing here are truth-evaluable. In chapter 6, I laid out three ways of understanding such existence claims: (1) as using ‘thing’ sortally, (2) as using ‘thing’ in a covering sense, such that it analytically follows from the truth of ‘There is some S here’, for any sortal S, that ‘There is some thing here’ is true, or (3) as involved in an alleged neutral use of ‘thing.’ But (1) clearly turns the eliminativist’s denials that there is any ‘thing’ here into the uninteresting claim that—in her idiolect, on her sortal use of ‘thing’—tables are not things—a reading eliminativists are quick to disavow (van Inwagen 1990, 99–100). Nor can the covering use help the eliminativist here, since on the covering use of ‘thing’, disagreements about what or how many ‘things’ or ‘objects’ there are must be based on specific disagreements about whether or not there are things of a given sort or sorts (e.g. whether there are artifacts or organisms or . . .). But then, on these rules of use for ‘thing’, the eliminativist can’t say that there is no table in a certain situation on the grounds that there is no thing there: for there is no way to establish that there is no thing there except, at least in part, by establishing that there is no table there (and no chair, no fork . . .). Whether or not there is a table there must be establishable independently based on the fulfillment or nonfulfilment of application conditions that don’t themselves appeal to the existence of a thing. But, as I have argued above, disagreements about specific existence questions about whether there are artifacts, organisms, and so on are hard to come by if we address these questions by way of determining whether or not the standard application conditions are fulfilled. The only way to revive those debates seemed to be in terms of debates about whether there is some thing or object here. So, in short, while there may be a legitimate covering use of terms like ‘thing’ or ‘object’, such uses do not provide a promising way of reviving metaphysical debates about whether there are chairs or other ordinary objects. All hopes for reviving the debate between eliminativists and common sense realists about ordinary objects, then, rely on the idea that there is some neutral use of ‘thing’ on which the eliminativist may deny, and the realist affirm, that there is some thing composed by the properly arranged particles. But I have already given reason in chapter 6 to cast doubt on whether such completely neutral existence claims are really well formed and truth-evaluable. Just as it enabled us to diagnose problems with the arguments based on the special composition question and rivalry with a scientific ontology, the hypothesis that such generic existence claims are ill formed and not truth-evaluable
parsimony and ontological commitment
159
immediately enables us to diagnose where this response of the eliminativist goes wrong. It is often—aptly enough—remarked that our definitions do not make it the case that anything exists;5 as Stephen Yablo says, ‘‘the knock against this has been the same ever since Kant; from the conditions a thing would have to satisfy to be X, nothing existential follows, unless you have reason to think that the conditions are in fact satisfied’’ (2002, 221). Similarly, Theodore Sider notes that while ‘‘I am free to stipulate any necessary and sufficient conditions for falling under the extension of ‘keyboard’ that I like . . . no such stipulation will guarantee that there is something satisfying those conditions’’ (2001b, xix). That all seems correct, and is perfectly consistent with the view here6— the point is not that (as in ontological arguments for the existence of God) the terms in question are defined in ways that ensure that something meets the definition. On the contrary, terms for ordinary objects require that certain conditions in the world obtain in order for them to apply. Instead, the point is that once the relevant conditions for the existence of tables, chairs, and so on are laid out, it’s overwhelmingly obvious that the world does satisfy these conditions, and that denying that would require a conspiracy theory—or at least a form of eliminativism much more severe than that we have been discussing. And so it seems that if eliminativists accept the same sortals as the realist, and accept the same application conditions governing those sortals as are enshrined in the ordinary use of the terms, they do not really end up with fewer ‘things’ than the realist, even if we can get a comparative count of how many things there are on each theory by employing a covering use of ‘thing’. This also completes an argument for ordinary objects: once you note that the existence conditions for such things are established by our practices, and accept that those conditions are quite minimal, it is rather obvious that they are fulfilled, and so that there are such things.7
9.4 Trivial Transformations and Ontological Commitments I have argued that eliminativists do not genuinely accept an ontology of fewer ‘things’ than their rivals, for it turns out that their ontologies entail the existence of ordinary objects, so that they are committed to such
160
ordinary objects
things just as much as defenders of a common sense ontology are. This, however, will strike many readers (not just eliminativists) as a surprising conclusion. For after all, is there not at least this crucial difference between eliminativists and those who accept ordinary objects: eliminativists do not quantify over ordinary objects (van Inwagen 1990, 108–11), while defenders of a common sense ontology do? If there really is that difference, then, according to Quine’s criterion of ontological commitment, ‘‘we are convicted of a particular ontological presupposition if, and only if, the alleged presuppositum has to be reckoned among the entities over which our variables range in order to render one of our affirmations true’’ (1953/2001, 13), it simply follows that eliminativists do have a more parsimonious ontology than their realist rivals. Based on this criterion of commitment, eliminativists and realists have generally agreed on at least one thing: that the eliminativist does offer a sparer ontology. The main debating point, then, has been whether or not Occam’s razor enjoins us to accept that sparer ontology, for after all, we are bound by it to accept a more parsimonious theory only if it can ‘do’ all that more profligate theories can. One thing any decent philosophical theory must be able to do is to preserve various common sense intuitions, for example, that claims like ‘There are two chairs in my study’ and ‘There are two-bedroom houses in some cities that are more expensive than any houses in other cities’ are true (or acceptable, or nearly as good as true . . .) while others like ‘There are more cars in Alaska than California’ are false (or unacceptable, or not nearly as good as true . . .). The natural way to formulate such claims would involve quantifying over chairs, houses, cities, and cars, and so it looks at first glance as if the eliminativist must treat each as equally false, in gross violation of our ‘common sense’ evaluations of these claims. To meet this challenge without adding to their ontology, according to the traditional Quinean criterion, eliminativists must find a way to paraphrase sentences like these that will enable them to account for the apparent difference in truth-value of claims of these sorts without quantifying over chairs, houses, cities, or cars. For, according to Quine, if an apparently committing mode of speech can be paraphrased so ‘‘as to show that the seeming reference to [the offending entities] on the part of our bound variable was an avoidable manner of speaking,’’ then we are not really involved in ontological commitment to entities of that kind (1953/2001, 13). Thus, for example, van Inwagen takes pains to argue that despite his denial of the existence of composite inanimate material objects, his
parsimony and ontological commitment
161
ontology is nonetheless not in conflict with common sense. He does this by showing how one can paraphrase standard common sense claims about chairs, and so on, into a ‘language of refuge’ that does not quantify over any inanimate composite material objects, but is able to explain the sense in which they are true (or false). While such paraphrases are not presented as having the same meaning as the ordinary language claim (or expressing the same proposition), they are supposed to, in some sense, ‘describe the same facts’, and explain what it is that’s right about the claims that seem true, when expressed in ordinary language (1990, 110–2). Thus, for example, the ordinary language claim ‘‘There are two valuable chairs in the next room’’ might seem to conflict with van Inwagen’s claim that there are no chairs. Nonetheless, by paraphrasing the claim into a language of refuge, he can explain the sense in which there’s something right about it: it remains true on his theory that a certain (chair-shaped) region of space is filled with simple particles ‘‘arranged chairwise,’’ and so the prior sentence is truly reporting the existence of relevant facts (101–2 and 108–14). Merricks (2001, 171–2) and Horgan and Potrcˇ (2000, 254) similarly offer a sense in which such claims are ‘nearly as good as true’ or ‘semantically correct’ (respectively) despite the fact that they are strictly speaking false, or false according to the higher standards of ‘direct correspondence’. I have no intention of impugning the clever paraphrases devised by van Inwagen and others, which (when properly done) properly preserve the role human intentions, practices, and so on play in the truthconditions of these sentences. Instead, what I wish to impugn is the idea that a theory’s ontological commitments are exhausted by those entities that must lie in the range of its quantifiers for its sentences to be true. While this may be sufficient, it is (I will argue) not a necessary condition for ontological commitment, and so in many cases, such paraphrases do not relieve a theory of ontological commitments but merely sweep them under the carpet. This is precisely the case with the eliminativist’s paraphrases of common sense claims, and so, I will argue, we should reject the eliminativist’s claim to avoid ontological commitments by way of paraphrases, and (more generally) should reject Quine’s criterion of ontological commitment. I have argued in chapters 2 and 3 that to disambiguate whether or not singular and general terms refer, and if so to what they refer, these must have some minimal associated conceptual content. This conceptual content establishes the sort of individual (or kind) that the term is to pick out, outlining frame-level conditions for the existence and
162
ordinary objects
identity of the referent, if any. I have also argued that the conceptual content in these terms is enough to ground certain conceptual truths based on interrelations among these terms, since the conditions for application of one term (or set of terms) may be necessary, sufficient, or otherwise relevant to the application of others, or the satisfaction of the truth-conditions for a sentence may be sufficient for the application of a new nominative term. One way the analytic interrelations among elements of language become evident in natural language sentences is through redundancies. Thus, as we have seen in chapter 1, it is redundant to utter ‘He bought a house and a building,’ since (given the meanings of ‘house’ and ‘building’) his buying a house analytically entails that he bought a building—no more is required. Much the same goes for Ryle’s example: ‘He bought a [matching] left-hand glove and a right-hand glove and a pair of gloves’—in this case, not a single fact, but three (that he bought a left-hand glove, and he bought a right-hand glove, and they match) guarantee the truth of ‘He bought a pair of gloves,’ but nonetheless the second half of the sentence is redundant with respect to the first, and the truth of the first half of the sentence is sufficient to ensure the truth of the second half, even though here a new nominative term (‘pair of gloves’) is introduced. This sort of phenomenon can be observed even more widely in the pleonastic transformations (brought out by Stephen Schiffer, 1994, 1996, 2000) that permit us, in English, to make trivial transformations from, for example, (1) ‘Jane hit John’ to (2) ‘A hitting of John (by Jane) occurred’ and from (3) ‘The whale is white’ to (4) ‘The property of whiteness is possessed by the whale.’ In cases like these—considered as part of normal English speech—the latter statement is simply a wordier way of making the former statement, the truth of the former is analytically sufficient to ensure the truth of the latter, and stating both would be redundant. But the latter sentences apparently commit us to the existence of things of new sorts, hittings and properties, not mentioned in the former. In our terms, the truth of the former sentences 1 and 3 is in each case sufficient to guarantee that the application conditions for the new nominative terms introduced in the latter sentences 2 and 4 are met, ensuring that there are hittings and properties (respectively). Nonetheless, the two statements in each case would be represented rather differently in first-order quantified logic. Statement 1, rendered into standard form, would read ‘There is some x and there is some y, such that x Janeizes and y Johnizes, and x hit y’, requiring no
parsimony and ontological commitment
163
quantification over events, while statement 3 would read ‘There is some x such that x is whaline and x is white’. The transformed statements 2 and 4, however, introduce singular terms for kinds of entities not mentioned in the earlier statement—an event (a hitting) and a property (whiteness), which are apparently guaranteed to refer provided that statements 1 and 3, respectively, are true. But statements 2 and 4, by quantifying over entities of these sorts, make explicit the tacit commitments of the original sentences, depositing an event or property in the lap of even those who did not begin by quantifying over them. Elsewhere (2001b) I have suggested that we call sentences such as statements 1 and 3 ‘basic’ sentences, and sentences such as statements 2 and 4 ‘transformed’ sentences. Where singular terms for a given kind of entity may be derived by pleonastic transformations from basic sentences that don’t quantify over anything of that kind, such that the application conditions for those terms are guaranteed to be fulfilled provided the basic sentence is true, we may say that those entities are ‘minimal’ relative to the situation described in the basic sentence. Terms for states and events are guaranteed to refer provided the more basic sentence is true because the conditions required to make true, for example, ‘Jane hit John’ are analytically sufficient to make true ‘There was a hitting.’ Thus, although there are real conditions required to ensure the existence of such things as states and events, these conditions are laid out in a way that ensures they involve nothing more than what is required to ensure the truth of the more basic sentence (lacking state or event terms). Now consider the eliminativist about events or properties who attempts to avoid commitment to these entities by translating all sentences apparently requiring quantification over events or properties back into sentences of a form like statements 1 or 3, which avoid such quantification. What has really been achieved here? If statement 2 is (as English usage seems to ensure) really redundant with respect to statement 1, and statement 4 with respect to statement 3, then the truth of statements 2 and 4 is guaranteed by the truth of statements 1 and 3. To deny that they are true, asserting that, since there are no events or properties, statements 2 and 4 are false, although statements 1 and 3 are true, requires that we sever the trivial connections permitted in ordinary speech between basic sentences and their pleonastic transformations, by treating what seem to be equivalent statements as statements lacking the same truth-value. It also means that the eliminativist must be tacitly and artificially inflating the application conditions for ‘hitting’, exaggerating what is
164
ordinary objects
required to make it the case that some event (a hitting) occurred, or that some object have a property (whiteness). For according to ordinary usage, nothing more is required than what is already required for the truth of the basic sentences that do not explicitly quantify over events or properties. If, on the other hand, ordinary usage and the standard truth-conditions for sentences like statements 2 and 4 and application conditions for their nominative terms are retained, someone committed to the truth of sentences 1 and 3 is thereby already committed to the truth of sentences 2 and 4, and with it to the existence of events and properties, and so gains nothing by avoiding the forms of speech expressed in the transformed sentences 2 and 4.8 What has all of this to do with the paraphrases proposed by eliminativists, designed to make good on the claim that they can preserve many common sense intuitions about the acceptability or unacceptability of sentences involving singular terms for common sense objects without having to quantify over such entities? Consider the paraphrases offered by van Inwagen. A sentence like ‘There is a chair here’ he would paraphrase as ‘There are particles arranged chairwise here,’ thereby quantifying only over particles, not chairs, thus supposedly avoiding commitment to chairs while preserving the sense in which the original statement is true (it may be paraphrased as the second, which is true). But what does ‘There are particles arranged chairwise’ mean? In line with his desire to preserve common sense claims and practices, van Inwagen nicely describes conditions for the existence and ongoing maintenance of such arrangements as based on artisans rearranging objects in space, causing bonding relations to hold so that these particles can jointly perform the characteristic supporting activities of chairs, with the arrangement persisting as long as there is a continuous history of maintenance (1990, 127–35). By averting to function and intentions as part of the truth-conditions for the ordinary claim, van Inwagen can do far better at preserving these appearances of truth and falsehood than a simple reductionist could. But if we do adopt van Inwagen’s proposed manner of speaking in his ‘‘language of refuge,’’ it seems clear that sentences containing apparent reference to chairs may be arrived at through pleonastic transformations from the language of refuge. For according to our ordinary practices in discussing chairs and other artifacts, nothing further is required for there to be a chair than that certain objects are bonded and arranged by an artisan in order that they fulfill certain typical functions, and that they are maintained in that arrangement for a time. If there were not already such a term as ‘chair’, from the language of refuge
parsimony and ontological commitment
165
statement ‘there are particles arranged chairwise’ one could (according to standardly licensed hypostatizations) still form the nominalization ‘a chairwise arrangement (of particles)’ or, for short, ‘a chair,’ which, since it requires no additional conditions for its application, will be guaranteed to refer provided the initial statement is true.9 Chairs, on this analysis, are minimal relative to particles arranged chairwise (in van Inwagen’s sense).10 But if that is the case, then nothing is gained by disallowing ordinary modes of speech that explicitly quantify over chairs in favor of paraphrases that, while not explicitly quantifying over chairs, nonetheless are such that their truth is analytically sufficient to guarantee that there are chairs. The complicated paraphrases do not make way for a sparer ontology; they only hide the commitment to chairs. Treating the paraphrases as true and the direct claims about chairs as untrue could only be done by artificially inflating the application conditions for ‘chair’ beyond those enshrined in normal use of the term. But as I have argued in chapter 3, the application conditions for our terms are, at bottom, established by the concepts of grounders, and claims to have ‘discovered’ some further conditions should be met with suspicion. Trenton Merricks approaches the same challenge by treating common sense beliefs of the form ‘F exists’ as ‘nearly as good as true’ ‘‘if and only if (i) ‘F exists’ is false and (ii) there are things arranged F-wise. So, for example, ‘the statue David exists’ is nearly as good as true because (it is false and) there are some things arranged Davidwise’’ (2001, 171). Merricks provides a somewhat different reading of ‘arranged statuewise’ from van Inwagen’s: ‘‘atoms are arranged statuewise if and only if they both have the properties and also stand in the relations to microscopica upon which, if statues existed, those atoms’ composing a statue would nontrivially supervene’’ (4). Why the clause ‘if statues existed’? From Merricks’s viewpoint, it is clear—so that one can borrow whatever conditions are relevant (according to the defender of common sense) without (apparently) committing oneself to the existence of statues (see chapter 1). But the clause ‘if statues existed’ makes a difference to the meaning of ‘arranged statuewise’ only if something other than the presence of the properties and relations on which composing a statue would supervene (if there were any) might make the difference regarding whether or not there are statues—but what else could be thought to make such a difference (and so to make the supervenience ‘nontrivial’)? (The standard suggestions, that it would take possessing causal powers, something being composed, there being an additional object, and so on have of course been dealt with in
166
ordinary objects
chapters 1, 7, and 6, respectively.) To think that something other than the relevant physical objects, properties, and relations (e.g. some marble being intentionally carved in a certain way by an artist intending to make a statue . . .) is needed in a certain situation for a statue to really supervene seems to artificially inflate the conditions needed for there to be a statue. But if nothing else could make such a difference, then the existence of atoms arranged statuewise does analytically entail the existence of a statue. (It is important to note that saying that the existence of tightly bonded atoms being arranged in certain ways by people in certain contexts with certain intentions analytically entails the existence of statues is not a reductive move, to say that all talk about statues may be translated into talk about atoms in such situations. The entailment is only one of sufficiency [such that a competent speaker could, from the first situation description, infer that there was a statue in that situation without need of any further empirical information]. The claim is not one of necessity—the concept of ‘statue’ is plausibly thought to be applicable in a way that is not committal about whether statues are ultimately made up of atoms, or a continuum of stuff, or something else. But the existence of some such material basis appropriately arranged in the appropriate context is clearly sufficient for the application of the term ‘statue’.) The general principle of relative minimalism as I have been describing it is that, when we have a sortal term ‘S’ associated with consistent application conditions and coapplication conditions, we may derive new existential claims (of the form Ax(Sx)) from a theory that initially incorporates no mention of Ss, just in case some sentence of that theory makes claims, the truth of which guarantees that sufficient conditions for the existence of an S are fulfilled. So there may be existential introductions, enabling us to derive new true existential claims about Ss from other claims that originally made no mention of Ss whatsoever. While this may run against the grain of the Quinean approach to ontology, it is perfectly consistent with a great deal of our ordinary talk about existence. Thus, for example, we normally allow that from the fact that two people willingly and understandingly sign certain documents and make certain vows, a marriage comes into existence; by constructing a building and performing a consecration ceremony, a church may come to exist, and so on. Further truths about Ss may be derivable from the combination of the empirical facts reported in the basic statements of the original theory, plus the meaning/rules of use for the introduced term ‘S’. So
parsimony and ontological commitment
167
from the fact that there are (according to the eliminativist’s theory) simples arranged tablewise, we may infer that there is a table; from the fact that these simples were placed in their current bonding relations by Jones, we may infer that Jones created a table; from the fact that the sum of the weights of the simples is 20 pounds, we may infer that the table weighs 20 pounds, and so on. Claims derived from the meanings/ rules of use of ‘S’ (and any other relatively minimal terms related to it that are introduced along with it), the empirical facts reported in the basic sentences of the theory, and combinations of these two sources exhaust all there is to know about relatively minimal entities.11 If the above is correct, then we have reason to reject Quine’s criterion of ontological commitment—at least the ‘only if’ side that entails that ‘‘a man frees himself from ontological commitments of his discourse . . . [if ] he shows how some particular use which he makes of quantification, involving a prima facie commitment to certain objects, can be expanded into an idiom innocent of such commitments’’ (1953/ 2001, 103). For a theory does not avoid commitment to any entities by avoiding use of certain terms or concepts.12 So if a theory treats certain sentences as true that, when a new relatively minimal term ‘S’ is introduced, analytically entail the existence of an S, then the theory is (tacitly, all along) committed to Ss, regardless of whether or not any statement the theory asserts explicitly quantifies over Ss. Similarly, if a series of statements involves explicit commitment to Ss by quantifying over S things, that commitment cannot be avoided (it is merely hidden) by removing the term ‘S’ and rewriting all sentences that quantify over Ss in terms of statements that (when combined with the term ‘S’ and its rules of use) analytically entail that there are Ss. Quine’s test for ontological commitment ignores the fact that there are often implicit commitments to certain kinds of entities even where we are not yet quantifying over them—commitments that can be made explicit through pleonastic transformations, and that can’t be avoided merely by shunting offending noun terms (that require us to quantify over the relevant entities when rendered into standard form) back into other parts of speech, or by shifting discussion from entities of a given kind to the analytically sufficient conditions for their existence.13 This, of course, is not to suggest that Quine merely overlooked the possibility of such entailments bringing with them unwanted commitments—for Quine’s criterion of ontological commitment is of course closely tied to his rejection of analyticity. If there are no analytic truths, and no analytic entailments of the kind I have been describing, the above argument cannot be made, for the existence of things of one
168
ordinary objects
sort will never analytically entail that of others—there would be (on that view) no relatively minimal entities. Moreover, the very method of determining our ontology I have argued for above is clearly not one Quine could adopt, for only if there are analytic truths (based on frame-level conditions) can existence questions be posed as I suggest, by way of examining whether the application conditions associated with the relevant sortals are met. The work of this chapter thus relies on the results of chapter 2, in which analyticity was defended along with that of other relations among meanings such as the analytic entailments I have been averting to. Those who accept the existence of such relations among meanings, then, have reasons for concern about accepting Quine’s criterion of ontological commitment, and about accepting the idea that paraphrases like those offered by van Inwagen and Merricks may help avoid ontological commitments. Those who accept the existence of meanings (even in a form as minimal as that of the associated frame-level conditions advocated in chapters 2 and 3) also have the basis for accepting the alternative approach to existence questions defended here. The work of this chapter, then, is not meant to show that, within a Quinean perspective, Quine’s criterion of ontological commitment should be abandoned. It is, however, meant to show that those who have doubts about Quine’s rejection of analyticity (doubts I have tried to fuel in chapter 2) also have reason to doubt his standards of ontological commitment and to reject his method of addressing existence questions. More broadly, the work engaged in here (especially in chapters 2, 3, 6, and 9) is meant to show how a different, defensible, and even preferable total view can be developed on which questions of existence are handled in a very different way from the familiar Quinean paradigm.
9.5 The Point of Paraphrase The result that paraphrases typically do not provide any assurance of fewer ontological commitments goes against the grain of much contemporary metaphysics, where paraphrase has become a way of life.14 So at this stage, the question is likely to be raised: Are paraphrases never useful? Explicitly metaphysical paraphrases are often introduced by analogy with paraphrases offered in cases in which scientific investigation seems to have disproven common sense claims. Thus, for example, van Inwagen introduces his paraphrases of sentences about chairs
parsimony and ontological commitment
169
in terms of simples arranged chairwise first by analogy with postCopernican paraphrases of sentences such as ‘The sun moved behind the elms’ in terms of sentences involving the rotation of the earth and visibility of the sun (1990, 101). He also uses a second analogy of an imaginary situation in which a group of people called ‘Pluralians’ saw what they took to be large tiger-like animals on the edge of their territory, and used the term ‘bliger’ to (attempt to) refer to them, but where biologists later discover that what the people saw was not (in any case) a single animal, but rather six animals that arrange themselves to look like a single larger one, for mutual protection (104). In such a case, he argues, we can paraphrase claims about bligers in terms of claims about animals ‘arranged in bliger fashion’ and preserve the sense in which they are true, just as we can paraphrase the claims about chairs in terms of claims about simples arranged chairwise and preserve the sense in which they are true, even though (according to him) there are neither chairs nor bligers. These paraphrases do seem useful as a way of describing the sense in which some reports involving sunrise and sunset, or the presence of bligers, are better than others. The paraphrases also help make evident what genuine facts were being (somewhat mis-) reported in claims about sunsets and bligers. But note the critical difference between these cases and paraphrases such as those van Inwagen offers of claims about chairs, or such as might be offered of claims about properties or events. In the case of sunrise, claims about the rotation of the earth relative to the sun and (consequent) visibility of the sun do not analytically entail that the sun changed its position. So if that is included in the literal truth-conditions of claims such as ‘The sun rose at 7:02 a.m. today,’ then we can accept that that claim is false while the paraphrase is true, without disturbing the entailments that are part of our ordinary use of the terms and concepts involved. (Whether or not that is part of the literal truth-conditions of claims about sunrise I will leave open; certainly it does not seem so anymore, although it may have been considered to be in times of convinced Ptolemaists.) The case of bligers is easier still. If the term ‘bliger’ is introduced as a term for an animal of a specific kind, then it is part of the success conditions for reference that there be some single animal (of the same species) present in all or most of the cases in which the reference of the term is supposed to be grounded. Since, according to the story, that condition is never met, the Pluralians were making a mistake—a mistake that has no parallel in the case of people who say things like ‘there are two valuable chairs in the next room’. In the deceived
170
ordinary objects
Pluralians’ use, the truth of ‘there is a bliger’ is not analytically entailed by ‘there are six animals arranged in bliger-fashion’—there really is some other condition required for the truth of the first sentence, according to the way the term ‘bliger’ was introduced (as a term for something of the category animal). And so one can accept that the second is true and deny that the first is true without misconstruing (or artificially inflating) the application conditions for ‘bliger’. Notice, however, that if (after the biologists’ visit) Pluralians become hip to the strange behavior of (multiple) animals in their area, and adopt a new term ‘pliger’ (or explicitly change the meaning of the old term ‘bliger’), where this is meant simply to refer to a bligerwise arrangement of animals (or perhaps simply to any reasonably persistent bligerlike appearance, regardless of its cause), their term can refer. In this case, the existence of a pliger would be analytically entailed by animals arranged bliger-fashion, and accepting the truth of sentences describing the latter state, but denying the truth of sentences such as ‘There are pligers’ would be as pointless (and involve as much misconstrual) as accepting the truth of ‘there are simples arranged chairwise’ but denying the truth of ‘there are chairs’.15 So, in sum, paraphrases are useful when, and only when, the sentence to be paraphrased really is false and there really are some extra conditions required for its truth that are not required for the truth of the paraphrase. Even then, the use of the paraphrase is primarily to demonstrate how we can preserve some sense in which the original sentence was onto something, more acceptable than others of its kind, and so on, without saying something false—not (simply) to provide a more parsimonious ontology.
9.6 Are We Committed to Extraordinary Objects? I have argued that it is not the case that all of our ontological commitments are owned by the bound variables in quantified statements we affirm, for new noun terms can be introduced with application conditions that are guaranteed to be met as long as the original quantified statements (which made no explicit mention of, nor quantified over, any such thing) are true. This move invariably (though inaccurately) raises the specter of defining things into existence, so it is worth pausing to explain what sorts of term can and cannot be introduced this way, and what can and cannot be said about the things to which
parsimony and ontological commitment
171
reference is thus introduced. This will help make evident why this is not a matter of defining things into existence, and why one cannot use a similar move to defend the existence not only of ordinary objects such as tables and chairs but also of extraordinary objects such as witches, miraculously appearing dates, and the like. It will entail the existence of certain types of entities (such as mereological sums) that are not commonly discussed in ordinary speech, but I will argue in section 10.3 that this is not an implausible conclusion at odds with common sense. To formulate the initial objection more sharply, one might ask: Why can we not (following the pattern for introducing terms for events or properties, or for [re-]introducing terms for chairs, given van Inwagen’s paraphrases) simply define into existence a ‘hoverball’, where a hoverball is said to be a pink sphere 3 feet in diameter, hovering over a person’s head, that exists if and only if there is a person, so that, given the existence of a person, the existence of a hoverball is guaranteed?16 Stephen Schiffer discusses a similar objection, based on ‘‘the concept of a wishdate, which I hereby stipulatively introduce thus:x is a wishdate ¼ df x is a person whose existence supervenes on someone’s wishing for a date, every such wish bringing into existence a person to date’’ (2003, 53). But there is a crucial difference between cases like these and the case of such relatively minimal entities as I have been describing. Our terms for genuinely relatively minimal entities are guaranteed to refer, provided the truth of a basic sentence, because the conditions required to make the basic sentence true are sufficient to ensure that the application conditions for the introduced term are met. But this is not the case with wishdates or hoverballs. The objector claims that ‘wishdate’ is defined as being a person whose existence requires only that someone wish for a date, or that ‘hoverball’ is defined as being a pink sphere over person’s head the existence of which requires only that of the relevant person. But the concepts of ‘wishdate’ and ‘hoverball’, as defined, also include the concepts of there being a person, or a pink sphere—these are part of what is included in the definitions. There are established meanings and uses of these terms, which ensure that there are certain substantive causal and other conditions required for it to be true, in any situation, that there is a certain person or pink sphere. Since these are incorporated as parts of the sense of ‘wishdate’ and ‘hoverball,’ it turns out there must be substantive conditions for the existence of wishdates and hoverballs beyond those conditions that are assured of being met if basic sentences such as ‘John wishes for a date’ and ‘Jackie is in the
172
ordinary objects
room’ are true. So the concepts of ‘wishdate’ and ‘hoverball’ are both self-contradictory, as both require application conditions beyond those needed for the existence of a wish or a person (respectively), yet both stipulate that no more is required. So the supposed definitions are ill formed and self-contradictory, and do not constitute a threatened reductio of the view that terms for genuinely relatively minimal entities may be introduced according to the model described above.17 The question then reverts once more to whether there are any tacit requirements for the existence of, for example, tables and chairs—built into the very meanings of the terms—that are not assured of being met by the existence of particles arranged tablewise (e.g. as van Inwagen describes them), as the concepts of person and pink sphere are built into the concepts of wishdate and hoverball. I claim that there are not; the ball is in the eliminativist’s court to say what they could be, and the onus is on the eliminativist to try to show that the associated conditions embed tacit contradictions (as ‘wishdate’ and ‘hoverball’ do). But while we can avoid commitment to such extraordinary objects as wishdates and hoverballs (since the corresponding terms are not minimal relative to other claims we accept), if you accept (as I have in x 9.3) that you are committed to Ks as long as you accept the truth of claims that (given the application conditions for ‘K’ and permitted redundant transformations) analytically entail the existence of Ks, then you must also accept more than stones, artifacts, and other ‘common sense’ objects. For other sorts of terms may be introduced with minimal existence conditions that are guaranteed to be met provided that other claims we accept are true. Thus suppose, to use an example of van Inwagen (1990, 126), we introduce the term ‘gollyswoggle’ to refer to a lump of clay with a particular very complicated shape, where it is taken to have that shape essentially. Given a lump of clay of that shape, the term ‘gollyswoggle’ is guaranteed to apply (and since it has its shape essentially, the gollyswoggle can be identical neither with the lump nor with any statue the lump may constitute). So it looks as if we are committed to the existence of gollyswoggles as well as lumps and statues. But of course there can be an endless number of terms for all the different shapes that might be (essentially) involved, and so (as van Inwagen puts it), ‘‘you must, as you idly work the clay in your fingers, be causing the generation and corruption of the members of a compact series of objects of infinitesimal duration. That is what seems to me to be incredible’’ (126). So, similarly, if one accepts that all it takes for there to be a mereological sum of objects x and y is the existence of x and y, then we are
parsimony and ontological commitment
173
also committed to the existence of all manner of arbitrary mereological sums of the other objects we accept. But accepting gollyswoggles and mereological sums of noses and towers might seem not only extravagant, but contrary to common sense. As Merricks puts the objection in a slightly different context, ‘‘with its explosion of macrophysical objects, massive amounts of colocation . . .—[this] is not the ontology of the folk. So one question is whether this departure from folk ontology is more or less plausible than eliminativism’’ (2001, 78).18 I will return to this issue in section 10.3, where I defend the resulting view’s claims to preserve a common sense ontology.
9.7 The Price of Avoiding Ordinary Objects There is, however, another kind of eliminativism more severe than that advocated by eliminativists like Merricks and van Inwagen. As we have seen, the strength of their views lies in their ability to preserve the apparent differences between true and false (or nearly true and nearly false, acceptable and unacceptable) sentences involving terms that are supposed to refer to ordinary objects, despite their (ostensible) denial that such terms refer. They do this by offering paraphrases of such sentences in terms that make no use of terms for ordinary objects, but only refer to simples arranged ordinary objectwise, where this last locution appeals to the intentions of artisans, the maintenance practices of people in the community, the ability of the simples jointly to fulfill various intended functions, and/or whatever other features are ordinarily built into our claims about the existence of ordinary objects. But their ability to provide such nice paraphrases is also the Achilles’ heel of their position. For, as I have argued, if they build in all of the relevant conditions, then their view is one on which sufficient conditions for the existence of ordinary objects are met, and so one on which there are ordinary objects, whether or not they engage in explicit quantification over them. What it would take to really deny the existence of ordinary objects, then, on my view, would be to deny that some of the necessary conditions for the existence of such things (as built into the ordinary use of the terms) are ever met. Thus, for example, one way to deny the existence of artifacts, on my view, would be to deny the existence of human intentions, since these are essential to the existence of artifacts (see Thomasson 2003b, forthcoming b). Someone who insists that only those entities referred to by physics exist, and so accepts only fundamental
174
ordinary objects
particles in purely physical (spatial, temporal, causal . . .) interrelations, does not seem to be subject to the argument that we can arrive at statements referring to artifacts by simple pleonastic transformations from sentences whose truth they accept, and so may offer a genuinely more parsimonious view by refusing to accept artifacts. To avoid begging (or engaging in a long sidetrack about) any questions about the possibilities for reductionism in philosophy of mind (see x 1.4), the argument may be made as follows. For any kind of ordinary object O, from the basic sentences the severe eliminativist accepts (referring only, e.g., to fundamental particles in physical interrelations), either we can or we cannot analytically infer the existence of Os by means of pleonastic transformations from these sentences. If we can, then the arguments above apply, and the theory is not really more parsimonious than one that is explicit in accepting Os. But it seems more likely that we cannot infer the existence of at least many ordinary objects, such as artifacts and social objects, by pleonastic transformations from this spare basis. For, for example, the truth of sentences about fundamental particles does not seem to analytically entail any truths about intentional states, the existence of which is a necessary condition for that of artifacts. But then, if nonreductivist arguments in the philosophy of mind are correct (e.g. Chalmers 1996, 106–7) there is some further fact that is a necessary condition for the existence of these ordinary objects that the severe eliminativist cannot account for (e.g. the existence of intentional states), and so the severe eliminativist will have to count all claims about them as equally false. But then the severe eliminativist faces the difficulty of accounting for the differentiating features of our talk and experience that take some sentences about artifacts and social objects to be true, and others false, for it seems she must hold an error theory of all such discourse. If that is the case, then while severe eliminativists do genuinely have a sparer theory than realists about ordinary objects, the latter cannot be considered to be a theory that does ‘with many’ what the severe eliminativist does ‘with fewer’, for the realist can, while the severe eliminativist cannot, account for these differences in our wellentrenched common sense beliefs about the truth and falsehood of claims about ordinary objects. As a result, Occam’s razor does not constrain us to choose the severe eliminativist theory, and in fact, if we think of one job of philosophy as providing some account of ordinary thought and language and of the apparent difference in truth-values of various common sense claims, then we have reason to choose the realist view in spite of the fact that it is not as spare as its competitor.
parsimony and ontological commitment
175
The question then remains what reasons we could have for preferring such a severe view, given its costs. The standard arguments based on a supposed rivalry with a scientific view, or contradictions or incoherencies within the common sense view, have already been dealt with. So it is hard to see why we should go down such a spare path when it means giving up the idea that philosophy should in some way be able to account for the sorts of distinction made in ordinary thought and speech.
C c
ten
a coherent common sense view
I have now offered diagnoses of where a great variety of arguments against ordinary objects go wrong, including those based on alleged causal redundancy, problems of colocation, vagueness, composition, rivalry with science, and parsimony. For the friend of ordinary objects, it should have been heartening to see how each of these arguments individually may be defused. But what should be interesting for philosophers of all persuasions is the way these superficially quite diverse arguments fall into some very noticeable patterns. In this chapter I will begin (in x 10.1) by showing how the diagnoses are interrelated, bringing out the common mistakes lying behind each of the arguments against ordinary objects. Once we have isolated the mistakes, we can also see how to avoid them. So in section 10.2 I will suggest how a common sense view may be developed that can avoid the above arguments against ordinary objects, and meet the challenge of showing how we can reflectively make sense of our unreflective common sense worldview. This positive view is based on the understanding of reference and of claims about identity and modality preliminarily defended in chapters 2 and 3. Their ability to provide a unified diagnosis of where the various apparently diverse eliminativist arguments go wrong should—at least for those who accept that our common sense views that there are such objects carry some epistemic weight—lend them a great deal of plausibility in addition to whatever the plausibility the direct arguments of chapters 2 and 3 lent them. In 176
a coherent common sense view
177
sections 10.3 and 10.4 I will consider some prominent objections to the positive view developed and defend its claims to common sense.
10.1 A Unified Diagnosis According to the diagnoses offered above, the arguments against ordinary objects each rely on one of three common mistakes: (1) accepting as ‘completely general’ metaphysical principles that fail to apply where there are analytic entailments; (2) failing to note that the most basic claims about existence, identity, and persistence conditions for the objects we refer to are analytic; and (3) treating generic existence and counting claims as truth-evaluable. The first kind of mistake lies behind arguments based on causal redundancy, and some arguments based on problems of colocation, property additivity, and parsimony. Each of these arguments relies on some very general metaphysical principle, for example, the causal principle, the no coincidence principle, various principles of property additivity, and Occam’s razor. While these principles derive their plausibility from considering claims that are separate and independent—including claims to causation, spatial location, property possession, or existence of two particles, statues, or persons—this plausibility does not carry over when we substitute in claims that are analytically interrelated. The key to avoiding each of these arguments, then, lies in accepting that there are analytic entailments, and that many plausible principles are implicitly restricted to claims without analytic interrelationships. The failure to notice this problem is not surprising, since conversational prohibitions against redundancy, combined with our tacitly understood category restrictions (implicitly confining most discussions to a certain topic such as fundamental particles or lumps of matter, works of art or their prices) normally prevent us from asserting claims involving analytic entailments. But while we need not worry about analytic entailments in most practical discourse, they come to the fore in metaphysical discussion, which, being concerned with how things of different sorts are interrelated, is inherently cross-categorial. The key to seeing that there is neither problem nor profligacy in accepting scientific objects, collections of particles, common sense objects, social objects and cultural objects, lies in noting that where there are analytic interrelations among our claims, distinct ontological claims may be true without rivalry, redundancy, or reduction.
178
ordinary objects
Making use of this kind of diagnosis thus requires accepting that there are analytic relations among our claims. In chapter 2 I argued that we have independent reason to accept that there are such analytic interrelations, based on relations among the meanings of our terms, in spite of Quinean arguments to the contrary. Moreover, I argued that even those who are inclined to a broadly causal account of reference have reason to accept that reference is determinate and claims are truth-evaluable only to the extent that our terms are disambiguated by association with frame-level conditions of application and coapplication. This thesis has other important consequences of its own, for example, that the most basic claims about identity and persistence conditions and modal and categorial features of the things we refer to are analytic. The second common mistake behind arguments against ordinary objects (including those based on the grounding problem and on various problems of vagueness) lies in failure to acknowledge this. Those who press the grounding problem assume that an entity’s category, basic conditions of identity and persistence, or modal features must supervene on its microphysical nonsortalish properties and be explicable in those terms. But if the work of chapters 2 and 3 is correct, this turns out to be a mistake: the differences in the basic modal and categorial truths about, for example, a statue and lump are reflections of different analyticities for the sortals (or names) involved. The demand for a bottom-up explanation of the differences in basic modal truths about, for example, David and Lumpl, is then inappropriate, since these are fundamentally analytic truths, not requiring bottom-up truthmakers. Accepting that the basic conditions of identity and persistence for the objects we refer to (if any) are fixed with the determination of reference also enables us to explain why we should expect vagueness in our ordinary terms and their referents, although all such vagueness is ultimately explicable as a product of our ways of representing the world. The third common mistake, lying behind arguments from the special composition question, rivalry with science, and parsimony, is assuming that generic existence and counting claims about, for example, whether or not some ‘thing’ is composed in a certain situation, what things exist, or how many things there are, are truth-evaluable, and that the corresponding generic existence questions are answerable, even when ‘thing’ is used in an entirely category-neutral way. But the view of reference argued for in chapters 2 and 3 also gives us reason to think that that assumption is false and that the relevant claims are
a coherent common sense view
179
not truth-evaluable where the terms used are not categorically disambiguated. The fact that all of these apparently diverse arguments rely on this one crucial commonality should, I think, make us very suspicious—particularly because, as I have argued in chapter 2, we have independent reason to doubt that such claims are fully meaningful and truth-evaluable. If they are not, we cannot demand a completely category-neutral answer to the special composition question, considered as the completely general question of when there is some y composed by the xs—since we must supply some application conditions to establish what sort of y we are looking for before we can answer whether there is a y. As a result, the failure to answer the special composition question in its general form does not tell against the existence of composite inanimate material objects. Similarly, without coapplication conditions and the criteria of identity they establish for the objects (if any) referred to, we cannot establish whether any list is complete (containing all the things of the categories in question) or if some might have been omitted. And so, since the scientific and the manifest images use terms of different categories, there is no rivalry between them, even if each purports—in a sense—to be complete. Finally, comparisons about parsimony likewise cannot be made in completely category-neutral terms, enabling us to say merely how many ‘objects’ each theory has. For such comparisons require the use of categorial terms that (by supplying identity conditions) enable us to count objects. But when supplied with the same categorial terms with the same frame-level application and coapplication conditions, the differences between apparently competing ontologies begin to disappear. These results should be interesting even to those who remain agnostic about whether or not such generic existence claims are meaningful (and questions answerable)—for it at least demonstrates that all hopes for reviving genuine metaphysical debates about whether or not simples compose some thing, or how many things there are according to various theories, rely on the idea that there is some category-neutral sense of ‘thing’ or ‘object’ that does not involve treating it as a covering term whose application is analytically entailed by the application of any first-order sortal—or turning the term into a sortal of its own—and that can be used in offering a completely category-neutral inventory of what ‘things’ there are.1 The arguments above give us reason to doubt that such an account is available, and so to doubt that the apparent metaphysical debates that rely on them are genuine. Diagnoses like those above are sometimes dismissed as merely attempting to ‘provide linguistic solutions to metaphysical problems’ and
180
ordinary objects
sidestepping the deep metaphysical issues (e.g. about composition). But this is a misunderstanding. If a mother at work sends in an electrician after receiving a call from her young son reporting that the refrigerator isn’t running, the electrician who duly reports back that the refrigerator is fine—that it wasn’t running in the sense that it wasn’t doing laps of the garden, not in the sense that it wasn’t functioning—has not ‘provided a linguistic solution to an electrical problem’. Instead, he has shown that the belief that there was an electrical problem was based on a misleading use of language. So, similarly, the diagnoses offered above do not purport to provide linguistic solutions to metaphysical problems, but rather to show that what appear as problems for a particular metaphysical view (the view that there are ordinary objects) are in fact no problems at all, resulting as they do only from misunderstandings bred in misuses of language.
10.2 The Basis for a Common Sense View While it is interesting enough to see the patterns these arguments against ordinary objects fall into, what is still more interesting is that all of these mistakes (and thus all the diverse arguments that rely on them) may be avoided by adopting a unified picture centered on the simple thesis that our singular and general nominative terms have a basic conceptual content in the form of frame-level conditions of application and coapplication collectively established by competent speakers. If we accept this, we have the grounds for accepting that there are analytic entailments. Of course, to avoid mistakes of type 1 we must also accept that the metaphysical principles in question are restricted to cases in which there are not such analytic entailments, so accepting generally that there are analytic entailments is not quite enough on its own to avoid these mistakes, though it provides the crucial and contested claim behind this diagnosis. Second, as I have argued in chapter 3, this claim entails that the most basic claims about modality and about conditions of existence, identity, and persistence are analytic, and do not require any bottom-up worldly truth-makers, thus enabling us to avoid mistakes of the second kind. Third, this thesis shows why (as many had suspected) generic existence and counting claims using ‘thing’, ‘object’, and the like in neither a sortal nor covering use are simply not truth-evaluable, as the terms involved are too radically underspecified. Given this understanding of how all the arguments against ordinary objects come together, perhaps it should come as no surprise
a coherent common sense view
181
that arguments against the existence of ordinary objects have come into prominence over the past fifty years, as it has become increasingly common to reject the idea that terms have any form of conceptual content. The ability of this simple thesis about reference to help dissolve the mistakes behind each of the apparently disparate arguments against ordinary objects should lend a great deal of credence to the view that we have reached a proper diagnosis of the problems with arguments against ordinary objects, and the basis for a positive view subject to no such difficulties. According to that positive view, our language contains a wide variety of sortal terms, paired with a wide variety of application and coapplication conditions. For any sortal term ‘K’ with consistent and nonempty application and coapplication conditions, as long as the application conditions associated with ‘K’ are fulfilled, there are K-objects, and if there are, they are guaranteed to have the existence, identity, and persistence conditions fixed by the application and coapplication conditions for the terms. So we have reason to think that there are fundamental particles, plants and animals, sticks and stones, tables and chairs, and even marriages and mortgages, since (barring conspiracy theories) we have reason to think that the application conditions commonly associated with these terms are fulfilled. And we have reason to think that these have pretty much the identity and persistence conditions commonly associated with them. Those who defend (or attack) a common sense metaphysics often speak of a ‘layered’ ontology or of ‘levels of reality’, with physical particles at the base layer, biological entities at a higher level, social entities at a higher level still, and so on.2 There is, of course, much in this picture that fits with the view I have developed here, and with which I am sympathetic. Nonetheless, talk of a ‘layered’ ontology suggests altogether too much organization—as if we could then specify exactly how many layers there are and how they stack up, and ask, for any sort of thing, exactly what layer it belongs to. The view I am defending does not presuppose that questions like these are answerable. The picture is rather that application and coapplication conditions may be drawn out in various ways, and the conditions laid out may overlap in various ways, leading to all sorts of possible analytic entailments among existence claims and other claims. As a result of these entailments, there are also all sorts of relations of metaphysical dependence among the objects picked out by these terms, although these objects may be of too many sorts, with too many different and complex interrelations, to be capturable in a simple layered picture. Nonetheless,
182
ordinary objects
this picture makes it clear how distinct claims may be talking about features of the same world, although they do so using terms set up with different frame-level conditions. The work above shows how to develop a workable common sense ontology, and to demonstrate that it is not subject to the problems long thought to confront it. Its claims to being a common sense ontology are many. First, it enables us to accept the existence of ordinary objects, including common sense natural objects such as sticks and stones; and social, cultural, and institutional objects such as baseballs, statues, and dollar bills.3 Second, it enables us to retain common sense views about the conditions for the existence, identity, and persistence of various sorts of common sense object (e.g. that there cannot be a statue without intentions to make one, that a statue can survive the loss of small parts, etc.), neither artificially inflating these nor replacing them with conditions proper to other sorts of thing entirely (e.g. mere lumps of matter). Third, it enables us to preserve and take seriously the sorts of inference legitimated by ordinary language such as, for example, that if Fido bit Fifi, then we can say that Fido’s biting of Fifi occurred; that if there are particles arranged in a certain way (chairwise, say), by a certain person intending to make a chair, where the particles so arranged can perform the characteristic activities of chairs . . . then there is a chair; and so on. Fourth, it enables us to accept the existence of vague objects, allowing that Everest, for example, exists but simply has no sharp boundaries whatsoever. Finally (as I have argued in x 9.6) it enables us to do all of this without accepting the existence of extraordinary things we have no reason to believe in, such as hoverballs and wishdates. As we have just seen, the key to defending this lies in accepting that there is basic conceptual content to our terms, including frame-level application and coapplication conditions for sortal terms. This view itself, I admit, is no part of common sense. In fact, I think, our common sense experience is generally not focused on language (or other representations) and the relation between terms and referents at all (so it is also certainly not contrary to common sense), but rather is simply focused on the world. But while this thesis and its consequences for understanding identity and existence conditions, counting questions, and so on are not themselves part of common sense, if I am right they serve as the unnoticed bedrock on which an acceptable common sense ontology may be developed—or, if you prefer, they are the (at least sufficient) preconditions for a common sense view of the world to be made workable.
a coherent common sense view
183
10.3 Too Many Objects? Of course, there are objections to the claims that the ontology developed here is a common sense ontology. One common objection is that the principles I have argued for entail not only the existence of our familiar ordinary objects (along with physical objects, scientific objects, etc.) but also (as I have noted in x 9.6) that of many more objects than common sense accepts. This objection can take two forms. The first is to suggest simply that views that accept ordinary objects give us a counter-commonsensical number of objects. So, for example, Katherine Hawley describes the challenge facing constitution theorists as offering ‘‘some explanation of why we tend to say there is just one thing there’’ (2001, 163) when, for example, the theory officially accepts that there is a sweater and a length of thread. I think this is indeed an intuition that must be respected (though we should note, as I argued in section 4.1, that the claim that there is a sweater and a length of thread sounds bad largely because it violates the independence presuppositions that normally govern conjunctive claims). But by now it should be clear how a view like mine handles this intuition. As I have argued in chapter 6, the claim that there ‘is just one thing there’ is not well formed and truthevaluable if it uses ‘thing’ in a category-neutral sense; but the intuition that it’s true that there is just one thing there likely comes from our standard sortal use of ‘thing’ to count, for example, separate physical masses—using this sortal, there is only one. But the truth of the claim that there is one ‘thing’ here in that sortal sense does not undermine the idea that, if we use ‘thing’ in a covering sense involving both the sortals ‘piece of thread’ and ‘sweater’, we wind up with a count of two. The other form an objection like this often takes is to suggest not that the problem is simply countenancing more objects than there seem to be, but rather that we gain commitment to various specific kinds of things that have no place on a common sense view. Certainly it is true that common sense does not recognize the existence of gollyswoggles, mereological sums, and the like. Nor, of course, does it deny their existence—there are no terms in ordinary English for these things, and common sense understandably does not consider such things at all since, given our current range of practices, such entities would be quite irrelevant and uninteresting. But suppose one introduced the terms for such entities, and taught people the associated application and coapplication conditions. Would people then say that it’s just common sense that there are no such entities?
184
ordinary objects
You could try to lead them to do this with inflammatory remarks like ‘Look, then there are millions of objects there on your desk’ or ‘Then there is some object composed of the Eiffel Tower and my left ear’ and so on. But our earlier discussion of the term ‘object’ should make us suspicious of such rhetorical appeals. These appeals clash with our intuitions because the term ‘object’ has at least two different uses. First, there is the covering use of ‘object’ on which one can say that since there are sums (provided that the relevant parts exist) and there are gollyswoggles (provided there is properly shaped clay), there are such ‘objects’ as sums and gollyswoggles. This use of the term ‘object’ differs from the ordinary use in which ‘object’ is used sortally, roughly to pick out cohesive, enduring, medium-sized separate physical entities (as in the birthday party memory game ‘Name the objects on the tray’, when the tray is only briefly uncovered). The above rhetorical appeals make the philosophical thesis sound bad by tacitly appealing to the sortal sense of ‘object’, which is associated with existence and identity conditions that would rule out the existence of a disjoint ‘object’ like the sum of tower and ear, and would rule out the possibility that there be a million objects on my desk. But this is irrelevant to the question of whether or not there are sums or gollyswoggles (i.e. whether the application conditions associated with such terms are fulfilled, and thus whether there are such ‘objects’ given a covering use of the term). But suppose, less inflammatorily (and less misleadingly), that we explained to ‘normal’ people how the terms ‘gollyswoggle’ and ‘sum of x and y’ were to be used—such that, for example, the former applies just in case there is a properly shaped piece of clay, and the latter just in case x and y exist. Then simply ask them, for example, is there a gollyswoggle (here on the pedestal)? I think in this case ‘common sense’, with a vocabulary suitably expanded to include the new term, would certainly accept that there is. (And much the same, I think, would go for the case of sums, once the whole language game of mereology was sufficiently introduced.) So while everyday English may not include the relevant vocabulary, and may rightly neglect to have any interest in the referents of such gerrymandered (but referring) terms as we care to introduce, I do not think that its indifference to such terms suggests that it is contrary to common sense (once the terms are introduced) to allow that they refer.4 So I accept, and do not think that common sense denies (or would deny) that there are gollyswoggles, sums, and referents of whatever other terms may be introduced in a way that (unlike ‘hoverball’ or ‘wishdate’) genuinely guarantees that their application conditions are
a coherent common sense view
185
met, provided the truth of other sentences accepted (e.g. there is clay shaped in the following way; there is the Eiffel Tower and my nose). Indeed, wherever we have a sortal with coherent application and coapplication conditions, and the application conditions are fulfilled, we may then, if we use ‘object’ in a covering sense, say that there is an object of that sort. Is this then a ridiculously profligate, bloated ontology? To see that these accusations are inappropriate, one need only return to the discussion of counting and parsimony above. But doesn’t this entail massive amounts of colocation, and thereby violate common sense? The discussion of colocation in chapter 4 has already shown how to handle this objection. Barring any further lines of worry, then, we can conclude that although the approach to ontology I have recommended entails colocation and the existence of many more objects than are naturally mentioned in an inventory of the world, properly understood, these consequences do not undermine its claim to preserve a common sense ontology.
10.4 The Specter of Antirealism One final, recurrent worry is that the foundations I have endorsed above for forming a common sense view preserve the existence of tables, baseballs, and stones only at a much greater hidden cost: abandoning realism. There are at least three ways in which the worry might be put, but in each case, I will argue, the argument that the above position leads to antirealism is a red herring. One way of expressing the worry we have already encountered in section 6.4. Since I am committed to denying that there is an answer to the question ‘How many objects are there?’ (where ‘object’ is being used in the supposedly ‘neutral’ sense), the view superficially resembles Putnam’s endorsement of ‘internal realism’ in his claim that there is no answer to the question ‘How many objects are there?’ except internally to a conceptual scheme (which is at least commonly considered a form of antirealism). But, as I have already argued, the point that the question ‘How many objects are there?’ is ill formed, on my view, is simply an observation about the semantics of the term ‘object’ (that, in this supposedly neutral sense, it comes without any coapplication conditions that would enable us to count). As Hilpinen (1996) observes, however, this does not at all interfere with the realist view that, for a great many sortal terms ‘S’ (including those of the natural sciences),
186
ordinary objects
there is a mind-independent, determinate answer to the question ‘How many Ss are there?’ Another attempt to resurrect a case for antirealism in the view I have been developing here might be put like this: On your view, all objects turn out to be mind-dependent, for there would be no objects were it not for our stipulation of the application and coapplication conditions of sortal terms. And the idea that all objects are minddependent is surely a classic statement of idealism. But, as I argued in section 3.3, whether the claim that objects are mind-dependent is utilized in favor of antirealism or in defense of an ontology of ‘mere stuff ’, it is based on a confusion. The point I have been utilizing above is that the conditions relevant to the application (and coapplication) of various terms at least at a basic, frame level are established by speakers (whether individually, as in the case of technical terms, or collectively, as in the case of customary terms). This does not, however, mean—and must not be confused with the idea—that those conditions can only be fulfilled if there are human thoughts, practices, terms, and the like. It is the latter state that marks existential mind-dependence. In the case of most common sense natural terms such as ‘stick’, ‘stone’, and so on, although the application conditions are established by us, according to those very application conditions there could be a stone even in a situation in which there were no humans (for the conditions appealed to for the existence of stones make no reference to humans or representations as a necessary condition for their existence). This is the central contrast between terms for natural objects and terms for social and cultural objects, for although all are alike in requiring that framelevel application conditions are established by human convention, they differ insofar as the latter, but not the former, require for the fulfillment of these conditions that there be human (or similar intelligent) beliefs, practices, representations, and so on. Thus, for example, while (according to the application conditions set up for the respective terms) there may be stones in a world devoid of representations, there may not be money, laws, or governments in such a world. Here I have focused on the role of humans in establishing application conditions; a separate study is needed to investigate the role of human beliefs, practices, and other representations in fulfilling application conditions, and in drawing out the resulting important differences between the referents of our terms for natural versus social and cultural objects (see Thomasson 2003b).5 While I think there are important differences between objects that do and do not depend existentially on human intentionality, I also think (and have argued elsewhere, e.g.
a coherent common sense view
187
2001a, 2003a) that the dependencies of social and cultural objects on human intentionality do not show that these objects don’t ‘really’ exist, or that we must be in some sense ‘idealist’ about them; as in other cases, provided the associated application and coapplication conditions are substantive and consistent, there simply are such objects if the application conditions for the relevant terms are fulfilled (even if these conditions appeal to human intentionality). In any case, however, one crucial lesson of this study is that the role of human intentionality in establishing application and coapplication conditions must not be confused with its role (if any) in fulfilling these conditions. If we speak simply of objects being ‘mind-dependent’ without clarifying what we are talking about, we may land ourselves in a great deal of unnecessary trouble, and miss the opportunity to notice a view that, like that developed above, provides a coherent and comprehensive approach to metaphysics that also yields an ontology of ordinary objects.
C c
eleven
the methods of metaphysics
In some ways, the conclusions of this book may seem (to most) unsurprising: the Moorean suspicions mentioned at the outset proved justified, for there does turn out to be something wrong with each of the arguments against the entrenched common sense position that there are ordinary objects. As a result, we lack sufficient reason to abandon an ontology of ordinary objects and may conclude where most of us started: accepting that there are tables and chairs, sticks and stones. In a sense, though, this is not the most important result of the work of this book. And although the Moorean suspicions were on target, the time and words expended trying to determine where precisely the eliminativist arguments go wrong (rather than just dismissing them on grounds that something must have gone wrong somewhere) prove to be well spent. For our investigations have shown not only where the opposing arguments go wrong but also how to develop a reflective common sense worldview that is subject to none of the problems alleged to plague it. Perhaps more important still, the cluster of theses that underpin the reflective common sense worldview have significant consequences for our understanding of the proper methods and limits of metaphysics. The central questions of metaphysics include questions about identity and persistence conditions (e.g. of persons, artifacts, or animals), about the ontological nature of things of different sorts (e.g.: Are works of music concrete or abstract, created or discovered?), and about existence 188
the methods of metaphysics
189
(What is there? Are there numbers, fictional characters, mereological sums, temporal parts?). There is a common (if recent) tendency to view metaphysics on analogy with science, as engaged in searching for answers to questions about the identity, persistence conditions, or ‘nature’ of objects, or to more global questions about what exists, that may turn out to be genuine discoveries that overturn our prior common sense views. Similarly, metaphysical questions about, for example, the precise conditions under which a person can survive, a work of music can be performed, or a stone may exist are often treated as if they must have precise discoverable answers, just as much as questions about the chemical composition of a sample of gold must. And ontological theories are generally treated (like scientific theories) as competitors to be weighed up at least in part in terms of their relative parsimony, explanatory power, and so on. But the work above gives us reason to suspect that this way of thinking about what we are doing in metaphysics leads us astray, for the metaphysical side of each of these questions is to be addressed by a form of conceptual analysis whose proper methods and limits are very different from those of the empirical sciences.1 If that is so, I will argue, radically revisionary answers to any of these questions must be met with suspicion. Moreover, I will argue, if we properly understand what we are doing in asking questions about existence, we must suspect that many apparent debates in ontology are pseudodebates based on the disputants talking past each other or trying to answer unanswerable questions. I will close by sketching some of these important metaontological consequences that fall out of the work above.
11.1 Identity and Persistence Questions Some of the most enduring specific questions in metaphysics concern the identity and persistence conditions (or ‘nature’) of objects of various kinds, asking, for example, what are the conditions for personal identity, when can a ship persist, or what is the ‘nature’ of the work of art—where this is a matter of asking if works of art should be understood as concrete artifacts, action types, abstract objects, or entities of some other sort. I have argued above, however, that the most basic, frame-level identity and persistence conditions for the objects we refer to are fixed analytically in fixing the reference of the very terms we must use in asking the metaphysical question (‘When does a person survive?’ ‘Is a work of art an artifact?’ . . .).
190
ordinary objects
As a result, answering questions about the most basic identity and persistence conditions for entities of various kinds must take off from a kind of conceptual analysis, analyzing the proper use of the relevant terms, their application and coapplication conditions. While there may be other factors entering into identity and persistence conditions (e.g.: What changes in temperature can a rock survive?) these are empirical questions for the natural sciences, not metaphysical questions proper. The properly metaphysical side of answering identity and persistence questions simply involves uncovering identity and persistence conditions by way of analyzing the categorial concepts of those who competently use the terms. Although the main target in this book has been theories that eliminate ordinary objects, an equally popular move is to attempt to reduce problematic ordinary objects such as artifacts to entities of more tractable sorts, such as lumps of physical stuff, collections of atoms, and the like.2 Even among those without reductive ambitions, there has been increasing trade in revisionary metaphysical theories of certain everyday objects, for example, theories that paintings are not individual artifacts but rather action types (Currie 1989) or the performances artists engage in during their creative activities (Davies 2004). It follows from the view of identity claims defended in chapter 3, however, that reducing ordinary objects to, or identifying them with, entities of other sorts— where these have different frame-level identity conditions, and thus are of different categories—is a nonstarter.3 For wherever the frame-level identity conditions associated with the relevant sortal terms differ, the individuals picked out by singular terms associated with those sortals are of different categories and cannot be identical, and so identifying statues with action types or baseballs with collections of atoms is not an option. Similarly, we have reason to be suspicious of radically revisionary theories about identity or persistence conditions (see my 2004b, 2005). For if the most fundamental metaphysical questions about identity and persistence must be addressed by way of analyzing the frame-level conceptions of those who ground (and reground) the reference of the terms, then revisionary theories that would take everyone to be massively mistaken about the identity and persistence conditions of things of a certain kind (or other revisionary theories about what things ‘are’ that have these consequences) are wrongheaded. For at least the general, frame-level conditions of identity for things we refer to are fixed by those who ground and reground the reference of the term so that, if the term refers at all, what it refers to has these identity conditions. These, then,
the methods of metaphysics
191
cannot turn out to be wrong: given, for example, the customary use of the term ‘painting’ and associated painting names (‘‘The Mona Lisa,’’ ‘‘Broadway Boogie-Woogie,’’ ‘‘Guernica’’), it just could not turn out (as Gregory Currie, 1989, alleges) that these names in fact refer to types of action that can survive the burning of any canvas. Nor can it be true that (as Mark Sagoff, 1978, argues) even replacing minor damaged parts results in the destruction of a painting or sculpture—not its restoration. If our painting names refer at all, they must refer to things with the basic, framelevel identity conditions competent users of the painting names would associate with them, and these seem to allow for survival through minor restoration, but not through a complete destruction of the canvas. In general, on this view, answering questions about the identity conditions for entities of various kinds is not a matter of looking deep into the world, but rather must be based on a form of conceptual analysis. Questions about the ‘ontological status’ of various sorts of entity (e.g. of works of music, paintings, or laws of state) are similarly best understood as enquiring after whether or not the entities in question may be identified with, for example, abstract sound structures, pigmented canvases, propositions, and so on. Whether or not this is the case depends on whether the frame-level identity conditions determined by the coapplication conditions for the terms for symphonies, paintings, and laws are the same as those for the terms for sound structures, pigmented canvases, or propositions. Thus these, too, must be resolved by analyzing the categorial concepts associated with each term and establishing their compatibility or incompatibility. So one important consequence of the views I have argued for above is that radically revisionary views on any of these topics should be met with suspicion. Metaphysicians who propose surprising answers to questions about, for example, the identity conditions for persons or the ontological status of works of art often claim that their conflicts with common sense are nonproblematic on grounds that metaphysics, like science, may make ‘discoveries’ that are surprising and radical, and may overturn common sense views about these things (Currie 1989, 87). But if these questions must, at least at the ground level, be undertaken via conceptual analysis of the frame-level conditions of those who ground the reference of the relevant term and thereby establish what sorts of conditions are relevant, then any views that are wildly revisionary of these common sense frame-level conditions will be unacceptable, and revisionary proposals cannot be defended by analogy with surprising empirical discoveries of the natural sciences.4
192
ordinary objects
Similarly, if the work above is correct, it may not be the case that all metaphysical questions have precise answers awaiting our discovery. For once we see the role of human intentionality in establishing the relevant conditions, we can see why some specific questions about existence and identity may be unanswerable, whether these concern existence conditions for social facts about elections or runs in baseball, survival conditions for fictional characters, identity conditions for ships, and so on (see my 2003a, 2005). As a result, we should neither waste our time arguing over them nor consider it an objection to any theory that it fails to provide answers (or nonarbitrary answers) to these questions. We can at best propose decisions about how we might use these terms more precisely, not present discoveries about what the facts of the matter really are.
11.2 Existence Questions In chapters 2 and 6 I also argued that existence questions must be addressed by first determining the frame-level application conditions associated with the relevant terms, and then determining whether these are fulfilled. The philosophical side to answering these questions, too, then involves a kind of conceptual analysis, determining what it would take (according to the frame-level application conditions for the sortal) for there to be entities of the relevant kind, be they artifacts, fictional characters, numbers, or persons. The rest is the empirical task of discovering whether or not these conditions are in fact fulfilled (although it may be overwhelmingly obvious that they are). Philosophers may undertake conceptual analysis to uncover the frame-level application conditions that must be fulfilled for there to be the relevant thing (or things of the relevant kind), and perhaps to unravel how these existence conditions relate to those for other sorts of thing, so that we can reveal more precisely what our commitments already are, and formulate a consistent theory. Determining whether the conditions so revealed are fulfilled or not—and thus whether or not there are such entities—is, however, an empirical matter, not to be resolved by metaphysical debates using criteria such as which metaphysical theory is more parsimonious or has greater explanatory power. This should lead us to reevaluate debates wherever parsimony is invoked as a criterion for choosing among metaphysical (as opposed to scientific) theories—including debates about fictional characters, mathematical objects, the theoretic entities of the
the methods of metaphysics
193
natural sciences, and so on. It should also lead us to worry, in cases in which apparently familiar entities are denied existence, that the eliminativist is tacitly inflating the conditions it would take for there to be such things, rather than acknowledging that the basic existence conditions are established with the frame-level application conditions of the term. There is likely to be some resistance to treating existence questions as purely descriptive questions to be answered by way of determining what application conditions are criterially associated with the sortal term in question, and then determining whether those are fulfilled—a method that entails that even simple pleonastic transformations can reveal ontological commitments. For this might seem to some to give us entities too cheaply; some might ask: Why should we think that fulfilling the associated application conditions for ‘K’ is really sufficient for Ks to exist? There is a sense in which I don’t know how any substantive answer to this general question could be provided, any more than it can be provided for questions like ‘Is being an unmarried man really all it takes to be a bachelor?’ There seems to be nothing to say, since this seems to be just a report of how our ordinary concept of ‘existence’ (or better, our use of phrases like ‘There are . . .’) works. Nonetheless, deeper issues lie in the background that I want to address more directly in hopes of dissipating the lingering feeling that ontological commitment is just coming too cheaply here. The feeling that lies behind most calls to parsimony in metaphysics, I think, stems from an impulse to pursue a certain kind of ‘deep’ ontology, or, as Frank Jackson (1998, 1–5) has called it, ‘‘serious metaphysics,’’ that will perform something like the simplifying and explanatory function played by ‘serious’ science. Thus, for example, Jackson writes: ‘‘Metaphysicians seek a comprehensive account of some subject-matter— the mind, the semantic, or most ambitiously, everything—in terms of a limited number of more or less basic notions’’ (1998, 4). This picture of metaphysics also plays a prominent role in many of the arguments for eliminativism surveyed above, as eliminativists claim to provide a simpler basic account in terms of which to explain the phenomena at hand than the realist can. Such a drive for a simple basic account tends to go hand in hand with higher standards for saying that entities of a certain sort really exist—say, that they play an essential role in explanation, must figure in any complete causal story, or exist according to some uniform and nonarbitrary principle of composition. And those individuals who propose or presuppose such higher
194
ordinary objects
standards are, I suspect, precisely those who will find my own approach to existence questions far too liberal. While I think we must use caution about assimilating metaphysical (and other philosophical) enquiries too quickly to scientific ones, and in particular use caution about the sense of ‘explanation’ that metaphysical as opposed to scientific theories can provide, I have no quarrel with the ‘deep’ approach to ontology, properly understood. That is, it seems to me that there certainly can be a role for distinguishing entities that are in some sense (which of course needs to be carefully specified) basic from those that are higher order or derivative, and for offering some kind of account of how the latter relate to the former. And indeed elsewhere (e.g. 1999, 2003b) I have worked on some corners of this project, trying to show how fictional characters, social objects, and the like relate to the more basic physical objects, intentions, and human practices on which they depend. But identifying entities that are in some sense ‘basic’ gives us no reason to deny the existence of those that are not. The parallel move is certainly not made by scientists—accounts (in more basic terms) of how light bulbs work are not typically taken to show that there are no light bulbs, nor are accounts of the basic physical ground of solidity properly taken to show that really nothing is solid ( Jackson 1998, 3)—nor should they be so taken, if the work of chapter 8 is correct. Moreover, as I have argued above, adding additional criteria for tables, rocks, and the like to ‘‘really’’ exist (such as, for example, contributing unique causal powers or being indispensable to explanations) only twists ordinary speech, or rather replaces the normal criteria for there being such things with newly inflated criteria, so that failure to satisfy those criteria does not in the least tell against there being tables and rocks in the ordinary sense (when the relevant terms are being used with their standard application conditions). So while we can acknowledge a legitimate role for ‘deep’ ontology, the ‘deep’ ontological task of determining what the basic entities are and how they relate to the others must not be confused with the merely descriptive ontological task of answering existence questions by saying what things, and what sorts of things, there are. As I have argued in chapter 6, the latter task (if it is to be done straightforwardly, using words with their customary meanings) is to be approached by simply asking what the relevant frame-level criteria are, and whether they are fulfilled. In short, to move from saying that certain entities are not basic to saying that there are no such things, that they aren’t really real or the
the methods of metaphysics
195
like, merely courts confusion and does not relieve us of the ordinary commitment to ordinary objects, in the ordinary sense.
11.3 Genuine Debates and Merely Apparent Debates Adopting this understanding of existence questions potentially has wide implications for helping distinguish which apparent ‘debates’ in metaphysics are genuine and which are pseudodebates. Specific existence questions about the existence of artifacts, mereological sums, numbers, temporal parts, or fictional characters may involve genuine philosophical debates about how we should understand the application conditions for the associated terms, and thus about what it would take for there to be, say, fictional characters, numbers, or artifacts—and resolving these debates may be no trivial task, since the relevant conditions may be complex, confusing, and merely tacitly built into the way the term is used rather than explicitly accepted. In any case, however, these debates must be based on analyzing the criteria tacitly built into the ordinary use of these terms. (Of course there seems to be no ordinary—as opposed to stipulated philosophical—use of such technical terms as ‘mereological sum’ or ‘temporal part’. Nonetheless, debates of this sort may arise insofar as the stipulated conditions may appeal to other ordinary terms.) Once such criteria are agreed on, however, it an empirical matter whether or not they are fulfilled; if they are, the realist wins. While it is at least in principle possible that there be empirical debates about whether or not the relevant conditions are actually fulfilled, this is not a philosophical controversy (and in the cases at issue, there is generally no controversy about the relevant matters of empirical fact)—the philosopher’s share of the work lies in conceptual analysis. More typically, however, philosophical debates about existence involve more general questions about what ‘things’ or ‘objects’ there are—either directly by asking for an inventory of what ‘things’ there are, or indirectly by taking an answer to a question like ‘Is there any thing composed here?’ as grounds for an answer to a specific existence question (e.g. ‘Is there a chair here?’). Philosophical debates about existence framed this way may be interpreted in several ways. But, as I have argued, none of these seems suited to revive the idea that there are serious, philosophically resolvable, debates about what exists rather than
196
ordinary objects
miscommunications, shallow debates about meanings, or pseudodebates involving incomplete statements and unanswerable questions. As I have argued in section 6.1, if serious ontologists are using ‘object’ or ‘thing’ as sortals, so that they have certain application conditions and identity conditions explicitly or tacitly in mind, then the existence claims used in making their arguments may be straightforwardly truthevaluable. But this way of making existence questions answerable comes at the risk of making disputes about whether there is or is not a thing in some situation entirely shallow. For then differences of opinion about whether or not there is any thing composed in a situation with particles arranged tablewise seem to be based on the participants talking past each other by turning the term ‘object’ or ‘thing’ into a sortal in different ways; or, at best, into shallow debates about how best to turn it into a sortal, or what conditions are most standardly associated with it (see Sidelle, 2002, 141–2, and Hirsch, 2002b, 106).5 Disputants may, rather than using ‘thing’ as a sortal of its own, be using it as a covering term that applies provided some first-order sortal applies. But then differences in their claims about what ‘things’ exist must resolve into differences in their answers to one or more specific existence questions. But in that case, as I have argued above, the different judgments at bottom reflect differences in what application or identity conditions they associate with one or more of these sortal terms (or differences in which specific first-order sortals they consider), not serious differences in beliefs about what exists. The third way to attempt to understand ontological claims of existence is that which treats ‘thing’ purely neutrally (not as a sortal of its own or a covering term guaranteed to apply provided some other sortal applies). But if we understand ‘thing’ in that way, as I have argued, generic claims about what ‘things’ there are are incomplete and not truth-evaluable, and cannot genuinely conflict with any different answer to the question ‘What things are there?’ One way to show that a certain debate is merely apparent would be to show that (although the disputants may be speaking different idiolects) the claims of one may be ‘translated’ into the other’s language, and result in claims the second would accept, so that (with translation in place) the two would clearly be in agreement. Sider (2006) examines this move in the case of skeptical arguments to the effect that disagreements about (temporal) ontology are merely verbal, and argues that no translation function can be found to translate the claims of the disputants into terms both would accept (showing that they really are saying ‘the same thing’ in a different idiom).
the methods of metaphysics
197
But the arguments above do not presuppose that there is some translation function available from one ontologist’s claims into terms the other would accept. While it seems plausible that demonstrating intertranslability of the disputant’s claims would be sufficient to support skepticism, it hardly seems necessary. For two people may fail to disagree in many different ways, for example, if at least one is failing to make any complete and meaningful statement at all, or if they are simply talking about different things entirely— for example, if Bob says that Mary is at the bank when John says she is not, when the first means the financial institution, and the second the edge of a river. People may be talking past each other (and not engaging in any genuine debate) even if there is no translation function from one language to the other—imagine that John’s language does not include terms for any social or institutional kinds and suppose, as seems plausible, that there are no translations from mere natural terms to institutional terms. But while it is clearly unnecessary for lack of disagreement for there to be a way of showing that the disputants’ claims may be translated into shared terms (to which they would agree), to show that there is a genuine disagreement (rather than just people talking past each other), clearly it is necessary to be able to put the disputants’ claims in the same terms and show that they make conflicting assertions. Serious ontologists themselves seem to recognize this demand, for they often are at pains to show direct conflicts among their apparently competing theories. But most attempts to point to a direct conflict involve apparent differences in views about how many ‘things’ there are or what ‘objects’ are in a certain situation. Thus, for example, in attempting to specify the difference in views between those who do and do not accept statues, Merricks writes: ‘‘Suppose there are a million atoms arranged statuewise in a certain region of space. And suppose we ask how many (nonsubatomic) things there are in that region. My metaphysical opponents would say that there are at least one million and one (the atoms and the statue). I would say there are only one million’’ (2000, 48). Sider (2001a, 203; see also 2001b, xx) similarly describes the differences among the nihilist, the chaste endurantist, and the defender of temporal parts in terms of whether or not each would accept as a description of a situation with exactly two electrons such claims as ‘there are at least two/three/four things,’ and van Inwagen explicates the difference between his denial of composite inanimate material objects and the ontology of those who accept them by insisting that, on his view (by contrast with the latter) ‘‘there is no one thing that just exactly fills [the]
198
ordinary objects
region of space [said by realists to be occupied by such a composite]’’ (1990, 104). Given what has come before, the fact that those looking for conflicts almost always do so by appealing to differences in how many ‘things’ there are in a given situation should make us immediately suspicious. For if the results above are correct, conflicts symptomatic of genuine debates cannot easily be found this way. Different metaphysicians might provide different but nonconflicting answers to generic existence and counting questions like ‘How many objects are there?’ in several ways. First, the disputants may be turning ‘object’ or ‘thing’ into a sortal in different ways, preventing them from talking about (and disagreeing about) the same ‘things’. Second, they may be using ‘object’ or ‘thing’ as a covering term, but differ in which specific firstorder terms they consider, or what application or coapplication conditions they associate with one or more of these. Third, the two may be engaging in different types of use of ‘object’ or ‘thing’, with one using it as a sortal and the other as a covering term. Finally, at least one may be using ‘thing’ purely neutrally in a way that makes no complete claim that could genuinely conflict with any other answer. With all these ways of missing each other, it’s no wonder that apparent disputes are so rampant, and it would be a wonder if the disputants could actually meet up to start a fight. Serious ontologists themselves, of course, will not take kindly to this diagnosis—but even if they do not accept it, the attempted diagnosis can at least help focus the metaontological debate on the central questions of how to understand generic existence claims, and their relation to specific existence claims. A more neutral audience (less committed in advance to the view that such debates are genuine) might note that the view of existence questions motivated in chapter 6 (based on the work of chapters 2 and 3) has some additional attractions. Not only can it help provide a unified diagnosis of the problems behind a variety of arguments against ordinary objects and yield the basis for a coherent common sense ontology, it also has the virtue of dissolving a great many apparent disputes about ontology. This may be a significant advantage at least for those who feel themselves unmoored at the prospect of adjudicating debates about what ‘things’ there are (e.g. about whether organisms, artifacts, universals, mereological sums, the blobject, etc. really exist)—where this is not supposed to be a simple matter of empirical discovery of whether or not the application conditions associated with the terms are fulfilled. It also can explain why there seem to be so vastly many different ontological proposals, with no prospect for
the methods of metaphysics
199
converging on agreement. Finally, and perhaps most important, it relieves us of the epistemological embarrassments that come with a ‘serious metaphysical’ approach that takes facts about what exists and what modal features objects possess to be discoverable by some special means that is not simply exhausted by a combination of conceptual analysis and straightforward empirical enquiry—as they are on the model I have been defending.6 From a neutral perspective, these are advantages not to be lightly overlooked.
11.4 What’s an Ontologist to Do? Given the above method of handling identity and existence questions, how can we understand what ontologists are up to in their apparent disagreements about these questions, and what sorts of question are left for ontology to investigate? On the above proposal, fundamental existence questions must be understood as specific existence questions (involving category-specifying terms), though one can invoke a covering use of ‘thing’ or ‘object’ for more general answers built on these. In approaching such existence questions, as with approaching identity questions, a number of potential areas of investigation and disagreement remain, along with much work for the metaphysician to do, but the tasks must be understood properly. First, for specific existence questions, there may be disagreements about what category a relevant term is associated with, and so about what application conditions and coapplication conditions for a term are. These may be of either of two sorts, as follows. (1) There may be differences in the conceptual analysis of what the term, as ordinarily used in English, means, and so about what it would take for there to be, for example, tables, numbers, works of art, or fictional characters, or the conditions under which two artifacts, characters, or persons are identical. These are factual debates about the actual meaning or use of an English term, but they may be quite difficult to resolve, since the presupposed application conditions are generally merely tacitly built into the way the term is used, and they may be complex, somewhat arbitrary, or involve various only loosely related threads of usage. In fact, the conditions of application and coapplication associated with our ordinary terms may even be (and perhaps typically are) incomplete, so that they do not unequivocally decide every case, or even (in rarer cases) are inconsistent—either internally or as combined with application of related terms. As a result, such debates may easily
200
ordinary objects
grade over to or be replaced by (2) debates implicitly not about what a given term actually means as used in everyday English, but rather about what the best way would be to revise our terms’ meanings, to make them more precise, more clear or consistent in application, more cleanly linked to other terms’ meanings, and so on. These are fundamentally pragmatic debates about what new meanings would be best to adopt (given certain other goals), akin to debates about how the rules of NCAA basketball should be changed to improve the game (make it more efficient, exciting, fair, etc.). But ontologists engaged in this project must be clear that they are not discovering surprising facts about, for example, where composition really does or does not occur, but rather proposing revisions to our somewhat messy conceptual scheme. Projects 1 and 2 of course may grade into each other, and though there are clear cases at each end, perhaps most debates about the existence or identity conditions of things like persons, artifacts, and fictional characters can be seen as both explicating and to some extent fleshing out and regimenting the application and coapplication conditions of the associated terms. But while both of these may be legitimate, difficult, and interesting subjects of discussion, properly understood, both of these are shallow debates about meanings, not deep debates about what really exists in the world, or what it is like. Once the conceptual analysis is done, and the application and coapplication conditions associated with a term agreed on, the only further room for debate concerns whether or not (or how many times) the relevant criteria are fulfilled in a given world-situation. But while there may be genuine disputes about these issues (Are there four or five oaks in our yard? How many trout remain in the lake?), these disputes are to be resolved by ordinary investigative techniques—whether those of simply going and looking, or in more complex cases, those of the investigative journalist or research scientist. If the above is correct, then ontologists cannot rightly conceive of themselves as engaged in serious debates about deep issues regarding what things, and sorts of things, really exist in the world, and so cannot conceive of their inquiries as analogous to those of science in ‘limning the depths of reality’. Specific philosophical debates about existence must be seen as either genuine but shallow debates about the application conditions the terms in question have (or should have), or as mere pseudodebates about what exists, where the apparent conflicts between the views are based simply in associating different application criteria with the term in question.7
the methods of metaphysics
201
As a result, if the above is correct, the primary role of the ontologist addressing existence questions is to undertake a certain kind of conceptual analysis, not to engage in deep discoveries about what really exists, or what things there really are. Once that is acknowledged, we can avoid being misled into entering pseudodebates or trying to answer ill-formed questions, or mistaking proposed conceptual revisions for surprising discoveries about the world itself. In short, one of the most important results of the above investigations lies in clarifying the proper role and methods of addressing questions of existence and nature in metaphysics, and distinguishing which issues (or aspects of issues) are empirical, and which conceptual. Once the difference between metaphysical and scientific enquiries is made more evident, we can avoid being taken in by revisionary theories that claim to overturn common sense, avoid fruitlessly pursuing questions that have no answer, and focus our efforts explicitly on the difficult task of conceptual analysis. Thus, even for those convinced in advance of the existence of ordinary objects, taking eliminativist arguments seriously and getting to the root of the problems with them turns out to be a worthwhile enterprise. For the payoff may lie not only in a reflective common sense view that can clearly avoid the problems posed by its critics but also in reexamining the way we think about and practice metaphysics.
This page intentionally left blank
notes
Introduction 1. Of course it has also become common to deny the existence of ordinary animate objects. Many of the arguments I will consider have also been wielded against the existence of animals, some even against persons (e.g. Horgan (1993), Horgan and Potrcˇ (2000), Unger (1979b, 1979c)). My responses to these arguments often apply equally whether the points in question are used against inanimate or animate objects, but to keep things simple I will focus on the case of inanimate objects.
Chapter 1 1. Neither van Inwagen nor Merricks, however, thinks this form of argument applies to humans, since they agree that thought or the possession of mental properties cannot simply be accounted for by the cooperative activities of the relevant simples (see Merricks 2001, 88–89; van Inwagen 1990, 122). 2. Originally expressed in Plato’s Sophist (247e) and more recently by Samuel Alexander (1927, 8). Jaegwon Kim has recently popularized the principle under the name ‘‘Alexander’s Dictum’’ (1993, 348). 3. At least, provided ‘the mob’ is used as a singular expression rather than as a plural referring expression. 4. Although see E. J. Lowe (2003a) for an interesting way of calling into question the assumption that the atoms can be said to (collectively) cause whatever one wanted to say the baseball causes. 5. In fact, though, the reason this seems wrong may be—at least in part— the same sort of reason overdetermination claims seem wrong (see below): 203
204
notes to pages 13–17
claiming that the baseball, ‘alongside’ its atoms, is a partial cause of the shattering seems to presuppose that the atoms (arranged baseballwise) and the baseball are separate, independent entities. One cannot be alongside oneself, nor can a whole be alongside its parts. Nonetheless, there may be an additional problem with the claim that each is a partial cause, since that more explicitly requires that the baseball provide (e.g.) some causal work beyond what the atoms arranged baseballwise contribute. 6. Ryle’s claim that a category mistake is involved is not itself an explanation of the absurdity, since he defines ‘categories’ in terms of absurdity, treating two expressions as of different categories if there are sentences in which substituting one for the other results in absurdity (1938/1971, 181). 7. Thanks to David Barnett for suggesting the Gricean analysis. 8. Theodore Sider (2003b) similarly points out that the claim of overdetermination by the baseball and its properly arranged atoms does not parallel cases in which claims of overdetermination (e.g. of a death by two shots) would involve positing unexplained coincidences, and uses this as part of an argument that overdetermination of the former sort is not shown to be sufficiently worrying to justify rejecting ordinary objects. 9. Yablo (1992) uses the nonindependence claim to argue that determinables and determinates are not causal rivals, and goes on (more controversially) to argue that mental properties are determinables of physical property determinates, and so not in causal rivalry with them. 10. One such attempt (which doesn’t involve appeal to the idea of analytic entailments) might be to deny that there is causal redundancy in cases where one entity materially constitutes the other (or the same thing materially constitutes both of the alleged rivals). 11. Cases of purely logical entailments—considered as entailments just in virtue of rules of inference, knowable by competent speakers merely on the basis of their reasoning abilities—would also count trivially as analytic entailments (in which the meanings involved are making no essential contribution to the entailment). Nonetheless, it is substantive (not merely logical) analytic entailments that will be at issue here. 12. Note that this involves denying the view that if two claims can be made true by the same truth-makers, then they express the same proposition. Noting the (often one-way) analytic entailments between our claims gives us reason to deny this: ‘John bought a house’ does not express the same proposition as ‘John bought a building’ (and ‘house’ is not synonymous with ‘building’), although one and the same occurrence ( John’s purchase of 2729 West Lake Avenue) may make both true. For a somewhat different discussion of reasons to deny this relation between truth-makers and propositions, see Heil (2003, esp. chap. 7). (Thanks to Uriah Kriegel for raising this issue.) 13. In chapter 9 I will return to this issue. There I argue that the existence of statues trivially supervenes on the existence of atoms arranged statuewise on any acceptable way of fleshing out Merricks’s definition so as to
notes to pages 18–27
205
preserve the sense in which ordinary claims about statues are ‘nearly as good as true’. 14. The analytic entailments at least seem straightforward for claims of causal relevance. Claims to be ‘the cause’ may not be so straightforward, as these are often supposed to narrowly describe what made the crucial difference between the effect’s occurring and its not (Yablo 1992, 188). As some of Yablo’s (1992) examples show, while ‘x is red’ may analytically entail ‘x is colored,’ and the causal relevance of a triangle’s redness (e.g. to a pigeon’s pecking) may not preclude the causal relevance of its being colored, the coloration still may not have as good a claim to being the cause, if the bird would not have pecked at a yellow square. 15. A claim Lowe (2003a) argues against, thus providing a direct defense against the argument from parsimony. 16. Put in the terms that will be laid out in chapter 2, this cannot be done without violating the basic application and coapplication conditions associated criterially with the use of the term ‘baseball’. So shifting terms of discussion to ‘lumps of matter’ (which comes with different application and co-application conditions) would be changing the subject away from talk of baseballs altogether. 17. Or, as I will argue in chapter 4, to claims of location or property possession. 18. The set of claims F must be a minimal set, since otherwise, one could add in any irrelevant (but consistent) claims to a set F that is already analytically sufficient for c, and F would still analytically entail c, but there would be no assurance that the superfluous claims would have no rivalry with c. For example, the set of claims F—Jones bought a left-hand glove, Jones bought a right-hand glove, those gloves match, and Jones bought a scarf—still analytically entails that Jones bought a pair of gloves, though the last superfluous member of F may nonetheless be a rival or additive to the claim that Jones bought a pair of gloves. 19. As David Chalmers would put it, facts about baseballs are logically supervenient on facts about atoms arranged baseballwise, while facts about mental states seem not to be logically supervenient on facts about brain states (1996, chapter 2). 20. Nonetheless, accepting such analytic entailments does not mean accepting a reductive view of ordinary objects. See x 11.1. 21. See Chalmers 1996, 72, and Chalmers and Jackson 2001, for similar arguments that any failure of such ordinary objects to supervene on mere physical facts is attributable to their dependence (at some level) on conscious experience. In x 9.7 I return to consider the prospects for a more severe eliminativism about ordinary objects that also eliminates everything mental. 22. I have drawn out one such reply elsewhere (1998). 23. Or that similar restrictions apply not merely when one claim analytically entails another but when there are entailments of a weaker kind. Jackson
206
notes to pages 28–35
(1998, 24–5), for example, argues that a weaker entailment relation—the ‘fixing’ or ‘necessary determination’ entailment—must hold between physical and psychological accounts of the world, provided physicalism is true. Does the causal principle remain valid when applied across sets of claims such that one fixes the other? If not, this could provide another line of defense against causal redundancy arguments about the mental. Chapter 2 1. If one assumes it is talking about ‘real’ overdetermination. See x 1.2. 2. Chalmers and Jackson defend the similar idea that there are a priori entailments, and argue more broadly that all ‘‘ordinary truths about macroscopic phenomena are entailed a priori by the combination of physical truths, phenomenal truths, indexical truths, and a that’s-all statement’’ (2001, 358). 3. See also his ‘‘Carnap and Logical Truth’’ (in 1966/1976). 4. It also may leave the notion of logical truth in need of further explication; in particular, we may need an account of how logical truths are knowable a priori. This, however, is not Quine’s main target in ‘‘Two Dogmas.’’ For an elucidating discussion of how to explain a priori knowledge (of a certain sort) of logical claims, see Boghossian (1997). 5. The Carnapian idea that analytic statements are those true by convention is one Quine earlier seemed to accept (1966/1976, 90). 6. See Carnap’s apparent reply to Quine (1937, 26–9). 7. Of course the idea that rules of use for specific terms may be established individually or collectively, explicitly or tacitly, does not prejudge the issue of whether some abstract features of linguistic syntax, or even some basic semantic categories, may be universal features whose presence in language is biologically based. The crucial point here is merely the obvious observation that competent speakers may, in different ways, introduce novel terms with different (and interrelated) rules of use, establishing analytic interrelations among them. 8. Indeed Strawson and Grice suggest that understanding the former may be a necessary precondition to understanding the latter case: ‘‘the notion of synonymy by explicit convention would be unintelligible if the notion of synonymy by usage were not presupposed. There cannot be law where there is no custom, or rules where there are not practices’’ (1956, 153). 9. Including not only conventionalist views (properly formulated) but also Kantian or transcendentalist views. This is of course not to say that the positions I defend above are consistent with all available views of logical truth. 10. Of course, if minimalists about meanings are right, then there simply is no more to say about meanings than is contained in the platitudes involving such interrelated terms, and the failure to provide an explication that reaches outside that circle cannot be supposed to be a defect. On the contrary,
notes to pages 36–38
207
expecting and demanding such an explanation would just demonstrate a failure to grasp the genuinely minimal status of meanings and such interrelated entities (see Johnston 1988). 11. There is also reason to think that precise conditions could not be offered, since the notion of meaning the same as seems itself to be vague, an issue to which I will return in chapter 5. 12. I will have more to say, however, in chapter 3 about analyticity— specifically, about why we have reason to think that the most basic modal claims are analytic, and the sense(s) in which analytic claims may be said not to need truth-makers. 13. Putnam (2000) similarly argues that one can accept that there are analytic truths while agreeing with Quine that no statements are unrevisable and must be held true ‘‘no matter what.’’ 14. This view of the source of synonymies in natural language (and our potential for having a sort of a priori knowledge of them) is somewhat analogous to the view Boghossian (1997) defends of the source of a priori knowledge of logical claims, as based on a kind of implicit definition. 15. One important naturalistic approach to understanding meaning and reference I will not discuss here: the form of meaning externalism developed by Ruth Millikan (2000) (though there she is concerned primarily with understanding concepts rather than linguistic reference and associated issues of analyticity as such). The view she develops there apparently avoids the qua problem faced by causal theories of reference. For she holds that substance concepts involve two components: a recognitional component and a grasp of a ‘substance template’ (9–10) that disambiguates, e.g. whether one is conceiving of Fido or of the species Dog, or the kind Fur, etc. (77) The species template determines what sorts of meaningful questions can be asked about the substance conceived of, what it’s possible to learn about the substance, and how to generalize what we have learned. The substance templates provide the ontological disambiguation I insist above is required. A much longer digression would be needed to examine the extent to which Millikan’s view is and is not consistent with the picture of reference, identity, and modality I develop above. But it is worth noting that these substance templates play much the same role as the categorial conceptions I discuss above, and may even provide the basis for conceptual connections involving something like my analytic entailments (though Millikan no doubt would not put it in those terms). E.g. in light of the role of substance templates, she notes: ‘‘Tracking Mama is one of the means of tracking women. If it’s Mama again, it’s a woman again’’ (2000, 74). Thus although there are some obvious differences in how the two views understand concepts themselves, this form of externalism shares something crucial in common with the view I defend above: both views acknowledge that the ontological category of entity we think of or refer to is determined by the basic categorial concepts or substance templates we employ.
208
notes to pages 38–41
16. Versions of the qua problem are discussed in Papineau (1979, 158–68), Devitt (1981), Sterelny (1983), Devitt and Sterelny (1999), Dupre´ (1981), Kitcher (1982), and Stanford and Kitcher (2000). Apparently even Locke noted that this sort of problem would plague theories like direct reference theories. See Stanford and Kitcher (100). 17. Alan Sidelle (1992b, 420–1) argues that a ‘permissive’ ontology including entities that overlap spatio-temporally (such as pieces of wood and lawn chairs) threatens the determinacy of reference for causal theorists. Resolving this indeterminacy, he argues, requires that we accept that semantic features must be invoked to relieve this kind of indeterminacy—and thus that reference across possible worlds cannot just be determined by ‘real’ mindindependent cross-world identities. While I agree that such a permissive ontology is sufficient to create a threat for causal theorists, for the reasons given above I doubt that these threats can be overcome even if one defends the idea that there is a ‘privileged’ ontology, e.g. of a single level of basic entities. 18. There may be room in principle for a severe reductionist (rather than eliminativist) view that (combined with a minimal, one-thing-to-a-place ontology) would take all singular terms uttered in the same causal circumstances to refer to the very same thing. This would obviously lead to some highly counterintuitive consequences, for example, identifying cows with lumps of bovine matter, cow surfaces, cow parts, and perhaps even cow-locations, and so on. It also seems to make it completely mysterious how we discover which ontological features (e.g. persistence conditions) the one thing here has (whether those of cows, lumps of matter, etc.). Since this move does not seem immediately plausible, and is in any case not open to the eliminativist, my chief interlocutor, I will not pursue it further here. 19. These conditions may presumably vary in level of detail and specificity. The point here is not to rule out the idea that some terms may come with more substantive application conditions than the frame-level ones mentioned here, but rather to insist that even those inclined to pure causal theories of reference have reason to accept that our singular terms have at least this much conceptual content. 20. A view like this is drawn out in Dummett 1973/1981, 571; Lowe 1989, 20–27; Wright 1983, 2–4. (An early source of this view is expressed in Frege’s requirement that proper names come associated with a criterion of identity, 1884/1968, x 62; see also Dummett, 73–9). Most past authors, however, have spoken of sortals as being associated with criteria of application and ‘criteria of identity’ (rather than ‘criteria of coapplication’). I discuss reasons for my change in terminology in x 3.1. 21. Thus we must emphatically not (as one anonymous referee supposed) take ‘possession of certain identity conditions’ or of the modal properties that go with this as criteria for the application of a sortal term. 22. It is these highly general application and coapplication conditions associated with categories that form the ‘frame-level’ conditions that, I have
notes to pages 42–50
209
argued above, may disambiguate intended reference and help avoid the qua problem. Attempts to ground the reference of a name are only successful provided the categorial term’s application conditions are fulfilled, although in at least some cases we might allow that reference can be secured even if the more specific application conditions of an associated species-sortal are not fulfilled. Thus, for example, grounders intending to name that whale ‘Orky’ may still succeed in grounding reference if the creature before them is a porpoise, since porpoises (unlike clumps of seaweed or waves) are of the same category as whales (animal). 23. This, as we will see in chapter 6, rules out the idea that there is a highest category such as ‘object’ or ‘entity’, if these terms do not come with coapplication conditions. See also my 2004a. 24. The most generic, neutral use of these terms seems not to involve association with such conditions. I discuss these and other uses of these highly general terms in chapter 6. 25. Reference shifts should be understood as a matter of establishing a new term homonymous with the original, but with different associated frame-level conditions and a different referent (if any). If, for example, a confused later speaker uses ‘tenacity’ to refer to the city he has arrived in (thinking that is the local name), and that usage becomes entrenched, a new place-name may be established that only sounds like the (old) term for a character trait. These of course are not the only source of reference shifts, since we may also have a shift of reference, e.g. from one person to another. 26. The picture here is much like the two-dimensional view of concepts developed by David Chalmers and Frank Jackson (2001). Grasp of a concept, they hold, involves ability to evaluate the concept’s extension across various epistemic possibilities, but does not require knowledge of definitions or explicit conceptual analyses. 27. The possible incompleteness of such conditions will play an important role in diagnosing the source of vagueness, in chapter 5. 28. Another common approach is the ‘gappy proposition’ view developed, for example, by Braun (1993), Reimer (2001), and Adams, Fuller, and Stecker (1997). I have argued elsewhere (2003c) that metalinguistic approaches, suitably modified, are preferable to gappy proposition models. 29. This ‘handling’, though, is presented explicitly only as a view of the truth-conditions for these utterances, not of what these statements mean or what propositions they express. 30. See the discussion of different uses of the term ‘bliger’ (or ‘bliger’ and ‘pliger’) in chapter 9. 31. Nor will it do to say that we can avoid this problem if we deny the existence of one or the other, for even the spare ontologist who would accept, for example, that there may be only one thing in a given location at a time does not normally hold that we can discover which thing it is merely by going and looking.
210
notes to pages 51–57
32. Thanks to Mark Sainsbury and Dan Korman for helpful discussion of this point. 33. The idea that our terms have at least a basic conceptual content that plays a role in determining reference is, of course, not new or unfamiliar (though it remains unpopular in certain quarters). For developments of similar ideas, see, e.g., Devitt and Sterelny (1999), Dummett (1973/1981), Sidelle (1989), and the two-dimensional approach to semantics developed in Chalmers and Jackson (2001). Chapter 3 1. This is only an ‘only if’ condition, since, even if associated with a categorial concept, there may be cases in which the associated identity conditions are underspecified or vague in ways that leave the truth-value of the identity claim indeterminate. See chapter 5 for further discussion of vagueness. 2. Dummett (1973/1981, 75–6 and 546) calls ‘categorial’ terms the most general sortal terms subsuming others that share the same frame-level criterion of identity. My use of the term resembles his in requiring terms of the same category to share the same criterion of identity, although I do not count the sharing of a criterion of identity as sufficient for sortals to be of the same category. 3. Some might be inclined to resist the above view of identity on grounds that it resembles the doctrine of relative identity defended most famously by Peter Geach (1962/1980). But, as has often been pointed out, the two views are importantly different (see, e.g., Dummett 1973/1981, 570–78; Lowe 1989, 43– 63; and Wiggins 2001, chap. 1). Geach holds that (1) claims of identity are always relative to the criteria of identity that are associated with what he calls ‘substantival’ terms, so that simple identity claims of the form ‘a is the same as b’ are ill formed and incomplete, until supplemented with some substantival term ‘F’ so that we can make the claim ‘a is the same F as b’. He also maintains that (2) since identity claims may only be made and evaluated relative to some substantival term that supplies the relevant criteria, there may be cases in which ‘a is the same F as b’ is true, while ‘a is the same G as b’ is false. The present view, however, denies both of these theses. On the view defended here, individuals can only be picked out as individuals at all given association with a sortal term (or, more broadly, categorial conception) that supplies criteria of identity. But this does not mean that ‘same as’ taken alone is incomplete, or (in different contexts) has many different meanings—only that there are many different criteria of identity for things of differing categories (see Lowe 1989, 61). And although criteria of identity vary, we retain a unified understanding of identity in any case as ‘a is identical with b if and only if a and b are of the same category and fulfill the identity conditions associated with that category.’
notes to page 58
211
Moreover, while Geach holds that there are cases in which ‘a is the same F as b’ is true, while ‘a is the same G as b’ is false (e.g. a restored statue, David2, may be the same qua statue as the original (David1), though not the same qua lump of marble), the present view denies this. If ‘David1’ and ‘David2’ really are statue names, then (given that the identity criteria for statues permit small gradual changes in the statue’s constitutive matter) we can say simply that David1 is (absolutely) identical to David2. On the other hand, if they are used as lump names (or one is a lump name and the other a statue name), then we can say that David1 is (absolutely) not identical to David2. 4. This connection between application and persistence conditions holds for all categorial terms and for other sortals we may call ‘basic sortals’, but does not hold for all sortals. For terms we may call ‘derivative sortals’, the application conditions for the general term come apart from the persistence conditions for the individual named, allowing that the thing picked out by the name associated with the derivative sortal may continue to exist (as one and the same) even if the term ceases to apply to it. Thus, e.g., ‘redhead’ is a sortal term that may be used to pick out people (‘Call that redhead in the corner ‘‘Sam’’’), but Sam may continue to exist even if the term ‘redhead’ ceases to apply to him. These derivative sortals include sortals that (like ‘redhead’) involve restricting a more basic sortal, and phase sortals (e.g. ‘boy’ or ‘caterpillar’) that are formed by adding additional criteria for that term to apply at a given time to the individual named—so the application criteria for the term go beyond the persistence conditions for the individual named. The persistence conditions for the individual named by a derivative sortal then are not provided directly by the application conditions for that term, but rather by the application and coapplication conditions for an associated basic sortal. (Initial baptisms using derivative sortals may also successfully ground the reference of a name even if the derivative sortal fails to apply, provided the related categorial term applies.) While it is important to note the difference between derivative and basic sortals in order to avoid various difficulties and confusions, unless otherwise noted, I will be speaking of basic sortals whenever I use the term ‘sortal’ below. (Dummett, 1973/1981, 76, uses the term ‘sortal’ only to cover what I here call ‘basic sortals.’ Wiggins, 2001, 30–3, uses ‘sortal’ more broadly as I do, and uses the term ‘phase sortal’ roughly as I use ‘derivative sortal’.) 5. Sidelle (1992a) likewise argues that knowledge of identity conditions is a priori, not empirical, and is based on grasp of semantic structures that govern proper use of the terms. 6. I noted in x 2.6 that, in some cases, our categorial intentions may be conditionalized on ways the world turns out to be (i.e. on what application conditions are fulfilled). If, e.g., we intend our term to refer to an animal if the application conditions for ‘animal’ are fulfilled, and to ‘artifact’ if they are not, but those for an artifact are, then conceptual analysis can only give us information about the identity and persistence conditions for the name conditional on which application conditions turn out to be fulfilled. Nonetheless, even in
212
notes to pages 58–64
these cases, the identity conditions themselves are not discovered in the world; what is discovered is which application conditions are fulfilled and thus which of the identity conditions (conjoined with which categorial intention) are brought into play. 7. For a historical discussion of the methods of conceptual analysis in phenomenology and ordinary language philosophy, see my forthcoming c. 8. To borrow a phrase from Ayer (1952, 79). 9. See Sidelle (1989, 75–6). 10. See Hirsch (2002a, 107–8) for a similar reply. 11. This is the name Sidelle (1989) gives his position, and it is also used by Elder (2004) in criticizing the view. 12. As Sidelle (1989, 2) puts it, that ‘‘necessity is nothing beyond analyticity.’’ Later, he takes the core view to involve the idea that ‘‘general principles of individuation are analytic’’ (35, see 46). Rea (2002, 85) describes Sidelle’s view similarly, as the view that ‘‘some of our [modal property] beliefs are analytic and others are derived from the conjunction of an analytic truth with an empirical truth.’’ 13. Sidelle (1989, 1–2) also describes it as the view that ‘‘modality does not find its home in the mind-independent world, but rather in us, in our ways of speaking and thinking.’’ 14. Rea (2002, 85– 6). Crawford Elder (2004, 10) similarly describes Sidelle’s view as entailing that ‘‘there are, in the world as it exists independently of us, no modally qualified states of affairs.’’ 15. This change in terminology also has the virtue that it doesn’t seem to imply that all of the concepts associated with our basic categories are purely conventional rather than, e.g., natural for members of our biological kind. 16. Elder (2004, 9–10) presents these as the two exhaustive ways of developing the modal conventionalist view. 17. Similarly, Sidelle writes (1989, 57) that it is a consequence of his view that ‘‘if what it is to be an individual of a certain sort is to have certain features not only actually, but essentially, then the conventionalist has all the same reason to think that if there are any such individuals, they must also not be ‘fully independent’, but should arise out of our individuative practices, which is our way of articulating the world.’’ 18. Rea (2002, 85–96) similarly takes Sidelle’s view to entail antirealism about material objects, though he acknowledges that there are other ways (such as Lewis’s realist modal plenitude) of accepting that modal truths may be known through reflection on our concepts without being committed to objectual antirealism. 19. Sidelle (personal correspondence) apparently does not make this move himself, holding instead that theses 2 and 3 are motivated independently by their capacity to avoid the difficulties of modal realism while accounting for the modal data. The crucial point here, however, is not exegesis of the Sidelle/ Rea/Elder debate, but rather to show that accepting thesis 1 need not (as Rea
notes to pages 64–76
213
and Elder do seem to think) lead to any untoward consequences. As I will suggest in x 3.5, it will also gain us some of the desired advantages over serious modal realism without the costs of the objectual antirealism that thesis 3 (and some interpretations of thesis 2) lead to. 20. See Rea (2002, 85–6), citing Sidelle (1989, chaps. 2 and 3) as evidence. 21. Of course, if there are worlds with different laws of nature in which rocks have a much higher melting temperature, Rocky might survive the heating at those worlds. This also demonstrates the different forms such principles of individuation take: it is essential to rocks’ survival at a given world w that the conditions at w don’t result in liquification; it is essential to water’s existence at any world that its chemical structure at this (actual) world be preserved. 22. And Sidelle (1989, 5) clearly is committed to this view. 23. It may be that such claims about the existence of ‘substantive modal properties’ can’t even be given positive content without averting to mindindependence. Certainly I’m not sure what the remaining content is supposed to amount to, given that even on the minimal conceptualist position there are modal claims that are mind-independently true—what more (it might be asked) does the robust modal realist have in mind? The conceptualist, however, does not need to make sense of the opposing view, but may simply note that her view certainly provides a way of avoiding commitment to it, and can provide an account of the truth of modal claims in far more minimalist fashion. 24. John Heil defends a similar view, writing: ‘‘Talk of modal properties is a philosophically pretentious, and potentially confusing, way of describing constraints built into the concepts we deploy’’ (2003, 186). 25. In keeping with the approach to existence questions defended in chapter 6 and used in arguing for ordinary objects in chapter 9. Chapter 4 1. An associated worry is whether accepting ‘both’ entities would be unduly profligate. I discuss the objection based on concerns of parsimony in chapter 9; here I will focus just on the idea that positing, e.g. a statue ‘in addition to’ the lump is contrary to common sense. 2. While the spatial and mereological senses of colocation may not be equivalent (it may be thinkable to have entities with the same location but different parts), mereological colocation at least seems to entail spatial colocation. In any case, the analysis given below will apply equally to spatial and mereological formulations of the supposed problem of colocation. (Thanks to Simon Evnine for raising this issue.) 3. Though these are not exact quotes, the examples are Ryle’s (1949/ 1984, 22 and 16, respectively). 4. At least no additional spatial parts. If one accepts the wider notion of ‘logical parts’, and with it the idea that modal properties of objects may be
214
notes to pages 77–83
among their logical parts, then of course the statue may have different logical parts from the lump. See Laurie Paul (2002). 5. To borrow a phrase from Ryle (1949/1984, 23). 6. This is similar to Lowe’s claim that, e.g., the question ‘How many dogs and animals?’ lacks determinate sense (1989, 105). He analyses the sense failure, here, though, as being based on the requirement that for the question ‘How many Fs and Cs?’ to make determinate sense, F and C must be disjoint sorts of kinds (105), not that the sortals be analytically or conceptually unrelated in the ways I have described. 7. Alan Sidelle (1992b, 423) offers a different but complementary assessment of why our inclination to accept such ‘counts’ does not mean that we are committed to denying the (uncounted) entities. 8. Or, alternatively: the fact that there is a lump with all of the following properties, relations, history, context . . . made of these parts and present in this space analytically entails that there is a statue constituted by that lump, made of those parts and present in that space. Of course some (e.g. van Inwagen, Merricks) might be inclined to deny that there are the relevant analytic entailments in these cases. In chapter 9 I return to defend the claim that these analytic entailments do indeed hold; here I will rely just on the informal criterion suggested in chapter 1. 9. That is not to say that this is the only restriction such a principle should ultimately be subjected to; other restrictions (e.g. for cases of material constitution generally, not necessarily involving analytic interrelations between the terms involved in the claims) might also be argued for. 10. For other solutions to the weight problem, see Zimmerman (1995) and Baker (2000, 167–79). 11. This, I should note, is only presented as a sufficient condition for properties’ not ‘doubling up’, not a necessary condition. There may also be other cases, e.g. of material constitution where the relevant analytic entailments do not hold, in which there is also no such doubling up. For a more general reply to this sort of problem, see Baker (2000, 167–79). 12. For an excellent discussion of the grounding problem, see Bennett (2004). 13. Mark Heller similarly argues that the modal differences between, e.g., lumps and statues (e.g. the different conditions under which each could survive) must have some nonmodal ground in different microstructural parts or forces acting on them (1990, 30–32). 14. Alan Sidelle similarly suggests that, on his view, we should not expect identity conditions to supervene on mind-independent facts, so colocation can better be made sense of (1992a, 288). 15. Or something of that sort—the details of the application conditions for ‘statue’ and ‘lump’ may be left for another occasion. In any case, however, it’s useful to note that those application conditions need not just appeal to conditions at a time but may also appeal to conditions over time (as, e.g., the
notes to pages 87–95
215
application conditions for ‘statue’ may appeal to an artist’s past intentional activities). Lowe (2005) emphasizes the importance of diachronic identity conditions. Chapter 5 1. Recent uses of this form of argument may be found in Unger (1979b) and Wheeler (1979). 2. In this regard, however, they are similar to Merricks’s use of the causal redundancy argument. 3. Of course Unger himself might also be accused of making the concept of stone artificially detailed and precise by claiming that proposition 3 is part of the ordinary concept of stone, when it seems no less incredible that the ordinary concept involve commitment to an indefinite number of iterations of a recursive principle than that the ordinary concept involve some exact boundary. 4. Bertrand Russell (1923/1997) famously held that all natural language expressions are vague and, in fact, that all expressions are vague, since even logical propositions were ‘built on the substructure’ of natural language propositions and thus inherited the latter’s vagueness. Precision, on his view, is no more than an ideal that cannot be reached in practice (65). 5. Similar examples of stipulated definitions with areas of indeterminate application are provided by Soames (1989). 6. In fact, as Ted Cohen (1991) has ingeniously pointed out, the rules of baseball are in an even worse situation: The rules of baseball are apparently not incomplete but self-contradictory regarding what happens in the case of ties. 7. Martin Golding (2003) similarly argues that legal analog of the principle of bivalence fails—that is, that even within a particular legal system, there are some statements of law (e.g. whether or not Jones is liable) that are indeterminate (neither valid/correct nor invalid/incorrect within that system). 8. I have argued this point elsewhere for the limited cases of fiction (Thomasson 2003a) and art (2005). 9. For the other two cases, of course, vagueness may creep in in addition to the indeterminacies, thanks to the vagueness of defining expressions such as ‘hole’, ‘tagged’, etc. 10. Of course, technical terms also may suffer from the complete boundarylessness characteristic of true vagueness, since even explicit authoritative stipulations at least generally must utilize customarily defined terms, so that the vagueness of the latter infects the former. 11. The latter point echoes Laurence Goldstein’s (1988) observation that judgments involving vague predicates are liable to revision, and speakers need not be led down the path of self-contradiction in a forced-march sorites as long as they are permitted to change their minds.
216
notes to pages 97–107
12. Tye (1990) shows that the fact that proposition 3 is indefinite is not a separate ad hoc assumption to avoid the sorites problems, but instead is a consequence of certain natural assumptions of a 3-valued logic combined with the standard line of reply to sorites arguments that some statements ascribing the predicate ‘stone’ to an object are indefinite in truth-value. For any conditional with an indefinite antecedent and conclusion itself has an indefinite truth-value, and any universally quantified statement with an indefinite truthvalue for some assignment (and which is false for no assignment) is indefinite in truth-value. As a result, if there is even one case where ‘x is a stone’ and ‘(x minus one atom) is a stone’ are both indefinite, the conditional is indefinite, and the universally quantified statement 3 is indefinite. 13. Among the commonly discussed disadvantages for supervaluational treatments is that, although they enable us to preserve the truth of the law of the excluded middle (since, on any legitimate precisification of a vague predicate ‘P’, ‘(Pa) v (:Pa)’ is true), they require us to accept that some disjunctions— such as that one—may be true without either disjunct being true (since the truth values of ‘Pa’ and ‘:Pa’ will vary for different precisifications). Similarly, they must allow that some existentially quantified claims are true even though no substitution instance is true (as, e.g., it is true on any legitimate precisification that there is some height that is the first-tall height, though there is no height of which this is true) and some universally quantified claims that are false, though no substitution instance is false. Tye’s indeterminist solution, on the other hand, has the disadvantage that the law of the excluded middle has some substitution instances on which it is indeterminate (when P is indeterminate), although it is never false (Tye 1994/1997, 282). 14. See also Horgan and Potrcˇ (forthcoming). 15. See Hawley (2001, 105–11), however, for an interesting argument that (pace Tye) there is no important distinction between ‘‘ontic indeterminacy in boundaries [or parts] and ontic indeterminacy in other respects’’ such as properties (109). Ontic vagueness, she argues, should not be defined in terms of vagueness in objects, properties, or relations at all, but rather in terms of the possibility of there being indeterminate utterances about that world where the indeterminacy is not attributable to semantic indecision. 16. It is important to note, however, that tracing the source of such ontic vagueness as there is to indeterminacies in our representations does not lead to a kind of antirealist view that, as Sainsbury (1994, 79) describes it, holds that ‘‘our world, before we find it, is an undifferentiated sludge.’’ See xx 3.3 and 10.4 for replies to such charges of antirealism or commitment to a ‘stuff ontology’. 17. On one point, however, there is wide agreement: if one holds that the indeterminacy in question is only a semantic indeterminacy about exactly what (precise) entity the singular terms ‘a’ and ‘b’ refer to, then the argument may be avoided. For we then may no longer make the move from ‘r(a ¼ b)’ to say, of b, that it is such that the predicate being-indeterminate-whether-it-is-identical-to-a
notes to pages 108–122
217
applies to it, since ‘a’ would not be rigid, but rather denote different things on different precisifications (Lewis 1988/1997, 318–9). On the other hand, if we allow that vague terms may refer rigidly to vague objects, then this way around the problem cannot be utilized. 18. I have swapped the as and bs of Johnsen’s original to maintain consistency with Evans’s version of the argument. Chapter 6 1. Of course this is not to say that x and y are not governed by the identity conditions for Ks. The point is rather that at least some of the conditions for Ks to be identical (e.g. spatiotemporal continuity) are not fulfilled by x and y, so they are not identical. 2. Thanks to Alan Sidelle for this fine example. 3. Or at least not truth-evaluable by the same methods as other existence and counting claims. Below I will consider some alternate possibilities for understanding claims about the existence and number of things. 4. Sider does not endorse this view, but describes it as follows: ‘‘the fundamental quantificational notion is a[n] . . . amalgam of quantification and predication: ‘there is an F such that . . .’, where F must be replaced by a sortal’’ (forthcoming, 35). 5. Thanks to Ted Sider for raising this issue. 6. For criticisms of the doctrine of quantifier variance, see van Inwagen (2002, 187–8) and Sider (2001b, xx). 7. Thanks to Ota´vio Bueno for helpful discussion of this point. 8. Of course, others presuppose some ‘higher standards’ for admitting existence, such as contributing distinct causal powers. In x 11.2 I will return to discuss these standards vis-a`-vis the purely descriptive standard proposed above. 9. Sider (2001b, xx) argues that the only variations in the quantifier that make sense are different restrictions on the quantifier (not different senses of ‘there is’). The above suggestion is in effect that (rather than different restrictions on a quantifier that could ultimately be understood as unrestricted) the differences are in what categories of entity the quantifier is implicitly conjoined with. 10. As, e.g., Hirsch (2002b, 60) suggests that the antimereological use is clearly the normal English use of the quantifier. 11. Hilpinen aptly notes that one shouldn’t accept the strong version that would require realism to assert this for all individuating concepts, since it is clearly false (for example) for cultural objects ‘‘whose character is ‘constituted’ by people’s beliefs, e.g., chess pieces and traffic signs’’ (1996, 6). 12. It is important to use categorial terms rather than sortal terms generally in order to avoid ‘double-counting’, since entities a and b picked out using different sortals S and S’ may be identical as long as S and S’ are of the same
218
notes to pages 123–133
category (and thereby employ the same identity conditions), e.g. if S is ‘dog’ and S’ is ‘mammal’. 13. Except, obviously, such sortals as ‘word’. 14. Thanks to Mary Kate McGowan for initially raising this issue for me. One way around this problem, as Lowe has suggested (2003b), is to deny that categories are ‘things’. Chapter 7 1. Van Inwagen (1990, 287 n. 14) credits H. Scott Hestevolt (1980–81) with introducing the Special Composition Question to contemporary philosophical discussion. 2. Although he considers it not to really be an answer to the Special Composition Question (since it is really about sums rather than composition, and since it contains mereological terms and thus can’t be used to define ‘whole’), van Inwagen also considers and rejects the ‘universalist’ view, that for any disjoint xs, there is a y such that the xs compose it (1990, 74)—namely a mereological sum. He rejects it on grounds that, if it were true, then a human would be identified with the mereological sum of its parts—but then the atoms that made parts of a human ten years ago would have to still be identical with that human now. Yet quite evidently, they are not, since the parts are now widely scattered (if they all still exist) (78). Sanford (2003) argues convincingly that ordinary objects are not mereological sums (fusions), and thus in any case a universalist solution—taken alone—would be of little help to the defender of ordinary objects. 3. I return to discuss whether we should accept that there are such objects— even if they are not identical to statues—in x 10.3. 4. Sanford is not alone in suspecting that the problem with arguments from composition lies in assuming that there must be one single, uniform, nonarbitrary answer to the Special Composition Question. Ned Markosian (1998) argues that there is no true (uniform) nontrivial answer to the Special Composition Question, so that composition must just be taken as a brute fact. Eli Hirsch (1993) characterizes van Inwagen’s search for a uniform answer to the Special Composition Question as based on seeking a nonarbitrary concept of existence, but denies that there is such a concept. 5. This obviously doesn’t rule out entailments in the other direction, e.g. from the existence of simples arranged in a certain way, able to collectively perform certain functions, etc., to the existence of a plank. 6. Another option is, with Sanford (1993, 221–2), to deny transitivity, at least for our ordinary concept of x being a part of y (as opposed to being simply part of y, where the above recursive definition might do the job). 7. Horgan (1993) also provides a perspicuous discussion of these issues. 8. Those engaged in what Jackson (1998, 5–6) calls ‘serious metaphysics’ might legitimately be interested in offering a minimal account of what basic
notes to pages 139–143
219
ingredients are needed in terms of which we can make sense of all the rest— but (as Jackson notes) such accounts certainly do not automatically entail that the derivative or less basic entities do not really exist; indeed the privileged basic account may entail that they do. (see x 11.2) Chapter 8 1. Susan Stebbing’s assessment is that ‘‘it is hopeless to attempt to extract from Eddington any consistent view’’ of the relation between the ‘scientific’ and ‘common sense’ tables (1958, 60). Indeed Eddington does seem to oscillate among three or four different positions—including that only the scientific table really exists (x), that they are to be identified ‘‘after some fashion’’ (xii), that the ordinary table is the combined creation of the scientific table and the mind (xii), and the above view that the scientific table is only a symbol or shadow of the common sense table (which then alone is real?) (xiii–xv). The context of discussion in the book as a whole suggests that something like the last is closest to Eddington’s true view, as I treat it above. Carl Hempel (1966, 77–9), however, apparently takes Eddington to be endorsing the first view—that ordinary objects should be eliminated in favor of their scientific counterparts. 2. Heil (2003, 190–2) similarly argues that discoveries that the ultimate stuff of the universe is a ‘‘collection of elementary particles, or fields, or perhaps a single field, a single space-time manifold’’ would not undermine the idea that there are ordinary objects such as trees, mountains, and human beings; such discoveries of physics may tell us more about the underlying physical nature of such things, but does not establish that they do not really exist. 3. On the other hand, possessing the microscopic nature described by physics may be sufficient to ensure that planks and tables are solid in the everyday sense. As Jackson puts it, ‘‘the story in the favoured terms will, we may suppose, tell us that these lattice-like arrays of molecules exclude each other, the intermolecular forces being such as to prevent the lattices encroaching on each others’ spaces. And that is what it takes, according to our concept, to be solid. Or at any rate it is near enough—perhaps pre-scientifically we might have been tempted to insist that being solid required being everywhere dense in addition to resisting encroachment. But resisting encroachment explains the stubbing of toes, the supporting of cups of coffee and the like, quite well enough for it to be pedantic to insist on anything more in order to be solid’’ (1998, 3). 4. I do not, however, mean to endorse Ryle’s whole treatment (1954) of the reasons why there is no rivalry between scientific and everyday statements about the world. For ultimately, he seems to defend their lack of rivalry by denying that the scientific theories are describing objects and properties at all. Instead, he seems to assume the Carnapian view that scientific theories are not in the business of describing the world at all, but simply of specifying the objects’ role in calculations and predictions. Such a view would be inconsistent
220
notes to pages 143–159
with a robust scientific realism and is not required to make the above reply to claims of a rivalry between the scientific and manifest images. 5. Of course this might be better described as an outdated scientific claim that had come to enjoy widespread belief than as a pure claim of common sense. 6. Sellars uses the more old-fashioned phrase ‘‘man first encountered himself—which is, of course, when he came to be man’’ (1963/1991, 6). Since he is self-consciously alluding to existentialist analyses here, the best way to put it might be as the image of Dasein and the World that enabled Homo sapiens to become Dasein. 7. The claim of a ‘redundancy’ here seems closely analogous to claims that positing ordinary objects and scientific objects (or ‘simples’) is implausible, for surely there is really ‘nothing over and above’ or ‘in addition to’ the simples— claims to which I have responded in x 4.1. 8. Sellars discusses a further problem that arises with this option: if the manifest image is to be accounted for in terms of appearances to human minds, we must ask what accounts for the relevant appearance of, e.g. pinkness when we seem to observe a pink ice cube. Sensations, however, seem to be homogeneous in much the same way as the original pinkness of the ice cube seemed to be, and so to resist identification with neurophysiological complexes (1963/ 1991, 35). This, then, threatens to make the scientific image incomplete, since it could not explain how things could even appear to have sensible properties. Nonetheless, Sellars holds out hope that a completed scientific image may bottom out not in particles but in a continuous, nonparticulate foundation that might be more plausibly identified with sensory qualities (37). Chapter 9 1. Except in the obscure sense that those theories that can capably account for our linguistic practices and intuitions (as I argue below) invariably turn out to be committed to ordinary objects (despite the fact that they often explicitly deny this). 2. This not to say, however, that the chair is identical with the particlesso-arranged (they may, e.g., have different identity conditions), or that the two phrases have the same meaning (other situations might also be analytically sufficient to ensure the truth of ‘there are chairs’ on the ordinary use of the term—e.g., if it turned out that there was continuum stuff rather than particles arranged in a certain way, for a certain purpose . . .). 3. This sort of ‘minimalist’ argument is developed at greater length in my 2001b. 4. A similar argument for accepting fictional characters may be found in my 2003a. 5. Except, perhaps, for very special cases like definitions themselves. 6. Though there is one difference from my view, insofar as the phrasing in these quotations suggests that truth-conditions for existence questions involve
notes to pages 159–170
221
first, there being some thing, and then its meeting certain conditions, unlike the view I have defended above. Nonetheless, they are alike in denying that we may define things into existence. 7. Hirsch (2002a) offers a compatible argument for the existence of common sense objects: Revisionary ontologists who deny the existence of common sense objects violate the principle of interpretive charity for English. Thus Hirsch and I agree that eliminativists arrive at their conclusions only by misinterpreting standard English; I would say they do so by artificially inflating the application conditions for ordinary terms. 8. Chalmers and Jackson (2001, 355) make a similar point against eliminativisms that nonetheless accept the existence of conditions sufficient for the existence of the entity they deny: ‘‘We might imagine that there are eliminativists who deny the existence of Mark Twain. They accept the existence of Samuel Clemens, and the description of his exploits: that he went under the name ‘Mark Twain’, that he published many of the works that we know of under that name, and so on. But they deny that Clemens was Twain: Clemens used the name, but Twain never really existed. The response to such an eliminativist seems clear. Once the eliminativist concedes the full description of Clemens’s exploits and of his connection to our current use of the name ‘Mark Twain’, we should say: if you grant all that, you grant that Clemens was Twain. Once the qualitative description is given, there is no further fact about Twain’s existence of which we are ignorant.’’ 9. Of course van Inwagen would reject the first hypostatization (rejecting the transition from asserting that there are particles arranged chairwise to hold that there is a chairwise arrangement of particles). But this is even more obviously (than in the move from the latter to ‘there is a chair’) a pleonasm, so it even more clearly falls into the pattern of the cases described above. 10. John Heil’s efforts to ‘‘deflate a certain conception of what the truthmakers must be for claims about statues to be true’’ (2003, 189) are (though put in different terms) in much the same spirit as the above arguments. 11. Compare the similar claims in my 2001b and Schiffer (1996, 159). 12. John Searle makes much the same point in arguing against Quine’s criterion of ontological commitment, noting that ‘‘sometimes a statement couched in one notational form can involve a commitment which, in some intuitively plausible sense, is exactly the same as the commitment involved in a statement couched in quite a different notational form. . . . Yet on [Quine’s] criterion the two statements, though they involve the same commitments in fact, would involve different commitments’’ (1969, 107). 13. Of course this does not require us to reject Quine’s method of handling nonexistence statements, as, e.g., saying ‘Unicorns do not exist’ certainly does not analytically entail that the application conditions for ‘unicorn’ are fulfilled. 14. Dyke (forthcoming) also argues that the availability of paraphrases is ontologically insignificant. 15. Compare Frege on copses and trees (1884/1968, x 46).
222
notes to pages 171–191
16. Thanks to John Hawthorne for raising this objection. 17. See also Schiffer’s discussion of wishdates (2003, 53–61). 18. Merricks raises this as an objection to an ontology of unrestricted composition and colocation. While this is not identical to the ontology proposed above, the same sort of objection could well be raised for my view. Chapter 10 1. Sider similarly notes that many central metaphysical questions (including those about the existence of mereological sums and of temporal parts) can be considered genuine (as opposed to pseudo-) questions only if one assumes that there is ‘‘a single, objective, correct account of what things there are’’ (2001b, xvi). He argues that allowing that ‘‘the world comes ‘ready-made’ with a single domain D of objects: the class of all the objects there are,’’ provides us with an especially eligible referent of the unrestricted quantifier, thus enabling us to back up the claim that, e.g., the universalist and nihilist are using the quantifier in the same way and genuinely disagreeing rather than simply talking past each other by using the quantifier differently. 2. In defense of a layered view, see, e.g., Smith (2004); in attack, see, e.g., Heil (2003) and Kim (1993, 337–9). 3. The same sorts of argument from chapter 9 also may be wielded in defense of various abstract cultural objects such as governments, literary works, and fictional characters. See, e.g., my 2001a, 2003a. 4. So, similarly, Hirsch (2002a, 106) accepts that, if universalists are seen as introducing a new technical language, they are not in conflict with common sense in claiming that their terms refer. 5. Even if we focus on social and cultural objects, however, as I have argued elsewhere (2003b) the realist may also accept these mind-dependent social and cultural objects, and accept that there are substantial differences between them and independent natural objects (differences that make a difference to the applicability of central claims in metaphysics, epistemology, and semantics), without any problems whatsoever. Chapter 11 1. For another defense of conceptual analysis, see Jackson (1998). 2. Thanks to Liz Giles for suggesting the need to discuss reduction views more directly. 3. At least if ‘reduction’ is supposed to involve identification. If some weaker sense of ‘reduction’ is in view, then it may be admissible, but does not relieve us of the need to give a separate account of what ordinary objects are. 4. Elsewhere I have drawn out some of these consequences for debates about the ontological status of fictional characters (2003a) and works of art (2004b, forthcoming d).
notes to pages 196–200
223
5. Of course, genuine disputes about ontology could still arise if the disputants are disagreeing about whether (in a given situation) a particular set of criteria is met. Debates about composition and ontology, however, seldom seem to be of the second sort—the realist about ordinary objects, van Inwagen, and Horgan do not seem involved in an empirical disagreement about the total world-situation that could make a difference to whether various stable criteria were or were not fulfilled (compare Sidelle 2002, 134–5). Instead, they disagree about whether various situations are properly described as, e.g., one in which there is a dog and a bowl, a dog and some simples arranged bowlwise, or mere perturbations of the blobject. So the disagreements then seem to be about the (properly) associated criteria for objecthood rather than about whether any clear and stable set of criteria is fulfilled. 6. On modal features, see x 3.5. 7. See, e.g., Hirsch (2002a) and Sidelle (2002) for use of similar ideas to reevaluate debates about universalism, presentism, and so on.
This page intentionally left blank
bibliography
Adams, F., G. Fuller, and R. Stecker. (1997). ‘‘The Semantics of Fictional Names.’’ Pacific Philosophical Quarterly 78: 128–48. Alexander, Samuel. (1927). Space, Time and Deity. Vol. 2. London: Macmillan. Austin, J. L. (1962). Sense and Sensibilia. Oxford: Oxford University Press. Ayer, A. J. (1952). Language, Truth and Logic. New York: Dover. Baker, Lynne Rudder. (2000). Persons and Bodies: A Constitution View. Cambridge: Cambridge University Press. Bennett, Karen. (2004). ‘‘Spatio-Temporal Coincidence and the Grounding Problem.’’ Philosophical Studies 118: 339–71. Boghossian, Paul. (1997). ‘‘Analyticity.’’ In Bob Hale and Crispin Wright, eds., A Companion to the Philosophy of Language. Oxford: Blackwell. Braun, David. (1993). ‘‘Empty Names.’’ No^ us 27, 4: 449–69. Burke, Michael. (1992). ‘‘Copper Statues and Pieces of Copper: A Challenge to the Standard Account.’’ Analysis 52, 1: 12–7. ———. (1994). ‘‘Dion and Theon: An Essentialist Solution to an Ancient Puzzle.’’ Journal of Philosophy 91, 3: 129–39. Carnap, Rudolf. (1937). Foundations of Logic and Mathematics. Chicago: University of Chicago Press. Chalmers, David. (1996). The Conscious Mind. Oxford: Oxford University Press. Chalmers, David, and Frank Jackson. (2001). ‘‘Conceptual Analysis and Reductive Explanation.’’ Philosophical Review 110, 3: 315–60. Cohen, Ted. (1991). ‘‘There Are No Ties at First Base.’’ In Bill Henderson, ed., The Pushcart Prize, vol. 16. Yonkers, N.Y.: Pushcart Press. Creath, Richard. (2000). ‘‘The Initial Reception of Carnap’s Doctrine of Analyticity.’’ In Føllesdal (2000). 225
226
bibliography
———. (2004). ‘‘Quine on the Intelligibility and Relevance of Analyticity.’’ In Gibson (2004a). Currie, Gregory. (1989). An Ontology of Art. New York: St. Martin’s Press. Davies, David. (2004). Art as Performance. Malden, Mass.: Blackwell. Devitt, Michael. (1981). Designation. New York: Columbia University Press. ———. (1991). Realism and Truth. 2nd ed. Princeton, N.J.: Princeton University Press. Devitt, Michael, and Kim Sterelny. (1999). Language and Reality. 2nd ed. Cambridge, Mass.: MIT Press. Dummett, Michael. (1973/1981). Frege: Philosophy of Language. 1st ed. New York: Harper & Row. 2nd ed. Cambridge, Mass.: Harvard University Press. Dupre´, John. (1981). ‘‘Natural Kinds and Biological Taxa.’’ Philosophical Review 87: 519–47. Dyke, Heather. (Forthcoming). The Representational Fallacy. London: Routledge. Eddington, A. S. (1928). The Nature of the Physical World. New York: Macmillan. Elder, Crawford. (1989). ‘‘Realism, Naturalism and Culturally Generated Kinds.’’ Philosophical Quarterly 39: 425–44. ———. (2004). Real Natures and Familiar Objects. Cambridge, Mass.: MIT Press. Evans, Gareth. (1978/1997). ‘‘Can There Be Vague Objects?’’ Analysis 38: 208. In Keefe and Smith, eds. (1997). Fine, Kit. (1975/1997). ‘‘Vagueness, Truth and Logic.’’ Synthese 30: 265–300. In Keefe and Smith, eds. (1997). Føllesdal, Dagfinn, ed. (2000). Philosophy of Quine: General, Reviews, and Analytic/Synthetic. New York: Garland. Frege, Gottlob. (1884/1968). Die Grundlagen der Arithmetik. Breslau: Wilhelm Koebner. The Foundations of Arithmetic. Translated by J. L. Austin. Evanston, Ill.: Northwestern University Press. French, Steven. (2003). ‘‘Scribbling on the Blank Sheet: Eddington’s Structuralist Conception of Objects.’’ Studies in History and Philosophy of Modern Physics 34: 227–59. Geach, Peter. (1962/1980). Reference and Generality. 1st ed. Ithaca, NY: Cornell University Press. 3rd ed. Ithaca, N.Y.: Cornell University Press. Gibbard, Allan. (1997). ‘‘Contingent Identity.’’ In Rea (1997b). Gibson, Roger F. (2004a). The Cambridge Companion to Quine. Cambridge: Cambridge University Press. ———. (2004b). ‘‘Quine’s Behaviorism cum Empiricism.’’ In Gibson (2004a). Gilbert, Margaret. (1996). Living Together. Lanham, Md.: Rowman and Littlefield. Golding, Martin. (2003). ‘‘The Legal Analog of the Principle of Bivalence.’’ Ratio Juris 16, 4: 450–68.
bibliography
227
Goldstein, Laurence. (1988). ‘‘The Sorites as a Lesson in Semantics.’’ Mind 97, 387: 447–55. Grice, Paul. (1989) Studies in the Way of Words. Cambridge, Mass.: Harvard University Press. Griffin, Nicholas. (1977). Relative Identity. Oxford: Oxford University Press, 1977. Hawley, Katherine. (2001). How Things Persist. Oxford: Oxford University Press. Heil, John. (2003). From an Ontological Point of View. Oxford: Oxford University Press. Heller, Mark. (1990). The Ontology of Physical Objects. Cambridge: Cambridge University Press. Hempel, Carl G. (1966). Philosophy of Natural Science. Englewood Cliffs, N.J.: Prentice-Hall. Hestevolt, H. Scott. (1980–81). ‘‘Conjoining.’’ Philosophy and Phenomenological Research 41: 371–83. Hilpinen, Risto. (1992). ‘‘On Artifacts and Works of Art.’’ Theoria 58: 58–82. ———. (1996). ‘‘On Some Formulations of Realism, or How Many Objects are there in the World?’’ In R. S. Cohen, R. Hilpinen, and Qiu Renzong, eds., Realism and Anti-realism in the Philosophy of Science. Netherlands: Kluwer. Hirsch, Eli. (1982). The Concept of Identity. Oxford: Oxford University Press. ———. (1993). Dividing Reality. New York: Oxford University Press. ———. (2002a). ‘‘Against Revisionary Ontology.’’ Philosophical Topics 30, 1: 103–27. ———. (2002b). ‘‘Quantifier Variance and Realism.’’ In Sosa and Villanueva (2002). Horgan, Terence. (1991). ‘‘Metaphysical Realism and Psychologistic Semantics.’’ Erkenntnis 34: 297–322. ———. (1993). ‘‘On What There Isn’t.’’ Philosophy and Phenomenological Research 50: 693–700. ———. (1994). ‘‘Transvaluationism: A Dionysian Approach to Vagueness.’’ Southern Journal of Philosophy 23, supp.: 97–132. ———. (1998). ‘‘The Transvaluationist Conception of Vagueness.’’ Monist 81, 2: 313–30. Horgan, Terence, and Matjazˇ Potrcˇ. (2000). ‘‘Blobjectivism and Indirect Correspondence.’’ Facta Philosophica 2: 249–70. ———. (Forthcoming). Austere Realism: Contextual Semantics Meets Minimal Ontology. Cambridge, Mass.: MIT Press. Husserl, Edmund. (1954/1970). The Crisis of European Sciences and Transcendental Phenomenology. Die Krisis der europa¨ischen Wissenschaften und die transzendentale pha¨nomenologie. Translated by David Carr. Evanston, Ill.: Northwestern University Press.
228
bibliography
Jackson, Frank. (1998). From Metaphysics to Ethics: A Defence of Conceptual Analysis. Oxford: Clarendon Press. Johnsen, Bruce. (1989). ‘‘Is Vague Identity Incoherent?’’ Analysis 49, 2: 103–12. Johnston, Mark. (1987). ‘‘Is there a Problem about Persistence?’’ Proceedings of the Aristotelian Society 61, supp.: 107–35. ———. (1988). ‘‘The End of the Theory of Meaning.’’ Mind and Language 3, 1: 28–63. ———. (1992). ‘‘Reasons and Reductionism.’’ Philosophical Review 101, 3: 589–618. ———. (1997). ‘‘Constitution Is Not Identity.’’ In Rea (1997b). Keefe, Rosanna, and Peter Smith, eds. (1997). Vagueness: A Reader. Cambridge, Mass.: MIT Press. Kim, Jaegwon. (1993). Supervenience and Mind. Cambridge: Cambridge University Press. Kitcher, Philip. (1982). ‘‘Genes.’’ British Journal for the Philosophy of Science 33: 337–59. Kornblith, Hilary. (1980). ‘‘Referring to Artifacts.’’ Philosophical Review 89, 1: 109–14. Kriegel, Uriah. (Unpublished). ‘‘For Your Eyes Only: Perception and the Realm of Appearances.’’ Kripke, Saul. (1980). Naming and Necessity. Oxford: Blackwell. Lakoff, George. (1987). Women, Fire and Dangerous Things. Chicago: University of Chicago Press. Lewis, David. (1986). On the Plurality of Worlds. Oxford: Blackwell. ———. (1988/1997). ‘‘Vague Identity: Evans Misunderstood.’’ Analysis 48: 128–30. In Keefe and Smith, eds. (1997). ———. (1997). ‘‘Counterparts or Double Lives.’’ In Rea (1997b). Lowe, E. J. (1989). Kinds of Being: A Study of Individuation, Identity and the Logic of Sortal Terms. Oxford: Blackwell. ———. (2002). A Survey of Metaphysics. Oxford: Oxford University Press. ———. (2003a). ‘‘In Defense of Moderate-Sized Specimens of Dry Goods.’’ Philosophy and Phenomenological Research 67, 3: 704–10. ———. (2003b). ‘‘Recent Advances in Metaphysics.’’ Facta Philosophica 5: 3–24. ———. (2004). ‘‘Some Formal Ontological Relations.’’ Dialectica 58: 297–316. ———. (2005). ‘‘How Are Ordinary Objects Possible?’’ Monist 88, 4: 510–33. Markosian, Ned. (1998). ‘‘Brutal Composition.’’ Philosophical Studies 92: 211–49. Martin, C. B., and John Heil. (1999). ‘‘The Ontological Turn.’’ Midwest Studies in Philosophy 23: 34–60. Martin, C. B. (1997). ‘‘On the Need for Properties: The Road to Pythagoreanism and Back.’’ Synthese 122, 2: 193–231. Martin, Michael, and Lee C. McIntyre, eds. (1994). Readings in the Philosophy of Social Science. Cambridge, Mass.: MIT Press. Mates, Benson. (2000). ‘‘Analytic Sentences.’’ In Føllesdal (2000).
bibliography
229
Merricks, Trenton. (2000). ‘‘No Statues.’’ Australasian Journal of Philosophy 78, 1: 47–52. ———. (2001). Objects and Persons. Oxford: Clarendon Press. Miller, Richard. (1992). ‘‘A Purely Causal Solution to One of the Qua Problems.’’ Australasian Journal of Philosophy 70, 4: 425–34. Millikan, Ruth. (1984). Language, Thought and Other Biological Categories. Cambridge, Mass.: MIT Press. ———. (2000). On Clear and Confused Ideas. Cambridge: Cambridge University Press. Moore, G. E. (1959). ‘‘Four Forms of Scepticism.’’ In Philosophical Papers. London: Allen and Unwin. Myro, George. (1997). ‘‘Identity and Time.’’ In Rea (1997b). Noonan, Harold. (1997). ‘‘Relative Identity.’’ In Bob Hale and Crispin Wright, eds., A Companion to the Philosophy of Language. Oxford: Blackwell. Papineau, David. (1979). Theory and Meaning. Oxford: Oxford University Press. Paul, Laurie A. (2002). ‘‘Logical Parts.’’ No^ us 36, 4: 578–96. Putnam, Hilary. (1975). Mind, Language and Reality. Cambridge: Cambridge University Press. ———. (1987). The Many Faces of Realism. LaSalle, Ill.: Open Court. ———. (2000). ‘‘Two Dogmas Revisited.’’ In Føllesdal (2000). Quine, W. V. O. (1953/2001). From a Logical Point of View. 1st ed. Cambridge, Mass: Harvard University Press. 2nd rev. ed. Cambridge, Mass.: Harvard University Press. ———. (1966/1976). The Ways of Paradox and Other Essays. 1st ed. New York: Random House. Rev. Enl. ed. Cambridge, Mass.: Harvard University Press. ———. (1974). The Roots of Reference. La Salle, Ill.: Open Court. ———. (1990). Pursuit of Truth. Cambridge, Mass.: Harvard University Press. ———. (1991). ‘‘Two Dogmas in Retrospect.’’ Canadian Journal of Philosophy 21: 265–74. Rea, Michael C. (1997a). ‘‘Supervenience and Colocation.’’ American Philosophical Quarterly 34, 3: 367–75. ———. (2002). World without Design: The Ontological Consequences of Naturalism. Oxford: Clarendon Press. ———, ed. (1997b). Material Constitution: A Reader. Lanham, Md.: Rowman and Littlefield. Reimer, Marga. (2001). ‘‘The Problem of Empty Names.’’ Australasian Journal of Philosophy 79, 4: 491–506. Ruben, David-Hillel. (1985). The Metaphysics of the Social World. London: Routledge. Russell, Bertrand. (1923/1997). ‘‘Vagueness.’’ Australasian Journal of Philosophy and Psychology 1: 84–92. In Keefe and Smith, eds. (1997). ———. (1927). The Analysis of Matter. London: Kegan Paul.
230
bibliography
Ryle, Gilbert. (1938/1971). ‘‘Categories.’’ Proceedings of the Aristotelian Society 38 (1938). Reprinted in Collected Papers, vol. 2. Bristol: Thoemmes. ———. (1949/1984). The Concept of Mind. London: Hutchinson. Reprint, Chicago: University of Chicago Press. ———. (1954). Dilemmas. Cambridge: Cambridge University Press. Sagoff, Mark. (1978). ‘‘On Restoring and Reproducing Art.’’ Journal of Philosophy 75, 9: 453–70. Sainsbury, Mark. (1989). ‘‘What Is a Vague Object?’’ Analysis 49, 2: 99–103. ———. (1994). ‘‘Why the World Cannot Be Vague.’’ Southern Journal of Philosophy 33, supp.: 63–81. Sanford, David. (1993). ‘‘The Problem of the Many, Many Composition Questions, and Naı¨ve Mereology.’’ Nouˆs 27, 2: 219–28. ———. (2003). ‘‘Fusion Confusion.’’ Analysis 63, 227: 1–4. Schiffer, Stephen. (1994). ‘‘A Paradox of Meaning.’’ Nouˆs 28: 279–324. ———. (1996). ‘‘Language-Created Language-Independent Entities.’’ Philosophical Topics 24, 1: 149–67. ———. (1998). ‘‘Two Issues of Vagueness.’’ Monist 88: 193–214. ———. (2001). ‘‘Meanings.’’ In J. Campbell, M. O’Rourke, and D. Shier, eds., Essays on Meaning and Truth. New York: Seven Bridges Press. ———. (2003). The Things We Mean. Oxford: Oxford University Press. Schroeter, Laura. (2004). ‘‘The Limits of Conceptual Analysis.’’ Pacific Philosophical Quarterly 85: 425–53. Searle, John. (1969). Speech Acts: An Essay in the Philosophy of Language. Cambridge: Cambridge University Press. ———. (1995). The Construction of Social Reality. New York: Free Press. Sellars, Wilfrid. (1956/1997). ‘‘Empiricism and the Philosophy of Mind.’’ In Herbert Feigl and Michael Scriven, eds., Minnesota Studies in the Philosophy of Science Vol. 1: The Foundations of Science and the Concepts of Psychology and Psychoanalysis. Minneapolis: University of Minnesota Press: 253–329. Empiricism and the Philosophy of Mind. Cambridge, Mass.: Harvard University Press. ———. (1963/1991). Science, Perception and Reality. London: Routledge. Atascadero, Calif.: Ridgeview. Sidelle, Alan. (1989). Necessity, Essence and Individuation: A Defense of Conventionalism. Ithaca, N.Y.: Cornell University Press. ———. (1992a). ‘‘Identity and the Identity-Like.’’ Philosophical Topics 20, 1: 269–92. ———. (1992b). ‘‘Rigidity, Ontology and Semantic Structure.’’ Journal of Philosophy 89, 8: 410–30. ———. (1998). ‘‘A Sweater Unraveled: Following One Thread of Thought for Avoiding Coincident Entities.’’ No^ us 32, 4: 423–48. ———. (2002). ‘‘Is There a True Metaphysics of Material Objects?’’ In Sosa and Villanueva (2002).
bibliography
231
Sider, Theodore. (2001a). ‘‘Criteria of Personal Identity and the Limits of Conceptual Analysis.’’ Philosophical Perspectives 15: 189–209. ———. (2001b). Four-Dimensionalism. Oxford: Oxford University Press. ———. (2003a). ‘‘Reductive Theories of Modality.’’ In Michael J. Loux and Dean W. Zimmerman, eds., The Oxford Handbook of Metaphysics. Oxford: Oxford University Press. ———. (2003b). ‘‘What’s So Bad about Overdetermination?’’ Philosophy and Phenomenological Research 67, 3: 719–26. ———. (2006). ‘‘Quantifiers and Temporal Ontology.’’ Mind 115: 75–97. ———. (Forthcoming). ‘‘Ontological Realism.’’ In David Chalmers, David Manley, and Ryan Wasserman, eds., Metametaphysics. Oxford: Oxford University Press. Smith, David W. (2004). Mind World: Essays in Phenomenology and Ontology. Cambridge: Cambridge University Press. Soames, Scott. (1989). ‘‘Presupposition.’’ Handbook of Philosophical Logic. Vol. 4. Topics in the Philosophy of Language. Edited by D. Gabbay and F. Guenthner. Dordrecht: Reidel. Sosa, Ernest, and Enrique Villanueva. (2002). Philosophical Issues 12: Realism and Relativism. Oxford: Blackwell. Stalnaker, Robert. (1973). ‘‘Presuppositions.’’ Journal of Philosophical Logic 2: 447–57. Stanford, P. Kyle, and Philip Kitcher. (2000). ‘‘Refining the Causal Theory of Reference for Natural Kind Terms.’’ Philosophical Studies 97: 99–129. Stebbing, L. Susan. (1958). Philosophy and the Physicists. New York: Dover. Sterelny, Kim. (1983). ‘‘Natural Kind Terms.’’ Pacific Philosophical Quarterly 64: 110–25. Strawson, P. F. (1950). ‘‘On Referring.’’ Mind 59: 320–44. ———. (1959/1963). Individuals. London: Routledge. London: Methuen. Strawson, P. F., and H. P. Grice. (1956). ‘‘In Defense of a Dogma.’’ Philosophical Review 65, 2: 141–58. Thomasson, Amie L. (1998). ‘‘A Non-reductivist Solution to Mental Causation.’’ Philosophical Studies 89: 181–191. ———. (1999). Fiction and Metaphysics. Cambridge: Cambridge University Press. ———. (2001a). ‘‘Geographic Objects and the Science of Geography.’’ Topoi 20: 149–59. ———. (2001b). ‘‘Ontological Minimalism.’’ American Philosophical Quarterly 38, 4: 319–31. ———. (2003a). ‘‘Fictional Characters and Literary Practices.’’ British Journal of Aesthetics 43, 2: 138–57. ———. (2003b). ‘‘Realism and Human Kinds.’’ Philosophy and Phenomenological Research 67, 3: 580–609. ———. (2003c). ‘‘Speaking of Fictional Characters.’’ Dialectica 57, 2: 207–26.
232
bibliography
———. (2004a). ‘‘Categories.’’ In The Stanford Encyclopedia of Philosophy. http://plato.stanford.edu/entries/categories. ———. (2004b). ‘‘The Ontology of Art.’’ In Peter Kivy, ed., The Blackwell Guide to Aesthetics. Oxford: Blackwell. ———. (2005). ‘‘The Ontology of Art and Knowledge in Aesthetics.’’ Journal of Aesthetics and Art Criticism 63, 3: 221–30. ———. (Forthcoming a). ‘‘Answerable and Unanswerable Questions.’’ In David Chalmers, David Manley, and Ryan Wasserman, eds., Metametaphysics. Oxford: Oxford University Press. ———. (Forthcoming b). ‘‘Artifacts and Other Human Creations.’’ In Stephen Laurence and Eric Margolis, eds., Creations of the Mind. Oxford: Oxford University Press. ———. (Forthcoming c). ‘‘Conceptual Analysis in Phenomenology and Ordinary Language Philosophy.’’ In Michael Beaney, ed., The Analytic Turn: Analysis in Early Analytic Philosophy and Phenomenology. London: Routledge. Tye, Michael. (1990). ‘‘Vague Objects.’’ Mind 99, 396: 535–57. ———. (1994). ‘‘Vagueness: Welcome to the Quicksand.’’ Southern Journal of Philosophy 33: 1–22. ———. (1994/1997). ‘‘Sorites Paradoxes and the Semantics of Vagueness.’’ Philosophical Perspectives 8: 189–206. In Keefe and Smith, eds. (1997). ———. (2000). ‘‘Vagueness and Reality.’’ Philosophical Topics 28, 1: 195–209. Unger, Peter. (1979a). ‘‘I Do Not Exist.’’ In G. F. MacDonald, ed., Perception and Identity. Ithaca, N.Y.: Cornell University Press. ———. (1979b). ‘‘There Are No Ordinary Things.’’ Synthese 41: 117–54. ———. (1979c). ‘‘Why There Are No People.’’ Midwest Studies in Philosophy 4: Studies in Metaphysics. Ed. P. A. French, Jr., T. E. Uehling, and H. K. Wettstein. Minneapolis: University of Minnesota Press: 177–222. ———. (1980). ‘‘The Problem of the Many.’’ Midwest Studies in Philosophy 5: Studies in Epistemology. Edited by P. A. French, Jr., T. E. Uehling, and H. K. Wettstein. Minneapolis: University of Minnesota Press: 411–67. van Inwagen, Peter. (1990). Material Beings. Ithaca, N.Y.: Cornell University Press. ———. (1998). ‘‘Metaontology.’’ Erkenntnis 48: 233–50. ———. (2002). ‘‘The Number of Things.’’ In Sosa and Villanueva (2002): 176–196. Wheeler, Samuel C. (1979). ‘‘On That Which Is Not.’’ Synthese 41: 155–73. Wiggins, David. (1968). ‘‘On Being in the Same Place at the Same Time.’’ Philosophical Review 77, 1: 90–5. ———. (2001). Sameness and Substance Renewed. Cambridge: Cambridge University Press. Williamson, Timothy. (2003). ‘‘Everything.’’ Philosophical Perspectives 17: 415–65.
bibliography
233
Wright, Crispin. (1983). Frege’s Conception of Numbers as Objects. Aberdeen: Aberdeen University Press. Yablo, Stephen. (1992). ‘‘Mental Causation.’’ Philosophical Review 101: 245– 280. Reprinted in David J. Chalmers, ed., Philosophy of Mind: Classical and Contemporary Readings. Oxford: Oxford University Press. ———. (2002). ‘‘Abstract Objects: A Case Study.’’ In Sosa and Villanueva (2002). Zimmerman, Dean. (1995). ‘‘Theories of Masses and Problems of Constitution.’’ Philosophical Review 104: 53–110.
This page intentionally left blank
index
a priori entailments, 206 n. 2 Adams, F., 209 n. 28 Alexander’s dictum, 4, 203 n. 2 analytic/synthetic distinction, 6, 29–37. See also analyticity analyticity, 6, 28–38, 59–60, 68–70, 84–85, 167–68 conventionalism about, 32–34 of basic modal claims 6, 54, 59–64, 178 (see also modal conceptualism) analytic entailments, 6, 16, 21–24, 28–30, 36, 38, 44–45, 52–53, 59–63, 162, 177–78, 180, 181–82, 204 n. 11, 214 n. 8 between claims about the physical and the mental, 24–26 between claims about simples arranged p-wise and claims about ps, 16–18, 164–66 and arguments from parsimony, 153, 177 and causal redundancy claims, 16–24, 25, 177 and ontological commitment, 159–68
and problems of colocation, 79–80, 177 and problems of property additivity, 80–81, 177 antirealism, 60–61, 63–68, 119–21, 185–87, 212 n. 18, 212–13 n. 19, 216 n. 16 application conditions, 39–45, 55–58, 83–84, 181, 199, 208 n. 19, 211 n. 4 for common sense terms, 141–42, 155–59 discovery versus revision of, 48–53, 199–200, 211–12 n. 6 as established by human intentionality, 91–95, 186–87 and existence conditions, 48, 55, 58, 155, 193 frame-level versus empirical, 39–40, 91, 94 indeterminacies in, 41, 92–95, 99, 106–7 inflated by eliminativists, 163–66, 170, 193, 221 n. 7 art, ontology of, 189–92, 215 n. 8 Austin, J. L., 14
235
236
index
Baker, Lynne Rudder, 214 nn. 10, 11 Barnett, David, 204 n. 7 Bennett, Karen, 82, 84 bligers, 169 block (in name use chain), 46–48 Boghossian, Paul, 59–60, 70, 206 n. 4, 207 n. 14 Braun, David, 209 n. 28 brevity, conversational maxim of, 13 Burke, Michael, 82–83 Carnap, Rudolf, 31, 206 nn. 5, 6 categorial concept, 39, 41–43, 57, 59, 106, 111, 116 role of in existence and counting questions, 111, 113–17, 119–25, 148, 154, 179, 217–18 n. 12. See also categories categorial quantification. See quantification, categorial categories, 7, 38–39, 42, 209 n. 23 discovery versus revision of, 50–52 and identity conditions, 57, 210 n. 2 and reduction, 190 See also categorial concept category mistake, 13, 204 n. 6 causal principle, 11, 19–20, 25, 27, 177, 205–6 n. 23 causal redundancy arguments, 4, 6, 9–27, 77–78, 177 Chalmers, David, 26, 174, 205 nn. 19, 21, 206 n. 2, 209 n. 26, 210 n. 33, 221 n. 8 coapplication conditions, 40–44, 51, 54–55, 56–57, 83–84, 180–82, 199–200 discovery versus revision of, 51–53 as established by human intentionality, 91– 92, 186–87 and identity conditions, 56–59, 190–91
role in counting questions, 112, 114–15, 120–21, 179 Cohen, Ted, 215 n. 6 coincidence. See colocation colocation, 4, 6, 73–86, 177, 213 n. 2, 214 n. 14 completeness (in accounts of what there is), 146–50, 179 composition, 110–11, 126–36, 156, 179, 218 nn. 2, 4, 223 n. 5 conceptual analysis, 54, 58–59, 61, 104, 111, 189–93, 199–201, 211–12 n. 6, 212 n. 7 conceptual content, 29, 37, 38, 40, 45, 53, 58–59, 65–66, 95, 161–62, 180–82, 210 n. 33. See also application conditions; categorial concept; coapplication conditions conceptualism, modal. See modal conceptualism constitution, 79–81, 183, 214 n. 11 contextual semantics, 100–104 convention, truth by, 31, 32–37, 60, 206 n. 8 conventionalism about analyticity, 33, 59–62, 70 about logical truths, 33–34, 206 n. 9 about modality, 62–63, 67–68 counting, claims and questions about, 7, 111–15, 118–25, 154–55, 158–60, 177, 178–79, 180, 185, 198 Creath, Richard, 31, 34 criteria of application. See application conditions criteria of coapplication. See coapplication conditions cultural objects, 186–87, 217 n. 11, 222 nn. 3, 5 Currie, Gregory, 190–91 customary terms, 88, 94–95, 98–99, 186
index Davies, David, 190 definitions, as grounding synonymies, 30–37 ‘deep’ ontology, 193–95, 218–19 n. 8 Devitt, Michael, 39, 42–43, 208 n. 16, 210 n. 33 direct reference theories. See reference, causal theories of Dummett, Michael, 38, 40–41, 56, 84, 124, 154, 208 n. 20, 210 nn. 33, 2, 3, 211 n. 4 Dupre´, John, 208 n. 16 Dyke, Heather, 221 n. 14 Eddington, A. S., 137–44, 147, 219 n. 1 Elder, Crawford, 63, 64, 67–68, 212 nn. 11, 14, 16, 212–13 n. 19 eliminativism rhetorical strategies of, 75–78 moderate forms tacitly committed to ordinary objects, 152, 155–60, 173 severe forms, 152, 173–75 epiphenomenalism about ordinary objects, 9–12, 20, 151 in philosophy of mind, 24–27 See also causal redundancy arguments Evans, Gareth, 107–9 existence, claims and questions about, 110–25, 136, 156–59, 178–79, 192–93, 195–201. See also nonexistence claims existence conditions, 55–62, 85–86, 159, 180–82, 192–93 extraordinary objects, 170–73, 183–85 fictional characters, 195, 220 n. 4, 222 nn. 3, 4 fictional names, reference of, 46–48
237
Fine, Kit, 92–94 Frege, Gottlob, 208 n. 20, 221 n. 15 French, Steven, 139 Fuller, Gary, 209 n. 28 Geach, Peter, 210–11 n. 3 Gibson, Roger, 34 Golding, Martin, 215 n. 7 Goldstein, Laurence, 215 n. 11 gollyswoggle, 172–73, 183–85 Grice, Paul, 13–14, 35, 37, 206 n. 8 grounding problem, 4, 6–7, 81–86 Hawley, Katherine, 60–61, 183, 216 n. 15 Hawthorne, John, 222 n. 16 Heil, John, 204 n. 12, 213 n. 24, 219 n. 2, 221 n. 10 Heller, Mark, 74, 78, 214 n. 13 Hempel, Carl, 219 n. 1 Hestevolt, H. Scott, 218 n. 1 Hilpinen, Risto, 120–21, 185–86, 217 n. 11 Hirsch, Eli, 118–19, 120, 123, 217 n. 10, 218 n. 4, 221 n. 7, 222 n. 4 Horgan, Terence, 75, 100–104, 126–27, 132–33, 161, 218 n. 7, 223 n. 5 hoverball, 171–72, 182, 184–85 ‘How many objects/things are there?’. See counting, claims and questions about Husserl, Edmund, 70 identity conditions, 6, 54, 55–62, 63, 67–68, 72, 82–86, 92, 103–4, 106, 111, 178–79, 180–82, 189–92, 208 n. 20, n. 21, 210–11 n. 3, 211 n. 5, 211–12 n. 6, 214 n. 14, 217 n. 1
238
index
identity, indeterminacy of, 107–9 independence presupposition, 13–16, 76–81, 152–53, 177 Jackson, Frank, 193–94, 205 n. 21, 205–6 n. 23, 206 n. 2, 209 n. 26, 210 n. 33 Johnsen, Bruce, 107–8, 217 n. 18 Johnston, Mark, 100, 157, 206–7 n. 10 Keefe, Rosanna, 98 Kim, Jaegwon, 15, 24, 203 n. 2 Kitcher, Philip, 208 n. 16 Kriegel, Uriah, 204 n. 12 layered ontology, 181–82 Lewis, David, 75, 95, 216–17 n. 17 logical truth, 33–34, 206 nn. 4, 6 Lowe, E. J. 56, 57, 114, 119, 203 n. 4, 205 n. 15, 214 n. 6, 214–15 n. 15, 218 n. 14 manifest image, 145–50, 220 n. 8 Markosian, Ned, 218 n. 4 Martin, C. B., 141 McGowan, Mary Kate, 218 n. 14 meanings, 28–30, 33, 37, 38, 45, 53, 65–68, 168, 178, 200, 206–7 n. 10. See also conceptual content mental properties, causal exclusion arguments about, 24–27 mereological sums, 183–85, 218 n. 2 Merricks, Trenton, 9–21, 24–25, 73–74, 75, 77, 78, 151, 152, 156, 165–66, 173, 197, 203 n. 1, 204–5 n. 13, 222 n. 18 metaphysics genuine and pseudodebates in, 112–15, 119–21, 158, 178–79,
189, 192–93, 195–201, 222 n. 1, 223 n. 5 methods of, 8, 188–201 limits of, 8, 188–89, 192 revisionary theories in, 157, 189, 190–91, 221 n. 7 Miller, Richard, 49–50 Millikan, Ruth, 207 n. 15 mind-dependence, for establishing versus fulfilling application conditions, 186–87 modal conceptualism, 54, 63–68, 70–72, 82–83, 213 n. 23. See also analyticity, of basic modal claims modal properties, 63–65, 67, 70–72, 81–86, 208 n. 21, 213 n. 24 modal realism, 70–72, 212–13 n. 19, 213 n. 23 modal truths, 62–68, 70–72, 82–86, 178, 180 Moore, G. E., 3, 188 Myro, George, 78 no coincidence principle, 78–80 nonexistence claims, 38, 45–48 ‘nothing over and above’, 73, 75–78
‘object’, sortal, neutral and covering uses of, 7, 112–15, 117, 120–25, 134–36, 142–43, 154, 158–59, 180, 184, 209 n. 24. See also ‘thing’ Occam’s razor, 151–53, 160, 174, 177 ontological commitment, 159–68, 193–95, 221 n. 12 overdetermination, 10–20, 23–25, 80–81, 204 n. 8, 206 n. 1 paintings, ontology of, 190–91 Papineau, David, 208 n. 16 Paul, Laurie A., 213–14 n. 4
index paraphrase, 151–52, 153, 160–70, 164, 221 n. 14 parsimony, 4, 20, 151–55, 177–79, 189, 192–93, 205 n. 15, 213 n. 1 persistence conditions, 57–62, 72, 82–86, 181, 188–92, 211 n. 4 pleontastic transformations to talk of modal properties, 71–72 and ontological commitment, 162–67, 174, 193, 221 n. 9 Potrcˇ, Matjazˇ, 75, 100 –103, 126, 132–33, 161, 203 n. 1, 216 n. 14 property additivity, 80–81, 177 Putnam, Hilary, 118–21, 185, 207 n. 13 qua problem, 38–44, 48, 52, 91, 117, 208 n. 16 208–9 n. 22 quantification bare versus categorial, 115–18, 121, 217 n. 4, n. 9 unrestricted, 117, 121–25, 135 quantifier variance, 118–19, 217 nn. 6, 9 Quine, W. V. O., 29–37, 167–68, 206 nn. 4, 5, 6, 207 n. 13, 221 nn. 12, 13 Rea, Michael C., 63–67, 71, 212 nn. 12, 14, 18, 213 n. 20 reduction of ordinary objects, 190, 208 n. 18, 222 n. 3 reference causal theories of, 6, 29, 38–48, 53, 58, 178–79, 208 nn. 16, 17, 19 hybrid theories of, 39–44, 47–53, 72 reference shifts, 43, 209 n. 25 Reimer, Marga, 209 n. 28 relative identity, 210–11 n.3
239
relatively minimal entities, 163–68, 171–73 Russell, Bertrand, 107, 215 n. 4 Ryle, Gilbert, 13, 14, 20, 77, 143, 162, 204 n. 6, 213 n. 3, 214 n. 5, 219–20 n. 4 Sagoff, Mark, 191 Sainsbury, Mark, 105, 210 n. 32 Sanford, David, 130, 136, 218 nn. 2, 4, 6 Schiffer, Stephen, 90, 94, 104, 162–63, 171–72, 221 n. 11, 222 n. 17 Schroeter, Laura, 50–52 scientific image, 144–50, 220 n. 8 scientific ontology alleged conflict with common sense ontology, 138–44, 150 alleged rivalry with common sense ontology, 144–50 Searle, John, 221 n. 12 Sellars, Wilfrid, 137, 142–43, 144–50 serious metaphysics. See ‘deep’ ontology Sidelle, Alan, 62–64, 74, 82, 114, 208 n. 17, 210 n. 33, 211 n. 5, 212 nn. 9, 11, 12, 13, 14, 17, 18, 212–13 n. 19, 213 nn. 20, 22, 214 nn. 7, 14, 217 n. 2, 223 nn. 5, 7 Sider, Theodore, 69–70, 159, 196–97, 204 n. 8, 217 nn. 4, 5, 9, 222 n. 1 Smith, David W., 222 n. 2 Soames, Scott, 14, 78, 215 n. 5 social objects, 91, 174, 186–87, 217 n. 11 solidity, 139–42, 194 sorites arguments against ordinary objects, 87–90, 95–101, 215 nn. 3, 11, 216 n. 12
240
index
sortal terms, 39–42, 44–45, 54–59, 91, 182, 210 n. 2, 211 n. 4, 217–18 n. 12 role in existence and counting claims of, 45, 48, 113–14, 120–21, 181, 185–86, 192 special composition question, 110–11, 126–36, 179, 218 nn. 1, 2, 4 Stalnaker, Robert, 14, 78 Stanford, P. Kyle, 208 n. 16 Stebbing, L. Susan, 140–42, 219 n. 1 Stecker, R., 209 n. 28 Sterelny, Kim, 39, 42–43, 208 n. 16, 210 n. 33 Strawson, P. F., 14, 35, 37, 206 n. 8 supervenience principle, 82–86 synonymy, 30–37, 45, 206 n. 8 transvaluationism, 100–101 ‘thing’ and disambiguation of reference, 42, 142–43, 148–50 and existential, quantificational, and counting claims, 112–19, 134, 154–55, 178–79, 180 role in apparent ontological disputes, 196, 198–99 sortal, neutral, and covering uses of, 134–36, 158–59, 183 See also ‘object’ trivial transformations. See pleonastic transformations truth-makers and analytic entailments, 16, 79, 204 n. 12 not needed for analytic truths, 67–70, 180, 207 n. 12 two-dimensional semantics, 209 n. 26, 210 n. 33
Tye, Michael, 95, 96–100, 104–8, 216 nn. 12, 13, 15 unanswerable questions, 58, 112–15, 117, 119, 126–27, 136, 154, 189, 192, 195–99 Unger, Peter, 88–89, 100, 203 n. 1, 215 nn. 1, 3 universalism, 218 n. 2, 222 nn. 1, 4, 223 n. 7
vagueness indeterminist treatments of, 96–100, 216 nn. 12, 13 ontic, 100–109, 216 nn. 15, 16, 216–17 n. 17 role in sorites arguments, 87–89 source of, 90–95, 178 supervaluational treatments of, 95–96, 98–99, 216 n. 13 van Inwagen, Peter, 9, 16–17, 75, 115, 126–36, 152, 157–58, 160–61, 164–65, 168–70, 172–73, 197– 98, 203 n. 1, 217 n. 6, 218 nn. 1, 2, 4, 221 n. 9 Wheeler, Samuel C., 215 n. 1 Wiggins, David, 41, 74, 78, 211 n. 4 Williamson, Timothy, 122 wishdate, 171–72, 182, 222 n. 17 Wright, Crispin, 208 n. 20 Yablo, Stephen, 15, 159, 204 n. 9, 205 n. 14 Zimmerman, Dean, 74, 82, 214 n. 10