Proto Sociology An International Journal of Interdisciplinary Research VOL. 14, 2000 Folk Psychology, Mental Concepts and the Ascription of Attitudes On Contemporary Philosophy of Mind
CONTENTS
Alvin I. Goldman Folk Psychology and Mental Concepts
4
Philip Pettit How the Folk Understand Folk Psychology
26
Jane Heal Understanding Other Minds from the Inside
39
Christopher S. Hill From Assertion to Belief: The Role of Linguistic Data in the Practice of Belief-Ascription
56
David M. Rosenthal Content, Interpretation, and Consciousness
67
Jay L. Garfield Thought as Language: A Metaphor Too Far
85
Robert M. Gordon Sellars’s Ryleans Revisited
102
Louise Röska-Hardy Self-Ascription and Simulation Theory
115
Brian P. McLaughlin Why Intentional Systems Theory Cannot Reconcile Physicalism With Realism about Belief and Desire
145
Gerhard Preyer Primary Reasons: From Radical Interpretation to a Pure Anomalism of the Mental
158
Rebecca Kukla How to Get an Interpretivist Committed
180
David Pitt Nativism and the Theory of Content
222
Raffaella De Rosa On Fodor’s Claim that Classical Empiricists and Rationalists Agree on the Innateness of Ideas
240
Consuelo Preti Belief and Desire Under The Elms
270
Erwin Rogler On David Lewis’ Philosophy of Mind
285
Barbara Von Eckardt, Jeffrey S. Poland In Defense of the Standard View
312
ON CONTEMPORARY PHILOSOPHY OF LANGUAGE, ONTOLOGY AND MORAL THEORY Thomas Baldwin Nearly Logic
332
James E. Tomberlin Logical Form, Actualism, and Ontology
346
Arend Kulenkampff, Frank Siebelt What a Noncognitivist might tell a Moral Realist
355
Contributors
378
Imprint
380
On ProtoSociology
381
Published Volumes
382
Bookpublications
387
Subscription
404
Digital Publications – Special Offer
405
4
Alvin I. Goldman
ALVIN I. GOLDMAN Folk Psychology and Mental Concepts Abstract There are several different questions associated with the study of folk psychology: (1) what is the nature of our commonsense concepts of mental states?, (2) how do we attribute mental states, to ourselves and to other people?, and (3) how do we acquire our concepts and skills at mental-state attribution? Three general approaches to these questions are examined and assessed: theory theory, simulation theory, and rationality theory. A preliminary problem is to define each of these approaches. Alternative definitions are explored, centering on which questions each approach tries to answer and how it answers them. For example, simulation theorists substantially agree on the answer to question (2) but not on the answer to question (1). The paper then turns to some serious problems facing both rationality theory and theory theory. Rationality theory is faulted for its inadequate treatment of question (1) and for its implausible answers to question (2). Theory theory is faulted for the problems it encounters in explaining first-person attribution, and for its treatment of attributed reasoning about change (the “frame problem”). Turning to simulation theory, the paper argues against Gordon’s “ascent routine” account of first-person attribution and in favor of an inner detection account. Finally, the paper addresses the question of the contents of our mental-state concepts. How do these concepts incorporate both behavioral features and inner features? A dual-representation hypothesis is advanced, and linked speculatively to mirror neurons.
I By “folk psychology” I mean the commonsense understanding and deployment of mentalistic concepts, especially the propositional attitudes. Three (or four) principal questions are of interest in the study of folk psychology: (Q1) How do ordinary people understand, or represent to themselves, the various mental states? That is, what are the contents of their concepts of the mental states? (Q2) How do they go about attributing these states? This question decomposes into two subquestions: (Q2)(A) how do people attribute such states to others, and (Q2)(B) how do they attribute such states to themselves? (3) How do people acquire their concepts of mental states and their skill at applying these concepts? The study of folk psychology is the attempt to
Folk Psychology and Mental Concepts
5
answer these questions, an attempt which should be (and is) a combined effort among philosophers and cognitive scientists. Large literatures on the topic already exist in both philosophy and developmental psychology, and other branches of cognitive and neuroscience also have much to contribute. The study of folk psychology might profitably be viewed as a branch of “descriptive epistemology”, since epistemology is (partly) concerned with how beliefs are formed (Goldman 1986) and folk psychology involves the formation of beliefs about a special subject-matter: mental states. By contrast with other portions of philosophy of mind, the study of folk psychology is not a firstorder examination of the nature and constitution of the attitudes (and other mental states). Rather, it is interested in second-order attitudes: beliefs (and concepts) of the folk about the attitudes. These two subjects, of course, are connected. A first step in trying to identify the true nature of mental states, it may well be suggested, is to identify the ordinary understanding of mental states, or what ordinary people mean by mental-state terms. What needs emphasis, however, is that the study of folk psychology is not directly or primarily concerned with the “proper” metaphysical status of mental states, only with the folk’s understanding (or misunderstanding) of mental-state phenomena. For example, it is a metaphysical question whether mental states are identical to neural states. But if mental-neural identity is not part of the folk’s understanding of mental states, then this matter does not come directly under the purview of the study of folk psychology.1 Admittedly, metaphysical questions have traditionally been paramount in philosophy of mind. Are mental states identical with brain states? Do they supervene on the physical states of the agent, or on a wider class of physical states? Jerry Fodor surveys a large set of metaphysical questions in “Fodor’s Guide to Mental Representation” (Fodor 1985): Are propositional attitudes “real”? If so, are they functional states? Do they essentially involve relations to other entities such as propositions, sentences of natural language, or sentences of a language of thought? These metaphysical questions will not occupy me here, because they do not belong to the study of folk psychology as I construe it. To be sure, I adopt the working assumption that beliefs exist (in some sense) because the subject-matter of folk psychology presupposes that ordinary people have beliefs about the attitudes; but this assumption is in principle defeasible.
1
Again, there are important connections between the the study of folk psychology and metaphysics. If mental-state concepts are theoretical concepts, for example, it seems easier to mount an argument for (metaphysical) eliminativism by arguing that the supposed laws believed by the folk are not genuine laws at all, and no states instantiate them.
6
Alvin I. Goldman
Three general approaches to folk psychology will be examined and assessed in the present discussion: the theory theory (TT), the simulation theory (ST), and the rationality (or charity) theory (RT). The first two of these approaches will take center stage, partly because they have been most prominent in recent years. Philosophical formulations of TT and RT are not usually presented in exactly the guise they are considered here. For example, analytical functionalism, which is a version of TT, is typically presented only as an answer to question (Q1), not as an answer to either (Q2) or (Q3). But I am interested, at least in principle, in how TT would answer all three (or four) of the questions we have posed. Our three questions are not wholly unrelated, of course. The content of mental-state concepts may constrain how these concepts can be plausibly deployed in attribution, and evidence about their contents can be obtained by inquiring into their manner of deployment. Which attributional operations are used is (evidentially) relevant to what the contents of the concepts might be.
II The first problem I shall explore is how the three positions on the table should be defined. This is a non-trivial exercise. Within each position – especially TT and ST – different adherents characterize their approaches differently, even when they share the same label. It is standardly assumed that the three approaches are disjoint, but some definitions might imply otherwise. Finally, the possible features that might be used in defining any of the positions are quite varied. A given approach might be defined in terms of its answer to (Q1), its answer(s) to (Q2) (both (A) and (B)), its answer to (Q3), or by some combination of such answers. Theory theory. At least two distinct forms of TT are on offer: a strict and a loose version. The strict version – essentially equivalent to analytical functionalism – says that the commonsense understanding of mental states is wholly in terms of their place in a folk-psychological theory. A theory is construed as a substantial set of laws or generalizations, in the present case, laws connecting various mental states with other mental states, with external circumstances, and with overt behaviors. This version of TT is presented, for example, by David Lewis (1972) and Paul Churchland (1988). Churchland gives a sampling of the folk generalizations or platitudes of the type in question, including: “Persons tend to feel pain at points of recent bodily damage,” “Persons denied fluids for some time tend to feel thirst,” “Persons in pain tend to want to relieve that pain,” and “Persons who want that P, and believe that Q would be
Folk Psychology and Mental Concepts
7
sufficient to bring about P, and have no conflicting wants or preferred strategies, will try to bring it about that Q.” (1988: 58-59) TT suggests that ordinary people’s concepts of mental states are exhausted by their understanding of the roles these postulated states play within such a set of laws.1 Defenders of ST have challenged the strict version of TT by raising doubts about whether all mentalistic attributors possess the sorts of generalizations or platitudes that are usually invoked (Gordon 1986; Heal 1986; Goldman 1989). Mentalistic attributors include children under age (say) six, whose putative knowledge (whether explicit or implicit) of such laws is especially open to question. In light of these challenges, perhaps, certain defenders of TT soften their commitments by omitting any requirement of laws or generalizations. Stich and Nichols (1992) opt for a loose version of TT (what they call a “wide” interpretation), formulated as follows: “just about any internally stored body of information about a domain [is] an internally represented theory of that domain” (1992: 46). This is a somewhat tendentious definition of TT, however, because it implies that a position ascribing to attributors any information about mental states is an instance of TT. This definition threatens to swallow up virtually all opposition, by counting only extremely radical positions as part of the opposition at all. This is not a fruitful way to divide up the available options, because it lumps almost every position into a single category. If we want to avoid the definition-by-laws characterization of TT, but not be so loose as the Stich-Nichols definition, how might TT be characterized? The historical motivation behind TT was to provide an account of mental concepts that fits with traditional empiricist notions of cognitive or semantic respectability.2 The construal of theoretical terms in science was the model on which TT was patterned. So one component of the core conception of TT is that the content of mental concepts should be empirically definable, where this means that the basic terms of the definitions should refer to publicly observable events.3 Indeed, it is common for proponents of TT to remark that mental 1
Notice that TT cannot be distinctively characterized as the metaphysical thesis that mental states have functional properties, or even have them essentially. It is mainly the conceptual (or semantic) thesis – that the concepts of mental states are specified by their functional roles – that distinguishes TT from alternative positions. Thus, Jerry Fodor embraces the metaphysical thesis of functionalism but rejects conceptual, or semantic, functionalism: “[D]enying, as a point of semantics, that “believe” has a functional definition is compatible with asserting, as a point of metaphysics, that belief has a functional essence.” (1998: 8)
2
It is noteworthy that the first statement of TT was in Wilfrid Sellars’s paper “Empiricism and the Philosophy of Mind” (Sellars 1963).
3
I here make the working assumption that the contents of people’s mental-state concepts come in the form of definitions. This assumption is quite controversial, of course, in the philosophical and cognitive science literatures. However, I only make this as a working assumption, and mean to be open to other formats for the contents of concepts (see Goldman
8
Alvin I. Goldman
states are “unobservables” and must be understood by their relations to observables. The psychologists Alison Gopnik and Henry Wellman write: “All these characteristics of theories ought also to apply to children’s understanding of mind, if such understandings are theories of mind. That is, such theories should involve appeal to abstract unobservable entities, with coherent relations among them” (1992: 148, emphasis added). As our discussion of functionalism indicates, TT standardly appeals to specific sorts of publicly observable events to which an understanding of mental states is tied, viz., the subject’s behavior and stimulus conditions in her environment. These are, of course, the sorts of observables to which (logical) behaviorism traditionally appealed, and functionalism is a direct descendant of behaviorism. Thus, TT endorses the idea that an understanding of mental states is founded on appropriate logical and/or epistemological relations to publicly observable peripheral events. TT thereby avoids any reliance on “qualitative” or introspectible properties, which scientific empiricists characteristically regard as either epistemologically suspect, metaphysically dangerous (as a threat to physicalism), or both. Tying mental-state concepts to purely peripheral events effects a purely third-person conception of mentality, a perspective congenial to traditional empiricism because it avoids epistemologically problematic issues of “privacy”. If mentalistic concepts are grasped in terms of peripheral observables and their relata, there follows a natural-seeming answer to question (Q2). People go about the business of attributing mental states by inferring such states from observed peripheral events in accordance with the theoretical connections, i.e., the laws allegedly known to folk attributors. This is the standard TT story about mental attributions to others; and paradigmatic proponents of TT endorse the same story about self-attribution of mental states (Gopnik 1993). So we might define TT in terms of its answers to two questions. In answer to (Q1), it says that mental-state (MS) concepts are defined wholly in terms of their relationships to peripheral observables. In answer to (Q2), it says that beliefs about MSs are formed by making inferences to their occurrence from observations of peripheral events, using the posited nomological relationships. There are several problems with this way of defining TT. Is TT committed to holding that all cases of MS attribution feature nomological inferences of the types in question? Fuller (1995) points out that TT might accept the simulation method of MS attribution as an “epistemological tool”, or shortcut. This possibility would be excluded by the second part of the currently proposed 1993). The general problems of what concepts are and what it means to possess them transcend the scope of this paper. See Peacocke 1992 and Fodor 1998.
Folk Psychology and Mental Concepts
9
definition. One could, of course, simply delete this second element from the definition. The trouble with the resulting proposal is that even ST might accept the first part alone, viz., that MS types are defined in terms of relationships to peripheral events. At least one proponent of ST, viz., Gordon, might be willing to accept this constraint, yet he is no theory-theorist. A different way to configure TT is to drop the second part and replace it with some constraint on the account of concept acquisition. TT might be characterized as holding that the child’s formulation and grasp of MS concepts results from general-purpose scientizing procedures (in this case, processes of concept formation). This characterization would fit with the views of Gopnik and Meltzoff (1997), for example, and Gopnik and Wellman (1992). This definition, however, would exclude developmentalists of the modularist or nativist persuasion (Leslie 1991; Leslie and German 1995; Baron-Cohen 1995), who have typically presented themselves as proponents of TT. These psychologists reject the “child scientist” account of MS-concept acquisition.1 Rationality theory. RT partly derives from Quine’s (1960) articulation of a principle of charity in the context of his approach to “radical translation”. Variants of his charity principle have been endorsed by others, most prominently by Davidson (1984) and Dennett (1987), as an account of how attributors “interpret” others, i.e., assign intentional states to them. The core idea is that attributors make assumptions about the normative propriety of having certain intentional states under various conditions. Rationality is assumed to impose certain requirements on the propositional attitudes a person should have. For example, (1) rationality requires agents to believe truths; (2) it requires their belief-sets to be coherent; (3) it requires their belief-sets to be closed under entailment; and (4) it requires their desires to be aimed at things it is good for them to have. Interpretation proceeds by making the charitable assumption that people usually comply with these normative principles. That is, attributors allegedly assign intentional states on the assumption that their targets mostly conform with these dictates of rationality.2 Thus, according to RT, normative principles of rationality guide the process of intentional attri1
Another interesting question in the definition of TT is whether an approach that drops the peripheralist-empiricist constraint but retains the nomological inference account of attribution should qualify as a version of TT. Consider an account that configures MS-concepts in terms of directly introspectible (or internally detectable) features rather than publicly observable, peripheral evemts. Using these concepts people acquire nomological beliefs (based solely on evidence about their own case) relating introspectible events to one another and to behavior. Using these laws, they make inferences (whether justified or not) to mental states of other people. Would such a position qualify as a specimen of TT?
2
For a reading of Davidson, Lewis (1974), and Dennett as endorsing this approach, see Fodor and LePore 1992: 142-144.
10
Alvin I. Goldman
bution in roughly the way nomological generalizations allegedly guide attribution according to (strict versions of) TT. According to Dennett, for example, to adopt the “intentional stance” toward people (or other systems) consists in assigning intentional states to them in accordance with a default assumption that they conform with principles of rationality. On the surface RT is a theory of attribution, a response to question (Q2).1 Is there also a complementary answer to question (Q1) that RT can offer? What, according to RT, do the folk mean or understand by a desire for q or a belief that p? Adherents of RT are often silent or evasive on this question, but here are two possible approaches. First, RT might portray the folk understanding of these states in terms of the specific rules of rationality it imputes to attributors. For example, attributors might understand a belief that p as a state one ought to be in if one has other beliefs that entail p and one ought not be in if one has other beliefs inconsistent with p. A desire for q might be understood as a state one ought to be in if q is good for one. And so forth. A second possible answer might appeal to the general idea of the intentional strategy. This is suggested by Dennett in the following passage: [A]ll there is to being a true believer is being a system whose behavior is reliably predictable via the intentional strategy, and hence all there is to really and truly believing that p ... is being an intentional system for which p occurs as a belief in the best (most predictive) interpretation. (1987: 29)
A possible implication of this passage is that what the folk understand by a belief is a state that enables an attributor to make reliable predictions of its possessor’s behavior via the intentional strategy. The adequacy of both this approach and the preceding approach to answering (Q1) will be addressed in section III. How is RT related to TT and ST? Presumably, RT contrasts with TT in its emphasis on normativity. Of course, if one adopts the loosest possible definition of TT, i.e., the one cited above from Stich and Nichols 1992, then RT qualifies as a species of TT. But if more stringent conditions are imposed for being an instance of TT, e.g., conditions that exclude normativity from being a defining trait of intentional states, then RT and TT will be mutually exclusive categories. What about the relationship between RT and ST? This relationship is often quite intimate. Some theorists who characterize themselves as simulationists endorse rationality postulates as a core component of MS attribution. Heal 1
That Dennett intends his theory as a theory of how people actually deploy their intentional concepts is clear in the following passage: “Do people actually use this strategy [the strategy of the intentional stance]? Yes, all the time.” (1987: 21)
11
Folk Psychology and Mental Concepts
(1998) is a clear example of this. Coming from the opposite direction, when defenders of RT are pressed on the question of exactly which rationality postulates are utilized, they sometimes fall back into a position indistinguishable from simulationism. I’ll return to this point below. What about ST itself? What are its core commitments? The distinctive element of simulationism is its answer to (Q2)(A), concerning third-person MS attribution. The standard ST position is that one makes attributions by putting oneself in the target’s shoes and modeling her resultant mental activity (Gordon 1986; Goldman 1989). More precisely, one creates “pretend” states of oneself intended to correspond to initial states of the target. Next one feeds these pretend states into a cognitive mechanism of one’s own – e.g., a practical or factual reasoning mechanism – and lets it operate on inputs so as to produce a new state, for example, a decision or a belief. Finally, one attributes this newly outputted state to the target. Some simulationists, however, take exception to this depiction. Specifically, Heal (1998) prefers to characterize the simulation account as holding that the attributor must “co-cognize” with the target, but that co-cognition need not involve feeding pretend states into cognitive mechanisms. Simulationists disagree even more sharply about the method of first-person MS attribution. I endorse an introspectivist or direct monitoring approach to first-person (contemporaneous) MS attribution (Goldman 1993, forthcoming). Gordon rejects such an account and opts for an ascent-routine story (Gordon 1992, 1995, 1996), which will be examined below. Analogous differences persist on the topic of the contents of MS concepts. Thus, simulationists share common ground only on the third-person attribution question, and even on that question, unanimity is far from perfect.
III In this section I shall briefly present some serious problems facing TT and RT. These criticisms do not purport to be definitive – that would be too much to expect for the complexity of the subject and the limits of space. But they are major stumbling blocks for the rivals of ST, which help tilt me and should tilt others toward ST. I begin with problems for RT. The first problem facing RT is its failure to address an entire subclass of the mental states, viz., the sensations. The sensations are surely understood and attributed by the folk; but since there appear to be no questions of “rationality” pertaining to pains or tickles, for example, RT comes up empty in trying to explain how attributors ascribe them.
12
Alvin I. Goldman
A second problem for RT is the implausibility of either of its approaches to attitude concepts. The first approach, recall, is to analyze these concepts in terms of the posited norms of rationality. An initial problem with this approach is a threat of circularity. To say that believing p is a state one ought not be in if one holds other beliefs inconsistent with p is to explain one type of belief-state in terms of other belief-states, which looks flagrantly circular. This might be avoided by positing a recursive specification of beliefs in which the foregoing serves as a recursive clause but other formulas provide base-clauses, e.g., believing p is a state one ought to be in if state-of-affairs p obtains in one’s immediate environment and one is looking in the appropriate direction. The question is whether this recursive approach can provide a sufficiently complete characterization of our attitude concepts. Isn’t there more to our concepts of beliefs and desires than when we ought to have them? A second problem is that, in general, analyzing the concept of an attitude or action in terms of norms for its adoption or performance seems to get things backwards. We ordinarily understand norms for an action of type X only if we independently understand what it is to do X. To grasp a library’s rules for when to return books, one must already understand what a book is and what counts as returning one. It doesn’t seem possible to reverse the order of explanation and explain what it means to return a book solely in terms of rules that specify when to return them. Many theorists, of course, hold that meanings involve norms.1 But the normative approach to meaning poses two problems in this context. First, what attitudes or actions do the norms of meaning govern? Are they propositional attitudes like belief, for example? If so, there is another threat of circularity, because the targets of our theorizing – attitude concepts like belief – are presupposed in the theoretical account of them. Second, even if it is granted that the meanings of concepts in general are to be explained in terms of norms, a second level of norms would have to be invoked to explain the specific meanings of the attitude concepts. This second level is what would distinguish the contents of the attitude concepts from the contents of other (descriptive) concepts such as the concepts of “chair” or “mountain”. It is at this level that my previous worries arise. The second possible RT approach to attitude concepts, floated in the previous section, is to say that the attitudes are understood in terms of their predictive utility when adopting the intentional stance. According to this account, however, possessing the belief concept or the desire concept would require one to possess the concept of the intentional strategy. Is this plausible? 1
For a recent example, see Brandom (1994).
Folk Psychology and Mental Concepts
13
Does everyone who attributes belief or desire (and therefore possesses these concepts) also possess the concept of the intentional strategy? Notice that this concept is a highly reflective one. It requires a possessor not only to be a user of the intentional strategy but to understand, reflectively, what that strategy consists in. But according to developmental psychologists (e.g., Bartsch and Wellman 1995), two-year-olds use the desire concept and three-year-olds use the belief concept. So, according to the present approach, two- and three-yearolds must possess the concept of making-reliable-predictions-via-theintentional-strategy. This consequence is hard to swallow. A different type of problem for RT concerns the specifics of its rationality norms and the questionable grasp of those norms by ordinary people, even children. As mentioned above, quite young children show substantial mastery of attribution skills in their attitude ascriptions. According to RT, then, these children must understand and grasp the canons of rationality that RT postulates. Is it really plausible to suppose that they grasp the general notions of logical consistency and deductive closure? Actually, it is doubtful whether even untrained adults grasp these notions. Many scientific studies of deductive reasoning challenge the notion that untrained adults approach such tasks with abstract semantical or proof-theoretic concepts of the sorts used in formal logic (Cheng and Holyoak 1985, Cosmides 1989). Similarly, psychological studies of decision and choice challenge the notion that naive people utilize standard normative models (Tversky and Kahneman 1986). When challenged by worries of these sorts, Dennett often backs away from the logic-based rules that he and other RT proponents had previously postulated. One should not “cling”, he says, “to the ideals of Intro Logic for one’s model of rationality” (1987: 96). What canons, then, do guide intentional-state attribution? Dennett replies: “When considering what we ought to do, our reflections lead us eventually to a consideration of what we in fact do; this is inescapable, for a catalogue of our considered intuitive judgments on what we ought to do is both a compendium of what we do think, and a shining example ... of how we ought to think” (1987: 98). As Dennett himself immediately recognizes, this response makes it look as if his theory collapses into something similar to ST.1 He briefly tries to distance himself from this threatened collapse (1987: 99-100), but the attempt is
1
Specifically, Dennett worries whether his own view collapses into Stephen Stich’s view, a “projectivist” view very similar to ST that Stich endorsed in the early 1980s (including Stich 1983) but which he has subsequently criticized in many papers.
14
Alvin I. Goldman
not very convincing.1 So the question is: Can RT, without collapsing into ST, identify a set of normative rules of rationality that attributors actually employ? Thus far, the prospects look bleak. I turn now to TT. The first problem I wish to raise for TT is the difficulties it faces in trying to give a plausible account of first-person (current) mental attribution. According to TT, what are the “data” from which a person might infer one of her current mental states? Is it her own behavior, some current external stimulus, or a combination of the two? This seems hopeless as a full account of how people make self-attributions. I currently believe that I have the intention to cook shrimp for dinner tonight. But there is no external stimulus from which this intention state is inferable (I just thought of cooking shrimp without seeing any pictures of shrimp, or the like), and I have not yet taken any behavioral steps from which this intention could be inferred. So how can TT accommodate my epistemic access to this intention state? The problem is daunting when one considers what (strict) TT requires. As formulated in a suitably sophisticated functionalism, the concepts of mentalstate types are concepts of functional roles: dispositions to be caused by specified external events, to causally interact in specified ways with other internal events (which themselves have certain functional roles), and to cause specified behaviors. Strictly speaking, what has functional roles are state-types of a person or organism (see Schiffer 1987: 20-22). So if a person wishes to answer the question, “Am I now in mental-state M?”, she needs to determine whether there is any state token she is currently in which instantiates a type that realizes the functional role associated with “M”. To determine this, she will have to determine the dispositions associated with her current state-tokens, dispositions that are highly complex matters. The dispositions are complex for three reasons, each of which poses epistemic problems (see Goldman 1993). First, the dispositions involve relationships with other events, to which the person may not have current access. For example, they may involve propensities to cause later mental states and/or behavior, but the person will have no current epistemic access to later events. (Of course, later events might be inferred; but such inference is only possible after the current state is type-identified.) Second, functional roles involve subjunctive relationships. They specify interactions that would occur if certain 1
Dennett goes on (1987: 100-101) to consider an even clearer formulation of ST (using the very term “simulation”), and argues that it collapses into TT. His argument rests on the claim that simulation could only work by relying on a theory. But this is incorrect, as I argue elsewhere (Goldman 1989). While simulation could, of course, be “theory driven”, it also could be “process driven”, i.e., driven by psychological processes rather than a theory. The interesting and proper interpretation of ST is that it postulates intensive use of process-driven simulation.
Folk Psychology and Mental Concepts
15
other conditions obtained. For example, having desire D might involve a disposition to form plan P if one has belief B. But perhaps one does not have belief B. How is one to determine whether or not a present state-token would produce P if B, contrary to fact, were present? Third, there is a problem of combinatorial explosion for inferences addressed to propositional attitudes. Each attitude type is associated with a functional role that implicates an indefinitely large number of other attitudes types, each of which is in turn associated with a comparably complex functional role. The upshot for epistemic purposes is an inferential holism posing a severe threat of computational intractability (Goldman 1993). To avert these problems, a more plausible theory would say that selfattribution occurs by internally detecting some properties of mental-state tokens that are (A) categorical (i.e., non-dispositional) and (B) non-relational (at least not massively relational in the way that functional-role properties are). There are two types of candidates for such properties: phenomenological and non-phenomenological properties. The language of thought, for example, presumably has non-phenomenological characteristics that meet conditions (A) and (B), and these characteristics might be the ones directly detected in identifying the content components of mental states.1 In the past (Goldman 1993) I flirted with the idea that the pertinent properties are phenomenological – at least for the non-content components of mental states.2 In the present paper, I mean to remain neutral on this issue. I mean to endorse only the direct detection of some such properties, whether or not they are phenomenological. It is noteworthy that some long-time proponents of TT, viz., Nichols and Stich (forthcoming), now endorse a similar model, thereby abandoning the core TT answer to question (Q2)(B). They call their theory the “Monitoring Mechanism Theory”, and contrast it with TT. However, they don’t seem to appreciate the full implications of this approach. They adopt the standard boxological story of the attitudes, according to which beliefs are depicted as states “residing” in certain boxes. The standard lore on boxes (which they follow) is that box talk is merely short-hand for talk about functional roles. If this is correct, the properties that qualify a state as a belief, a desire, or an intention are not categorical or non-relational. So the Nichols-Stich story does not escape the computational problems mentioned above. What about TT’s approach to third-person attribution? I’ll be quite brief 1
The question of content attribution is greatly complicated by issues concerning externalism about content. There is no room in this paper to take up those issues, which in any case are somewhat tangential to the problems addressed here.
2
My discussion may have encouraged an interpretation under which attitude contents were also supposed to be phenomenologically detectable; but that was not my intention.
16
Alvin I. Goldman
here, and I’ll center my comments around the “frame problem” that Jane Heal (1996) has helpfully introduced in this context.1 Begin with the assumption that adults are pretty proficient at predicting the reasoning of other adults on the persistence or nonpersistence of various states of affairs over time. If a familiar type of change takes place in the world, adults will draw conclusions about what will change and what will remain the same in the next time period. Furthermore, if these are prosaic real-world scenarios, Sandy will be reasonably competent at predicting the conclusions Mandy will draw about such scenarios. Is it plausible that Sandy’s mindreading of Mandy is guided by a theory of such reasoning that Sandy knows? If Sandy really has such knowledge, she is way ahead of, or at least in the same league with, the most astute practitioners of Artificial Intelligence, who have wrestled with this knotty problem without reaching much consensus (Lormand 1999). Workers in AI have tried to handle nonchanges by letting a system assume by default that a state persists unless there is an axiom specifying that it is changed by an occurrence, given surrounding conditions. But new evidence often requires such an assumption to be retracted. When should such retraction take place?2 Ordinary people seem to be able to determine which changes are relevant to which other changes. But how, in general, is such relevance determined? Some theories of reasoning about change implausibly posit the use of enormous numbers of axioms about nonchanges; others avoid this cost but posit implausibly bold axioms about nonchanges. TT implies that ordinary folk have developed (presumably tacitly) an excellent theory about other people’s reasoning about change. That is how TT would explain their substantial success in predicting other people’s inferences about changes and nonchanges. It’s not just that one or two ordinary folk have made this theoretical breakthrough; most ordinary people appear to have done so. However, this seems to beggar belief, given the limited success that apparently brilliant minds have enjoyed (at the explicit level) when addressing the very same problem. To be sure, young children develop theories of a language’s grammar in a way that surpasses what linguists have yet accomplished. But there it is plausible to suppose that a dedicated, special-purpose module is
1
However, Heal’s discussion formulates the frame problem in very general terms, whereas I shall try to stick to its original, somewhat narrower, set of concerns.
2
In the “Yale Shooting Problem,” for example, Hanks and McDermott (1986) explore the following case. Let a system assume by default that (1) live creatures remain alive and (2) loaded guns remain loaded. Confront it with this information: Fred is alive, then a gun is loaded, then, after a delay, the gun is fired at Fred. If assumption (2) is in force through the delay, Fred probably vioates (1). But equally, if assumption (1) is in force after the shooting, the gun probably violates (2). Why is (2) the more natural assumption to enforce?
17
Folk Psychology and Mental Concepts
at work. Is it comparably plausible that there is a special-purpose module for developing theories specifically about the causal reasoning powers of other people? An entirely different account of mindreading in this domain seems far more promising, viz., that when Sandy seeks to predict Mandy’s inferences about changes and nonchanges, she (Sandy) goes through the target inferences on her own and expects Mandy to reach the same conclusions as she does. This, of course, is precisely what ST postulates. Sandy does these reasoning tasks as well as Mandy, and presumably in the very same fashion. If she uses the simulation heuristic to predict Mandy’s results, she doesn’t need a theory about how the task is executed. The same point can be made with a related example: counterfactual reasoning. Philosophers first identified the problem of counterfactual reasoning only in the 1940s or so, and only by 1970 (roughly) were moderately adequate theories developed (e.g., Lewis 1973). Presumably, however, ordinary people living in earlier periods were reasonably competent at predicting and explaining other people’s feats of counterfactual reasoning. Doesn’t it stretch credulity to suppose that they had a folk theory of such reasoning at times when philosophers had little if any inkling that there was even a problem in this territory?
IV In this final section I turn to my own favored approach, ST, and to some slight specifics of its configuration, especially as pertains to question (Q1). There is broad agreement among simulationists as to how ST answers (Q2)(A), the question of third-person attribution. There is far less agreement about how it should answer (Q1) or (Q2)(B). It seems natural, moreover, for positions on the latter two questions to be intertwined. For example, if one accepts the direct detection account of first-person (current) attribution, it is not unnatural to suppose that mental-state concepts involve some sort of categorical features or characteristics that are epistemically identifiable via direct, internal detection. This is the kind of view I have advocated elsewhere (Goldman 1993, forthcoming). This view, with its strong Cartesian or semi-Cartesian flavor, contrasts with another major tradition in philosophy of mind, rooted in the writings of Wittgenstein, which places overt behavior at the center of all comprehension of the mind. Another simulationist, Robert Gordon, takes a very different tack from mine, clearly along behaviorist lines. Gordon (1996, unpublished) contends
18
Alvin I. Goldman
that self-attribution relies, not on introspection or inner detection, but on what he calls ascent routines. The way adults ordinarily determine whether or not they believe that p, he says, is simply to ask themselves the question whether or not p. If someone were to ask me whether I believe Mickey Mouse has a tail, I would ask myself “Does Mickey Mouse have a tail?”. If my answer to this latter question is “yes”, then my presumptive answer to the beliefquestion will also be “yes”. Gordon calls this an “ascent routine” because it answers a question at the mental level by answering another question pitched at what he considers a “lower” level (Gordon 1996: 15). He claims that this enables one to answer a seemingly “internal” question by means of anwering an “outward-looking” question (unpublished: 3). To find out about one’s own mental states, he thinks, one does not consult “internal” information but “external” information. For example, when I tell the clerk, “I want two of these and three of those, “it is very likely that I do want possession of the items I point to. To attain reliability with regard to what I want, I don’t look inward, I look outward at the display case. (Unpublished: 3.)
I find several problems with Gordon’s ascent-routine approach. First, how does he hope to generalize from cases of belief to other propositional attitudes? In the case of belief, one substitutes for the question, “Do I believe p?” the question “Is p true?”. But what is the story for the other propositional attitudes? If I ask myself the question, “Do I want two croissants from the display case?”, what (“outward-looking”) question am I to substitute for that? Perhaps Gordon will reply. “Are the two croissants in the display case appetizing-looking?” But there are several problems here. First, how am I to tell whether they are appetizing looking – to me – without (tacitly) consulting an internal state of mine? Second, consider two different questions I might ask about my attitude toward the croissants: “Do I want two of the crossants in the display case?” and “Do I believe there are two croissants in the display case?” In both cases, I might be assisted in answering the question by looking at the display case, but the difference between what I need to know in the two cases is not available from the outward object, only from my mental condition. Third, consulting the display case only seems relevant to making up one’s mind what to believe or desire, not to determining an attitude already possessed. Fourth, what about other mental attitudes? Which outward-looking questions should I substitute for the following questions, “Do I now intend to bring it about that p?”, “Am I now wondering whether p?”, “Am I now recalling p?”, and “Am I now imagining that p?”? A different kind of problem with Gordon’s approach is that there should
Folk Psychology and Mental Concepts
19
be a tight connection or resemblance between information relevant to questions about past mental states and information relevant to present mental states. If Gordon is right that questions about present first-person mental states can be answered by information about outward objects, then questions about past first-person mental states should also be answered by information about outward objects. But that is surely wrong. If I ask myself, “Did I intend to bring about p (at such-and-such a time)?” or “Was I then imagining that p?”, I need not retrieve anything about outward objects to answer these questions. I can just recall my (internal) mental state at the time. This discussion confirms my earlier proposal, viz., that the story of firstperson present mental-state attribution should assign a salient role to the direct inner detection of some sorts of internal features or characteristics. If we now try to apply this result to question (Q1) on our list, it looks as if the concepts of mental states should somehow be composed of the relevant sorts of internally detectable properties that would account for first-person attribution. This proposal immediately runs into trouble on both empirical and philosophical grounds. There is empirical evidence, which fits with suggestions of Wittgenstein, that infants (under one year) have a sensitivity to certain patterns of movement associated with intentionality, agency, or goal-directedness. According to Premack (1990) and Gergeley et al. (1995), the core notion of intentionality is triggered in young children by perceptual patterns suggestive of self-propelled motion or motion with respect to a potential goal. BaronCohen (1995) seconds Premack’s idea by positing an “intentionality detector”, alleged to be an innate endowment for reading mental states off of behavior (1995: 32). As described by Baron-Cohen, the intentionality detector interprets almost anything that propels itself or makes a non-random sound as an agent with goals and desires (1995: 34).1 But if goal detection is stimulated by external motion or behavior of certain types, then the concepts of purposiveness or goal-possession do not, after all, seem to consist in inwardly detectable features. I shall approach this problem, in an admittedly speculative fashion, by advancing a dual-representation hypothesis about MS concepts. The focus will be on goals, desires, or plans, though similar proposals might apply to other 1
Some aspects of these ideas are challenged by a study of Woodward (1999). Woodward found that 9-month-olds differentiated between movements of grasping a toy and movements of merely letting the back of one’s hand drop onto a toy. In other words, by nine months of age infants selectively encode certain goal-related properties and not others; and this selective encoding appears to begin at roughly five months. So infants don’t seem to attribute goaldirectedness indiscriminately to any motion of a self-propelled entity. However, these sorts of qualifications still leave in place the fundamentally behavioral character of the cues used for third-person goal-state detection.
20
Alvin I. Goldman
types of MSs. The rough idea is that people develop two sorts of mental representations for at least some MSs. Representations associated with behavioral characteristics are an initial entry-way into a conception of desire and intentionality; but a full conception of desire and intentionality also involves representations associated with internal characteristics (of a phenomenological or sub-phenomenological variety). In tendering this hypothesis, I do not mean to propose that people have one type of representation for third-person desires and a different type of representation for first-person desires. That would yield the traditional puzzle in philosophy of mind of how people can conceptualize both themselves and others as having the same sorts of states. Rather, people coordinate representations of certain behavior and representations of certain inner features as representations of one and the same (sort of) state. My story is intended to allow for changes and development in the grasp of mentality at different (early) ages, but the changes I have in mind do not coincide with those standardly discussed in the developmental literature. The idea of dual, or multiple, representations of a single type of object, state, or category is quite common in cognitive science. For example, people might represent a single sortal in terms of both shape and function; and they may deploy representations utilizing different cognitive codes or modalities. Thus, Biederman (1987) proposes that visual object-classification commonly proceeds by means of shape-coded object types. In visually identifying something as a piano or a lamp, one exploits a stored model or prototype of what pianos or lamps look like. In addition, there are separate, modality-neutral codes representing pianos and lamps in terms of their functions, e.g., “keyboard musical instrument” and “portable source of (artificial) light”. Some pairs of matching representations are especially striking. Meltzoff and Moore (1983) discovered that infants as young as forty-two minutes can imitate the facial gestures of another person. These newborns apparently represent their own facial movements proprioceptively – after all, they have not previously seen their own faces. They produce facial movements of their own that mimic those of a target, so they must somehow match representations of their own movements to representations of the target’s facial gestures (Meltzoff 1999).1 But since they don’t see their own movements, it must be proprioceptive representations of their own movements that they link to visual representations of the target’s movements. Meltzoff and Moore therefore postulate an innate cross-modal matching between visual and proprioceptive
1
In this case, of course, the infants’ “matching” representations are not representations of numerically the same action, only similar actions (their own and that of the observed target).
Folk Psychology and Mental Concepts
21
representations of the same movements.1 In a somewhat analogous fashion, I conjecture that children learn to match inner features detectable in their own goal or desire experiences with behavioral cues utilized in identifying goals or desires. In principle, this could be achieved by noting their own goal-driven behavior and its regular association with these inner features. But this would leave us with the traditional question of why they should think that other people ever undergo analogous inner experiences. This would be a particularly vexing problem for ST, because ST postulates that third-person attributors make an inference from the output states of their own mental simulations to corresponding inner states of their targets. Why should they think that other people undergo such internal states at all? Philosophers of mind have shied away from internal features in their theorizing precisely because of this looming problem of other minds. To develop my conjecture further, I now turn to the recently discovered phenomenon of so-called “mirror neurons”.2 An interesting class of premotor neurons were initially discovered in macaque monkeys, neurons that discharge both when the monkey performs an action and when it observes another individual making a similar action (Rizzolatti et al. 1988). The discharge of these neurons is not correlated with individual movements, but rather with general types of actions, especially grasping, holding, tearing, poking, and manipulating. So these neurons seem to code motor representations of goaldirected actions involving objects. Mirror-neuron activity, then, is apparently a neural correlate of plans or goals for action. Mirror neurons are also triggered, however, when an observer watches a target agent act toward a goal. This correspondence, or “mirroring”, between observed and executed actions is sometimes described in terms of the metaphor of physical “resonance” (Rizzolatti et al., in press). It is as if neurons in certain motor areas start to “resonate” with their cousins as soon as the appropriate visual input is presented. This resonance phenomenon is not restricted to monkeys, but is found in humans as well, as confirmed by transcranial magnetic stimulation (TMS), MEG/EEG recordings, and brain imaging techniques. Experiments demonstrate that motor centers of adult humans resonate during movement observation. For example, Fadiga et al. (1995) stimulated the left motor cortex of 1
Elisabeth Pacherie (1998) argues that the cross-modal matching must be between motor and visual representations rather than proprioceptive and visual representations. This fits the hypothesis I shall advance below even better.
2
The reader is reminded that I am not addressing the traditional problems of normative epistemology in the mental domain. I am not asking what confers justification on people’s beliefs about other minds. I seek only a descriptive account of the generation of their beliefs, which is what the study of folk psychology, as here construed, aspires to provide.
22
Alvin I. Goldman
normal subjects using TMS while they were observing arm movements and hand grasping movements performed by the experimenter. Motor evoked potentials were recorded from various arm and hand muscles. A selective increase of motor evoked potentials was found in those muscles that the subjects normally use for producing the observed movements. One can make casual observations of the same underlying mechanisms in everyday life. When you see a cyclist about to have an accident, your own muscles tense in what feels like the way they would tense if you were on the bicycle about to have an accident.1 Findings in clinical populations further support these ideas. Lhermitte et al. (1986) document a phenomenon of “imitation behavior”, in which patients with prefrontal lesions compulsively imitate gestures or even complex actions performed in front of them by an experimenter. This behavior is explained as arising from an impairment of inhibitory control that normally governs motor plans. Apparently, when observing someone else perform an action, a normal human generates a “resonant” plan or image of doing that action himself. This plan is normally inhibited so that it does not yield motor output; but such inhibition is impaired in the relevant patient population. Vittorio Gallese and I (Gallese and Goldman 1998) have cited resonance phenomena as possible evidence for, or a precursor of, mental simulation. Here I mean to invoke them for the twin purposes sketched above: (1) to explain how there could be a “matching” between internal and external representations of (what is conceived of as) the same state, and (2) to explain how a child might come to interpret others as undergoing internal experiences of desiring or planning similar to her own. It is easy to see how mirror-neuron (MN) activity could facilitate the establishment of a correspondence between internal and behavioral formats for representing mental states. MN activity involves inner events associated with goal possession, desire, or planning. Externally triggered MN activity involves the observation of behavior and environmental circumstances (i.e., the presence of a goal object) associated with goal possession. So when one undergoes externally triggered MN activity, there is a (roughly) simultaneous occurrence of observed behavior and detectable inner events that might come to be linked or associated with one another. To be sure, the observed behavior is the behavior of another creature. How might a learner come to link certain inner experience in himself with the behavior of another? The learner is in a position to notice that the same sorts of inner experience occurs when he himself prepares to execute an action. This is because MN activity serves the primary function (along with other neuron 1
Thanks to Bill Child for this example.
Folk Psychology and Mental Concepts
23
groups in the premotor cortex) of preparing to execute an action. When MN activity is externally driven, the action is not actually executed; it is inhibited. But the appropriate sort of action is executed by the person being observed. So the learner might get the idea – perhaps is even “hard-wired” to get the idea – that inner events of the type he undergoes while watching the other’s behavior are also undergone by the observed actor. This would be no more remarkable than the matching of visual and proprioceptive (or motor) representations of behavior by Meltzoff and Moore’s neonates. Although the dual-representation hypothesis I am floating may be appealing, some might argue that it constitutes a major concession, or even capitulation, to TT. If behavioral features are admitted into the set of representations associated with desire and goal-possession, doesn’t this amount to belated agreement with TT on the proper answer to question (Q1)? This isn’t so, for two reasons. First, only a limited number of mental-state types, I suspect, have relatively simple and tight connections with stereotypical behavior. The story being told here about the concepts of desire, intention, or goal-possession probably does not generalize to all mental-state concepts (e.g., belief). Second, nothing in the story sketched above suggests that children acquire any farflung theory of systematic mental-state connectedness of the sort postulated, for example, by functionalism. And if they do develop a grasp of mentalstate/mental-state connections, this may only be with the help of internally detected features, which does not fit the standard mold of the TT approach. So although the dual-representation answer to (Q1) takes notice of some of the motivations behind traditional versions of TT, it keeps a considerable distance from all of the detailed versions of TT that have actually been proposed.1
References Baron-Cohen, S. (1995). Mindblindness. Cambridge, MA: MIT Press. Bartsch, K. and Wellman, H. (1995). Children Talk about the Mind. New York: Oxford University Press. Biederman, I. (1987). Recognition by components: a theory of human image understanding. Psychological Review 94: 115-147. Brandom, R. (1994). Making It Explicit. Cambridge, MA: Harvard University Press. Cheng, P. and Holyoak, K. (1985). Pragmatic reasoning schemas. Cognitive Psychology 17: 391416. Churchland, P. M. (1988). Mind and Consciousness. Cambridge, MA: MIT Press. Cosmides, L. (1989). The logic of social exchange: Has natural selection shaped how humans reason? Studies with the Wason selection task. Cognition 31: 187-276.
1
I am grateful to Elisabeth Pacherie and Brad Thompson for very helpful comments on this paper.
24
Alvin I. Goldman
Davidson, D. (1984). Inquiries into Truth and Interpretation. Oxford: Oxford University Press. Dennett, D. (1987). The Intentional Stance. Cambridge, MA: MIT Press. Fadiga, L., Fogassi, L., Pavesi, G., and Rizzolatti, G. (1995). Motor facilitation during action observation: A magnetic stimulation study. Journal of Neurophysiology 73: 2608-2611. Fodor, J. (1985). Fodor’s guide to mental representation: The intelligent auntie’s vade-mecum. Mind 94: 55-97. Fodor, J. (1998). Concepts: Where Cognitive Science Went Wrong. Oxford: Oxford University Press. Fodor, J. and LePore, E. (1992). Holism: A Shopper’s Guide. Cambridge, MA: Blackwell. Fuller, G. (1995). Simulation and psychological concepts. In M. Davies and T. Stone, eds., Mental Simulation. Oxford: Blackwell. Gallese, V. and Goldman, A. (1998). Mirror neurons and the simulation theory of mind-reading. Trends in Cognitive Sciences 2: 493-501. Gergeley, G. et al. (1995). Taking the intentional stance at 12 months of age. Cognition 56: 165193. Goldman, A. (1986). Epistemology and Cognition. Cambridge, MA: Harvard University Press. Goldman, A. (1989). Interpretation psychologized. Mind and Language 4: 161-185. Goldman, A. (1993). The psychology of folk psychology. Behavioral and Brain Sciences 16: 15-28. Goldman, A. (forthcoming). The mentalizing folk. In D. Sperber, ed., Metarepresentation. New York: Oxford University Press. Gopnik, A. (1993). How we know our minds: The illusion of first-person knowledge. Behavioral and Brain Sciences 16: 1-14. Gopnik, A. and Meltzoff, A. (1997). Words, thoughts and theories. Cambridge, MA: MIT Press. Gopnik, A. and Wellman, H. M. (1992). Why the child’s theory of mind really is a theory. Mind and Language 7: 145-171. Gordon, R. (1986). Folk psychology as simulation. Mind and Language 1: 158-171. Gordon, R. (1995). Simulation without introspection or inference from me to you. In M. Davies and T. Stone, eds., Mental Simulation. Oxford: Blackwell. Gordon, R. (1996). ‘Radical’ simulationism. In P. Carruthers and P. Smith, eds., Theories of Theories of Mind. Cambridge: Cambridge University Press. Gordon, R. (unpublished). Self-ascription: Jonesian conditioning vs. ascent routines. Hanks, S. and McDermott, D. (1986). Default reasoning, non-monotonic logic, and the frame problem. Proceedings of the American Association for Artificial Intelligence 328-333. Heal, J. (1986). Replication and functionalism. In J. Butterfield, ed., Language, Mind, and Logic. Cambridge: Cambridge University Press. Heal, J. (1996). Simulation, theory, and content. In P. Carruthers and P. Smith, eds., Theories of Theories of Mind. Cambridge: Cambridge University Press. Heal, J. (1998). Co-cognition and off-line simulation. Mind and Language 13: 477-498. Leslie, A. (1991). The theory of mind impairment in autism: Evidence for a modular mechanism of development? In A. Whiten, ed., Natural Theories of Mind. Oxford: Blackwell. Leslie, A. and German, T. (1995). Knowledge and ability in “theory of mind”: One-eyed overview of a debate. In M. Davies and T. Stone, eds., Mental Simulation. Oxford: Blackwell. Lewis, D. (1972). Psychophysical and theoretical identifications. Australasian Journal of Philosophy 50: 249-258. Lewis, D. (1973). Counterfactuals. Oxford: Blackwell. Lewis, D. (1974). Radical interpretation. Synthese 23: 331-344. Lhermitte, F., Pilon, B. and Serdaru, M. (1986). Human autonomy and the frontal lobes: I.
Folk Psychology and Mental Concepts
25
Imitation and utilization behavior: a neuropsychological study of 75 patients. Annals of Neurology 19: 326-334. Lormand, E. (1999). Frame problem. In R. Wilson and F. Keil, eds., MIT Encyclopedia of the Cognitive Sciences 326-327. Meltzoff, A. (1999). Imitation. In R. Wilson and F. Keil, eds., MIT Encyclopedia of the Cognitive Sciences. Cambridge, MA: MIT Press. Meltzoff, A. and Moore, M. (1983). Newborn infants imitate adult facial gestures. Child Development 54: 702-709. Nichols, S. and Stich, S. (forthcoming). Reading one’s own mind: A cognitive theory of selfawareness. Pacherie, E. (1998). Representations motrices, imitation et theorie de l’esprit. In H. Grivois and J. Proust, eds., Subjectivite et Conscience d’Agir: Approches Cognitive et Clinique de la Psychose. Paris: Presses Universitaires de France. Peacocke, C. (1992). A Study of Concepts. Cambridge, MA: MIT Press. Premack, D. (1990). The infant’s theory of self-propelled objects. Cognition 36: 1-16. Quine, W. V. (1960). Word and Object. Cambridge, MA: MIT Press. Rizzolatti, G. et al. (1988). Functional organization of inferior area 6 in the macaque monkey: II. Area F5 and the control of distal movements. Experimental Brain Research 71: 491-507. Rizzolatti, G., Fadiga, L. Fogassi, L., and Gallese, V. (in press). From mirror neurons to imitation: Facts and speculations. In W. Prinz and A. Meltzoff, eds., The Imitative Mind: Development, Evolution and Brain Bases. Cambridge: Cambridge University Press. Schiffer, S. (1987). Remnants of Meaning. Cambridge, MA: MIT Press. Sellars, W. (1963). Empiricism and the philosophy of mind. In Science, Perception, and Reality. New York: Humanities Press. Stich, S. (1983). From Folk Psychology to Cognitive Science. Cambridge, MA: MIT Press. Stich, S. and Nichols, S. (1992). Folk psychology: Simulation or tacit theory? Mind and Language 7: 35-71. Tversky, A. and Kahneman, D. (1986). Rational choice and the framing of decisions. Journal of Business 59: 251-278. Woodward, A. (1999). Infants’ ability to distinguish between purposeful and non-purposeful behaviors. Infant Behavior and Development 22: 145-160.
26
Philip Pettit
PHILIP PETTIT How the Folk Understand Folk Psychology Abstract Let folk psychology consist in the network of concepts, and associated beliefs, in terms of which we make sense of minded performance. This paper addresses the question of how we, the folk, come to understand those concepts: this, as distinct from the separate question as to how we come to apply them in the interpretation of particular minds, our own and those of others. The argument is that even though the network of concepts is akin to a set of theoretical, interdefined terms, still it is possible to explain how we, the folk, understand them without suggesting that we are proto-scientists. The understanding required can be based on a sort of know-how: that is, a practical, untheoretical, form of knowledge.
1. Background According to analytical functionalism, each type of intentional state is identified by its entry and exit conditions – by the conditions that trigger its appearance and disappearance in the normal subject – and by the way it connects with other states in generating further states and in leading to action. Thus the belief that p is characterised by its role in being sensitive to perceptual evidence for or against p, at least under normal conditions, and in connecting with other beliefs and desires to produce further results: in connecting, say, with the belief that if p then q to generate the further belief that q. According to analytical functionalism what makes a state an instance of believing that p is precisely that it plays that sort of role.1 Even if we reject analytical functionalism as a complete account of intentional states, it is plausible that it catches part of the truth about such states. For whatever else is needed to make a state a case of believing that p, it must typically display the sort of profile invoked by functionalists. Why should we think of a state-type as an instance of believing that p, unless in normal circumstances – under some plausible specification of what these are (Pettit forthcoming) – the normal subject would enter the state in face of perceptual
1
The belief might also be taken, though I think less plausibly, as the role-state rather than the realiser-state (see Jackson 1998; Jackson and Pettit 1988).
How the Folk Understand Folk Psychology
27
evidence that p, would exit the state in face of perceptual evidence that not p, and would perform otherwise in appropriate ways? I am going to assume here that in this sense analytical functionalism captures at least part of the truth about intentional states of belief and desire and the like. The assumption commits me to the view that when we, the folk, explain ourselves and one another in terms of intentional states, then we are playing a theoretical sort of game (Braddon-Mitchell and Jackson 1996). We posit statetypes of a kind that transcend observational recognition and that are each defined, so the assumption has it, by the network of connections whereby it is linked to entry and exit conditions and to other states and actions. We understand terms like ‘the belief that p’ and ‘the desire that r’, then, just so far as a shared background of belief – a shared theory – tells us what to expect of any states deserving those names. The expressions pick out types of intentional states after the broad pattern in which theoretical terms in science are often alleged to pick out theoretical entities (Lewis 1983). As the term ‘electron’ picks out something that is defined by its role in physical theory – it is the sort of particle that orbits the nucleus of an atom, that is negatively charged, etc. – so ‘the belief that p’ picks out a type of state that is defined, at least in part, by its role in the psychological economy of the normal subject. The view that we the folk are theorists of mind, however, has come in for some hard criticism. And with apparent reason. Robert Gordon (1992) is surely right when he tells us that we do not ‘regard other human beings as “black boxes” and theorise about their behavior control centers’. We are too unthinking a species, it seems, to count as theorists of mind; we do not have the scientific or proto-scientific cast that theorising requires. What to conclude, then? Does the observation that ordinary folk are generally an unthinking lot make nonsense of the theory-theory: the theory that we, the folk, are applied theorists in the area of human psychology? Or can the observation be reconciled with theory-theory? There are really two questions here. The first is how we the folk can be held to understand a general theory of mind, given we show so little evidence of theoretical focus. And the other is how we the folk can be held to endorse and apply the theory – to behave in the fashion, so it seems, of proto-scientists – in our everyday construals of one another, and indeed of ourselves. In this paper I will deal only with the first question.1 I want to try and show how it can be plausible to credit us, as we use psychological terms, with understanding a general theory of mind. 1
Victoria Mc Geer and I deal with this second question in a forthcoming paper on “The Self-regulating Mind”.
28
Philip Pettit
I defend a conciliationist line, arguing that the thesis that we the folk understand a general theory of mind is consistent without our not being theorists in the sense which Gordon has in mind. One way of describing the line in the paper is to say that we, the folk, are theorists only in a latent or implicit or tacit sense. But of course there are many ways of construing latency in epistemic matters – see, for example, Davies (1987) – and the paper is distinguished by the particular construal for which it argues (Jackson and Pettit 1995; Pettit 1998).
2. The Problem Beliefs and desires are defined and distinguished, at least in part, by the sorts of connections that they enjoy, under normal conditions, with one another, with behaviour, and with the perceptual world. So how can we the folk – we, the untheoretical folk – identify and discriminate such state-types without going in for theoretical definition? How can we know what the belief-that-p statetype is, for example – how can we present it to ourselves as a potentially explanatory state – without rehearsing or even being able to rehearse the litany of connections by which it is apparently defined? The problem is exacerbated by the fact that the sets of connections whereby any particular belief-role, or indeed any other intentional role, is distinguished will typically implicate other roles, as analytical functionalism insists. The roles will be holistically networked, so that no one set of connections can be properly identified without the identification of others as well. We cannot understand what it is to be a husband without understanding what it is to be a wife. And similarly we cannot understand what it is to have one belief without understanding what it is to have other beliefs and what it is to instantiate other intentional states: what it is to perceive things, to desire things, to act so as to fulfill those desires, and so on. Theory-understanding would not raise a problem, notice, if we ordinary folk gained discriminating knowledge of each intentional state-type just on the basis of observation. We may grasp what type of colour red is, or what type of shape square is – we may have adequate discriminating knowledge of the entities in question (Evans 1982) – just on the basis of seeing how they look and being able to recognise them. And there would be no problem to resolve if we had the ability to see what makes a state a case of believing that p on a similar observational and recognitional basis. But it will be granted on all sides that intentional state-types are not observationally discriminable in this way. And so the problem is to explain how we can effectively discriminate them
How the Folk Understand Folk Psychology
29
other than by resort to postulational definitions of the kind generally employed in scientific theorising.
3. Towards a Response This last comment is suggestive. It directs us to the fact that if the problem looks insoluble, that is because we endorse the common assumption that there are only two ways of knowing something and of discriminating it from relevant alternatives: by observation, as in the paradigm ruled out; or by description, as in the case of defining it theoretically by its connections. But this assumption, I now want to say, is false. And once we recognise this we can see our way towards a response to the problem. Suppose that a question arises as to whether I am acquainted with a certain complicated route across a city or landscape. Let us assume that there is nothing about how such a route looks that could make it observationally recognisable or salient to me, whether from a bird’s eye point of view or from the point of view of a traveller: this, perhaps, because it is very complex. And let us assume that the route might is not presented to me – is not made a matter of suitably discriminating knowledge – by means of abstract specification or description. The starting point and the other decision points along the way are not described for me in a set of instructions, nor are they encoded in a route map at my disposal. Could I be acquainted nevertheless with the route in question? Is it possible that in the absence of discrimination by observation, and short of discrimination by more abstract specification, I could yet be familiar with the route? Is there any way in which I could present it to myself as a route that is distinguishable from all relevant alternatives, any way in which I could know ‘which route it is’ (Evans 1982)? A little reflection suggests that there is. Suppose that I am disposed, in the event of wanting to get from start to finish, to form an inclination as I come to each decision-point to go on in a certain direction. Suppose that those inclinations are reliable indicators of how to go, at least when things are normal or favourable in an appropriate sense. Suppose, in short, that I defeasibly know how to get from the start to the finish, though only in a way that does not enable me to describe the route, either in abstract instructions or by means of a map. I only have practical knowledge – I have pure practical knowledge – of the route in question. Even though I know how to get from start to finish, I lack an independent capacity to describe or diagram what to do at each of the decision-points on the route.
30
Philip Pettit
In such a case it should be clear that I can still discriminate the route from relevant alternatives. If I recognise the know-how that I possess – if I do better in that respect than the home-seeking donkey – then I can present the route to myself as the route that corresponds to that know-how. Being alerted to the know-how in question, I can say or think: oh, that route, where the route in question is revealed in my navigational disposition; it is the route such that in normal operation my navigational disposition will provide me with any information that may be required to identify it. In the scenario envisaged I won’t be able to present the route to myself, all at once, as a discriminable entity. I won’t be able to present it quite in the way that I would be able to do so if I could recognise it observationally, in the manner of the colour red, or if I could identify it in the manner of a defined entity by a set of instructions or a map. Enfranchising the navigational disposition as a source of relevant information – provisionally certifying it as a relevant form of know-how – I will think of the route ostensively, as this or that route. But the mode of presentation will be promissory in character. I will not see straight away the difference between the route and every relevant alternative, as I would be able to do on the basis of observation, or a set of instructions, or a mapping. But for any question raised as to what distinguishes the route – as to where it goes at this or that point, for example – I will be able to rely on my know-how to tell me. I will be able to situate myself at that point and say then where the route leads. Or I will be able to situate myself at the corresponding point in a mock-up of the situation and say how the route goes. Or I will be able to imagine myself at the point and say which direction I should take on the basis of purely mental simulation.1 The lesson of this example, which may itself be a little strained, is that there is a third way of discriminating things – a third way of having discriminating knowledge of them – apart from the usual observational and descriptional modes. It is possible to have discrimination by know-how as well as discrimination by observation and descrimination by description. Take anything that corresponds to a certain conscious disposition – a conscious form of knowhow – on the part of those who are to be aquainted with the entity. Using whatever tags or examples that serve to cue them, we can lead those who are to
1
Such simulation may enable me to work out descriptions of how to go at each point, and so a set of instructions for the route as a whole; or it may enable me in a parallel way to construct a map. But the idea is that even if I manage to do this – and perhaps help others at the same time – the mode in which I will continue to discriminate the route for myself will be by means of the know-how, not the instructions or the map; I may be incapable of making use of those media, for example – they may just produce confusion in me – while continuing to know the route in the other sense.
How the Folk Understand Folk Psychology
31
be acquainted with the entity to think: oh, that entity, where the item in question generally answers to their disposition and the manifestations of the disposition generally supply reliable information about it. The subjects can think of the item ostensively but the basis of the ostension is practical rather than observational or descriptional.1
4. The Response This lesson impacts immediately on the problem we raised about the mode in which intentional states like the belief that p, or the desire that r, can be presented to us ordinary folk. I will draw out the lesson with reference to belief states, leaving it to be understood in the case of desires and other intentional states. In the geographical example there are two elements required for suitable presentation of the route: those who are to be acquainted with the route must have a generally reliable navigational disposition and they must enfranchise that disposition, in the sense of treating it as a reliable, if defeasible, source of information. By analogy, two parallel elements will enable us ordinary folk to have a corresponding knowledge of, say, the belief-that-p state. Suppose that I am generally disposed to form certain further dispositions in face of those inputs that should lead to believing that p according to the full or partial version of analytical functionalism assumed here. Suppose that I am generally disposed to drop those dispositions in face of the inputs that should lead to no longer believing that p. And suppose that the dispositions in question are those that full or partial functionalism associates with believing that p: they involve the dispositions, at least when conditions are normal, to respond in a distinctive manner to certain new evidence or new beliefs and to go about satisfying various desires in an equally characteristic fashion. Suppose, in short, that I instantiate the dispositions associated with being at least a potential believer that p: I believe that p or I am such that presented with suitable evidence I will come to believe that p. The pattern of disposition that I display in such a case – let us call it, the belief-that-p disposition – corresponds to the belief-that-p role in the same way that my navigational disposition corresponds in the earlier case to the geographical route. 1
I have argued elsewhere that we discriminate rules as entities that we can then try to follow, on the basis of just such a practical – I call it, ‘ethocentric’ – mode of discrimination; certain paradigms trigger a disposition and we pick out the rule as the counterpart of that disposition, assuming that the disposition is not affected by factors that come out as disadvantages when we try to negotiate convergence with others. See Pettit (1996), Chapter 2 and see also below on our access to common meanings.
32
Philip Pettit
If the enfranchisement of the navigational disposition enables me to present the route to myself, secure in (defeasible) knowledge of which route it is, then by analogy it would seem that just by enfranchising the belief-that-p pattern of disposition – just by taking it as a generally reliable, if defeasible, indicator of the rational connections that define the belief-that-p role – I can present that role to myself. I can know which role it is and I can know enough, therefore, to be able to assign the appropriate semantic value to the expression, ‘the belief that p’. It appears that as I can be acquainted with the route via my navigational ability so I can be acquainted with the belief-that-p role via the corresponding ability in this case. There will be a disanalogy between the cases, of course, so far as the beliefthat-p disposition is going to be present in someone only so far as they also instantiate other dispositions – that is the networking point which analytical functionalists emphasise – whereas the navigational disposition can be realised entirely on its own. But even this disanalogy may be exaggerated. For we might reasonably imagine that the sort of navigational disposition envisaged is itself networked with other dispositions and that someone will be able to get from A to B only so far as they are equally able to navigate the criss-crossing routes C to D, A to C, A to D, and so on. It might be, for example, that the crucial landmarks that trigger the disposition at various points on the A-B route are going to be recognisable only to someone who can use them equally on other routes. The ability of London taxi drivers to find their way around their complex city used to be described as ‘the knowledge’. There is every reason to think that having the knowledge in regard to one route inevitably goes with having the knowledge in regard to a whole network of routes. But do we ordinary folk possess dispositions of the kind required; and if we do, are we in a position to enfranchise them as sources of information about what is required of any state that is to count as a case of believing that p, or believing that q, or whatever? Yes, clearly we do. I am possessed of a disposition that is worthy of the relevant epistemic franchise to the extent that I have practical know-how in matters of reasoning: to the extent that I know how to reason in regard to p, in regard to q, and so on. We human beings do not just form and unform beliefs in a passive manner, with our brains supplying us automatically with the updatings imposed by experience and practice. That may be how it is with other animals and with current forms of artifical intelligance. But we human beings do more than just house a stream of belief; we perform an intentional activity, itself based on belief and desire, whose aim is to increase the chances, perhaps inter alia, of our beliefs being true rather than false (Pettit 1996, Ch.2). This is the activity of reasoning about questions like whether or not it is the case that p, or that q,
How the Folk Understand Folk Psychology
33
or whatever. We think about what perceptual evidence would support or undermine the claim that p, and see whether or not it obtains; we think about what things are consistent with the fact that p, what inconsistent, and reflect on whether we already believe or are disposed to believe any such things; and so on. And in doing this we display a distinctive and conscious form of knowhow: a conscious, practical knowledge of how we should be guided by matters of perception, and matters of actual or prospective belief, in determining whether or not it is the case that p. The know-how involved in being able to reason in regard to p gives me the ability – the, no doubt, defeasible ability – to say, for any possible evidential or belief scenario we care to envisage, whether or not it is likely that p in that world. And that means in turn that it enables me, defeasibly, to say for any such scenario whether a person who reasons perfectly in regard to p will come to believe that p there. But the person who reasons perfectly in regard to p will presumably come to form the belief that any rational believer – any believer, reasoning or not, whose beliefs come and go as functional specifications require – would form in the situation on hand: otherwise reasoning would not serve its point. And so the know-how involved in being able to reason in regard to p will give me the ability to say how a rational believer will respond to any possible evidential or belief scenario in regard to believing that p. In short, it will enable me, however defeasibly, to identify the entry, exit and performance conditions of any state that is to count in the rational believer as a case of believing that p. The way in which reasoning know-how will enable me to identify those conditions, of course, may be as purely practical as the way in which navigational know-how enables me to identify the twists and turns of a complex geographical route. I may be no good at giving abstract specifications of the entry, exit and performance conditions of the belief that p or at representing them in any general mode. But I may be consciously able to say for any relevant set of perceptual inputs, and concurrent intentional states, how the rational believer will respond in regard to p. I may be able to do that through imagining myself in the presence of such inputs or states – those of us who believe in philosophical thought-experiments think we can do this – but I will certainly be able to do it if I actually find myself in their presence. The emerging picture should be clear and in some respects familiar (Davidson 1984; Dennett 1979; Dennett 1987). As one of the ordinary folk I will not be able to discriminate a state-type such as that of believing that p either on the basis of observation or on the basis of abstract description and theory. But I can have discriminating knowledge of it on the basis of my practical know-how in reasoning as to whether or not p. I can think of it as
34
Philip Pettit
that state of belief, the one that is foreshadowed in my p-reasoning ability. I can pick it out as a determinate state-type, confident of being able to answer, case by case, any question that may be posed as to how it is distinguished from other beliefs in this or that situation. I can pick it out in just the same promissory way that we imagined me picking out the complex route in the geographical analogy. Of course I cannot know how to reason in regard to p without knowing how to reason in regard to other contents as well. In that respect knowing how to reason has the holistic character of the navigational knowledge that we attributed to the London taxi-driver. And if practical, reasoning knowledge is holistic then that will explain how, as analytical functionalism predicts, my knowing which type of state is a case of believing that p will go hand in hand with knowing at the same time which types of state are cases of believing that q, that r, and so on. Possessed of a networked pattern of practical knowledge of how to reason, I will be able to discriminate the different state-types in the networked pattern of beliefs that analytical functionalists take folk psychology to postulate. In this section I set out to explain how we untheoretical folk can present ourselves with a model of the believing-desiring subject in which the different types of belief and other states required by that model are effectively discriminated. Only if this is explicable, will it possible to say that the selection of folk-psychology as a model of explanation is an untheoretical sort of choice, not one involving scientific or even proto-scientific postulation and definition. The points I have made about the practical basis on which different belief state-types are available to us, the folk – and available, in particular, as explanatory posits – should be sufficient to discharge the task taken on.
5. A Challenge But the position outlined in the last section will truly vindicate the claim that the folk have an untheoretical way understanding intentional states only if it really is the case, as I suggested, that reasoning ability need be nothing more than a practical skill. It must be possible for a person to have reasoning ability – and so to have the practical know-how required for identifying belief-states – without having the ability to spell out and endorse the principles that govern that reasoning. A person must be able to reason according to modus ponens without being able to articulate or understand an articulation of the principle of modus ponens. And a person must be able to reason according to less formal principles – for example, according to the principle that relates being red to
How the Folk Understand Folk Psychology
35
looking red in normal circumstances – without being able to spell out or understand that principle. Is the ability to reason a suitably practical ability, then? I believe it is. Apart from the articulate or judgmental way of believing things – apart from the way in which we believe that p when we can assert that p and can live up to that assertion in our behaviour – there is a distinct, practical way in which we must be held to believe other propositions (Pettit 1998a). And this mode of believing is exemplified in the way we endorse principles of reasoning of precisely the sort that we have just been discussing. Consider a principle like modus ponens or indeed like the colour principle mentioned. All of us must be said to believe such principles since they govern how we reason and they dictate the sorts of beliefs – judgmental beliefs – that we come to form in different situations; they have an influence on our behaviour, in other words, of just the sort that we associate with belief. Yet we clearly do not have to believe those principles in the judgmental sense, since we may not be able to understand the words required for their articulation: we may not be able to understand the use of schematic letters involved in the ordinary way of expressing modus ponens, for example, and we may not understand the notion of normal circumstances required for articulating the colour principle. How then do we believe such principles? Do we live up to the principles in the minimal sense of acting as if they were true, for example, and believe them in that sense (Stalnaker 1984)? Yes, but that means nothing, since the principles are necessary truths and it is impossible to fail to act as if they were true. The important thing is that we live up to the principles in a richer sense, involving a contingent accomplishment. We live up to them in the sense that we do actually have the ability to make judgments in regard to potential premises and conclusions – something that simple believers like animals may presumably lack – and the requirements to which we prove responsive in the exercise of that ability are summed up in the principles. We prove responsive to those requirements as we spontaneously form judgments and, perhaps even more importantly, as we take account of difficulties and objections raised by others. We may never form an articulate judgment in regard to modus ponens or the colour principle but still those principles capture a pattern in matters to which we prove responsive in belief-formation. It used to be traditional to distinguish two ways of believing a universal proposition. People may believe that all dogs chase cats in sensu composito, as when they form an articulate judgment to that effect and live up to it in their behaviour. Or they may believe it in sensu diviso, as when they form only particular judgments but those judgments display a certain, robust pattern: for
36
Philip Pettit
every judgment that an animal is a cat, they form the paired judgment that it chases cats. The sense in which people may have a merely practical belief in modus ponens or the colour principle parallels this sense of belief in sensu diviso. For the idea is that people do not articulate the principle and then live up to it, in the manner of judgmental belief, but rather that their particular judgments display a pattern that the principle encodes. And that is precisely what happens with belief in sensu diviso. People do not articulate and then live up to the universal proposition but they form particular judgments which display a pattern that the universal proposition spells out; each judgment bears on a particular dog and each attributes to the dog a disposition to chase cats. It isn’t just a plausible thesis that there is a sense of practical belief that is weaker than judgmental belief, yet involves more than the behavioural belief involved in acting as if something is the case. Lewis Carroll’s fable of Achilles and the Tortoise shows that there are always some principles of reasoning that we believe only in this practical sense. Achilles believes he can force the Tortoise to admit that Z, given that he admits two premises, A and B, which logically imply Z. But the Tortoise insists that he needs more than A and B as premises; he also needs C: if A and B are true, Z must be true. Very well, says Achilles: you admit A and B and C, so now you must admit Z. No, replies the Tortoise, for I must also endorse another premise, D: if A and B and C are true, Z must be true. The lesson is that in order to reason from certain judgmentally believed premises to a judgmentally believed conclusion, you have to believe in the merely practical way that such premises support such a conclusion; or if you take that as itself a judgmentally believed premise, then you have to believe in the purely practical way that the new premises support the conclusion; and so on. It turns out then that the challenge raised for the position sketched in the last section can be readily answered. The practical know-how whereby I can find my way along a certain route may be untheoretical enough to vindicate the claim that I have an untheoretical understanding of the route is in question. And the practical know-how whereby I can reason from one proposition to another may be untheoretical in sufficient measure to vindicate the claim that I have an untheoretical understanding of belief states and, more generally, of the posits of intentional psychology. It is perfectly appropriate, then, to attribute an understanding of intentional states – an understanding of intentional psychology – to ordinary, untheoretical people, because such an understanding can be achieved without anything much in the way of theoretical thought. The theory that we ascribe to ordinary people when we speak of a folk psychology is a theory adumbrated in ordinary practice, particularly in the practice of reasoning. It is truly
How the Folk Understand Folk Psychology
37
a theory, and people do truly apply that theory in making sense of themselves and one another. But it is a theory mastered on the basis of practical knowhow, not something requiring scientific conjecture. It is a lived theory, a Lebenstheorie,1 to which we spontaneously and unthinkingly commit ourselves in learning the vocabulary of mind.2
References Braddon-Mitchell, David, and Frank Jackson. 1996. Philosophy of Mind and Cognition. Oxford: Blackwell. Burge, Tyler. 1979. “Individualism and the Mental”. Midwest Studies in Philosophy 4: 73-121. Davidson, Donald. 1984. Inquiries into Truth & Interpretation. Oxford: Oxford University Press. Davies, Martin. 1987. “Tacit knowledge and Semantic Theory: Can a Five Per Cent Difference Matter?”. Mind 96: 441-62. Davies, Martin. 1994. “The Mental Simulation Debate”. In Objectivity, Simulation and the Unity of Consciousness, edited by C. Peacocke. Oxford: Oxford University Press. Dennett, Daniel. 1979. Brainstorms. Brighton: Harvester Press. Dennett, Daniel. 1987. The Intentional Stance. Cambridge, Mass.: MIT Press. Evans, Gareth. 1982. The Varieties of Reference. Oxford: Oxford University Press. Geach, Peter. 1957. Mental Acts. London: Routledge. Gordon, Robert. 1992. “The Simulation Theory: Objections and Misconceptions”. Mind and Language 17: 11-34. Heal, Jane. 1994. “Simulation versus Theory Theory: What is the Difference?”. In Objectivity, Simulation and the Unity of Consciousness, edited by C. Peacocke. Oxford: Oxford University Press. Jackson, Frank. 1998. From Metaphysics to Ethics: A Defence of Conceptual Analysis. Oxford: Oxford University Press. Jackson, Frank, and Philip Pettit. 1988. “Functionalism and Broad Content”. Mind 97: 381-400. Jackson, Frank, and Philip Pettit. 1995. “Moral Functionalism and Moral Motivation”. Philosophical Quarterly 45: 20-40. Lewis, David. 1969. Convention. Cambridge, Mass.: Harvard University Press. Lewis, David. 1983. Philosophical Papers Vol 1. Oxford: Oxford University Press. Lewis, David. 1996. “Elusive Knowledge”. Australasian Journal of Philosophy 74:549-67. Locke, John. 1975. An Essay Concerning Human Understanding. Edited by P. H. Nidditch. Oxford: Oxford University Press.
1
I mean the use of this word, and the association with the notion of Lebenswelt, to signal a connection between the line defended in this paper and lines of thought that can be found in a number of German traditions: for example, in the Erlanger School tradition of constructivist analysis of concepts.
2
My thanks to those who provided comments on an earlier version of this paper at a workshop organised in Flinders University in December 1997, and at seminars in the Australian National University in May 1998 and the University of Konstanz in June 1998. I am particularly grateful to Martin Davies and Philip Gerrans, who provided detailed feedback on that paper, and to Dan Dennett, Michael Esfeld, Andrew Gleeson, Richard Holton, Peter Menzies, Juergen Mittelstrasse and Nick Zangwill who commented on it.
38
Philip Pettit
Pettit, Philip. 1995. “The Virtual Reality of Homo Economicus”. Monist 78:308-29. Pettit, Philip. 1996. The Common Mind: An Essay on Psychology, Society and Politics, paperback edition,. New York: Oxford University Press. Pettit, Philip. 1998. “Practical Belief and Philosophical Theory”. Australasian Journal of Philosophy 76. Pettit, Philip. “A Theory of Normal and Ideal Conditions”, Philosophical Studies, 1999, Vol 96, pp. 21-44 Pettit, Philip, and Michael Smith. 1996. “Freedom in Belief and Desire”. Journal of Philosophy 93:429-49. Stalnaker, Robert C. 1984. Inquiry. Cambridge, Mass.: MIT Press.
Understanding Other Minds from the Inside
39
JANE HEAL Understanding Other Minds from the Inside Abstract We find it natural to say that creatures with minds can (sometimes at least) be understood ‘from the inside’. The paper explores what could be meant by this attractive but, on reflection, somewhat mysterious idea. It suggests that it may find a hospitable placement, which makes its content and appeal clearer, in one version of the so-called ‘simulation theory’ approach to grasp of psychological concepts. Simulation theory suggests that ability to use imagination in rethinking others’ thoughts and in recreating their trains of reasoning is central to our grasp and use of psychological concepts. On this view to think of another’s mind is not to think of some intricate quasi-mechanical assemblage of items in the other’s head which causally explain her behavior. If this is all that the ‘inside’ of another person, i.e. her mind, were like, then there would be no question of anything being ‘from’ it. The simulation view, however, emphasizes that thoughts essentially have content and that identifying another’s thought, and working out its possible effects, involves identifying its content and oneself entertaining thoughts with the same content. So, on this approach, to think of another’s mind is to think of a complex but rationally unified set of thoughts, a set which is conceived as had by one subject but where the contents and relations can be grasped and appreciated by another. Some of these thoughts will be indexical and the whole can thus be said to constitute the subject’s point of view on the world, both literally and metaphorically. Grasping this point of view is, the paper suggests, what is meant by speaking of ‘understanding from the inside’.
Can we understand other minds ‘from the inside’? What would this mean? There is an attraction which many have felt in the idea that creatures with minds, people (and perhaps animals) invite a kind of understanding which inanimate objects such as rocks, plants and machines do not invite and that it is appropriate to seek to understand them ‘from the inside’. What I hope to do in this paper is to introduce and defend one version of the so-called ‘simulation’ approach to our grasp and use of psychological concepts, a version which gives central importance to the idea of shared rationality, and in so doing to
40
Jane Heal
tease out and defend one strand in the complex of ideas which finds expression in this mysterious phrase.1 Let us here recap the salient ideas of the simulation approach.2 Simulationism is to be contrasted with another approach to philosophy of mind which has, at least among Anglo-American analytic philosophers, been the dominant one of the last decades and which has also been an important influence on psychologists and cognitive scientists. We may call this familiar alternative the theory theory. The version best known to philosophers is functionalism in philosophy of mind. This says that to grasp psychological notions is to grasp that there are certain inner states of persons, which are typically caused by such and such external events, which interact among themselves to cause further inner states and events and which finally combine to cause behaviour.3 To possess the concept of some particular mental state is to grasp the particular causal-explanatory role associated with that state. When we use our understanding of psychological notions, for example in predicting what another will think or do, we deploy this theoretical knowledge. There are many ways of spelling out this general idea, depending on what view we take of how we acquire and represent this supposed theoretical knowledge – whether it is innate or learnt, whether explicitly or tacitly known etc. But let us leave all these issues on one side, just noting however that the theory theory does not seem at all hospitable to the ‘from the inside’ idea. Indeed part of its motivation is to find an account of the psychological which is naturalistic, i.e. which does away with certain deeply suspect forms of dualism and sees human beings as part of the natural order. Theory theory does this precisely by claiming the similarity of psychological concepts to nonpsychological concepts such as those of natural science, presenting the former merely as particular complex and interesting cases of the general style of thought invoked in the latter. So much for theory theory. Now for a thumbnail sketch of its rival, the simulation approach. This is by no means an entirely new idea. A version of it goes back to Vico in the early eighteenth century; it gets a passing mention in
1
Talk of persons ‘having a point of view’ and of there being such a thing as ‘what it is like to be that person’ are also part of the same set of ideas. But I would like to stress that the whole issue of the existence of ‘qualia’ is not touched on at all in what follows.
2
For more on this see the two collections edited by M. Davies and T. Stone Folk Psychology and Mental Simulation, (both Oxford: Blackwell, 1995) and Theories of Theories of Mind, edited by P Carruthers and P. K. Smith, (Cambridge: Cambridge University Press, 1996).
3
For a collection which includes many of the classic papers arguing for this view see Readings in the Philosophy of Psychology, Vol 1, ed. N. Block (Cambridge, MA: Harvard University Press, 1980).
Understanding Other Minds from the Inside
41
Kant; it is associated with Dilthey and is forcefully defended by Collingwood.1 And under the name ‘Verstehen’ one broadly simulation-style approach is familiar, and has been extensively debated, in the philosophy of history and social science. But in the last ten years the idea has been revived in the context of psychology and philosophy of mind. And here it provides a new perspective on a great number of familiar topics – for example the nature of imagination, the differences between practical and theoretical reasoning, the nature of emotion as well as initiating an interesting body of empirical work in psychology and suggesting new models in cognitive science.2 I shall touch on only a very small part of this. A way of putting the central idea of the simulation approach is this. When we think about another’s thoughts or actions we somehow ingeniously exploit the fact that we ourselves are or have minds. What we do is to make our own mind in some way like the mind of the one we seek to predict or understand. We simulate his or her thoughts, we recreate in ourselves some parallel to his or her thought processes. Many simulationists further articulate this by talking of my having ‘pretend beliefs’ and of my ‘inference mechanisms’ being run ‘off-line’. But others (and here I include myself) would prefer to use more everyday vocabulary and to talk of my using my imagination and of my thereby entertaining the same thoughts and making the same inferences as the others. We shall come to the significant differences here in due course. I shall not here consider in detail the reasons for preferring a simulation approach to a theory theory one. Let me just indicate one central and immediately apparent advantage of simulationism. It is this. Others’ thoughts are very varied and numerous and interact with each other in countless different ways. The remarkable thing is how successfully we deal with this, correctly adjusting our expectations of others’ thoughts, feelings and actions in an immense variety of circumstances. Clearly any theory adequate to systematise our competence here would itself be immensely complex. But simulation can explain our competence without crediting us with knowledge of any such vast, and very probably unwieldly, body of information. Rather in thinking about another’s thoughts, in order for example to predict his or her intentions, we harness our own cognitive apparatus and make it work in parallel with that of the other and then use the result we arrive at to ground our prediction. And for this to occur all that is required is, first, that we have cognitive apparatus 1
I. Berlin, Vico and Herder (London: Hogarth Press, 1976); I. Kant, Critique of Pure Reason, trans. Norman Kemp Smith (London: Macmillan, 1953), p. 336; W. Dilthey, Selected Writings, H. P. Rickman ed. (Cambridge: Cambridge University Press, 1976); R. G. Collingwood, The Idea of History, (Oxford: Oxford University Press, 1946), esp pp. 282-302.
2
These themes are all illustrated in the collections mentioned above in footnote 2 p. 40.
42
Jane Heal
which is sufficiently similar to that of the other to produce usefully similar results and, secondly, that we can make it work in a parallel way. It is not required that we have some representation of the apparatus itself or its workings. We do not need to possess a ‘know that’ about the processes of thinking, what thoughts lead to what others and so forth, provided that we can harness relevantly our own ‘know how’ of doing the thinking itself and can thus follow through in ourselves the same train of thought as the other has pursued. The economy of the proposal is striking. The phrase ‘thinking about others’ thoughts’ covers a great variety of importantly different kinds of reflection which we now need to distinguish. There is something very importantly right in the overall picture painted by functionalism, namely the facts which it highlights that psychological states may be caused by events in the world, that such states interact with each other to give rise to further states and that they may give rise to bodily behaviour. This gives us a useful framework for considering the different sorts of issue which may arise for me concerning another’s thoughts. 1. I may wonder what effect the circumstances around her will have on the psychological states of another person. (E.g. She is being whirled round on a fairground ride; will she feel sick? There is a disturbance in the corner; will she notice it?) Let us call the connections I focus on here ‘world-mind links’. 2. I may wonder what further thoughts she will have, given some thoughts about which I already know. (E.g. She believes thus and such about the cash flow of our firm; will that lead her to think that we are about to go bankrupt? She endorses these and those principles: what decision will she reach in this particular case?) Let us call these ‘mind-mind’ or ‘intrapsychic’ links. 3. Thirdly I may wonder what behaviour, i.e. what actual bodily movements, she will exhibit, given her thoughts. (E.g She will hear a balloon popped behind her: will she jump? She intends to smash her opponent’s ball away to the side line: will she succeed in jumping high enough to get the needed angle?) Let us call these ‘mind-world’ links. But let us note also that there is a fourth thing I may be doing under the general heading ‘thinking about another’s thoughts’. 4. I may try to work back from the behaviour she has produced to a view about the psychological states from which the behaviour arose. (E.g. She pulled a funny face; was she really amused? She said that such and such; was she annoyed?) These are four extremely different contexts in which psychological concepts are used; and competence at each of them may well call upon different aspects of the skill which is grasp of those concepts. We should beware of lumping them all together and supposing that a philosophical account of our
Understanding Other Minds from the Inside
43
competence with such concepts, whether simulationist or theory theory, should say the same about each. And the claim I want to make about simulationism is that it is particularly at home, its strengths and plausibility particularly apparent, in the second of the listed circumstances, i.e. in an account of our grip of mind-mind or intrapsychic links. And this will be mainly what I shall discuss below. I do not think that simulationism has anything distinctive to say about our ability to answer the third sort of question. It does have distinctive ideas to contribute on the first and fourth, but I shall not discuss them in detail here. However, a few remarks about the fourth may help to ward off some misunderstandings. It is important to note how, on any view, the fourth context – that of interpreting and explaining behaviour – must be a very different matter from the others. All philosophers, whatever their theory of mind, acknowledge that many alternative explanations of the same behaviour are possible. For example, even if we can identify something with fair confidence as an intentional raising of the arm, when we move to identify the purposes behind the raising, and to the feelings, goals, beliefs which in turn lie behind that purpose, it is clear that many accounts could be given. Even on a functionalist or theory theory view there is no such thing as just ‘applying the theory’ in some fixed and algorithmic way to derive an interpretation. In the other cases (those falling under 1-3 above), as conceived by functionalism, there is such a fixed and algorithmic procedure: if one has sufficient information about the prior conditions then one just has to identify the bits of the theory which deal with those conditions and apply them. Ones prediction may be hedged because one knows that further information may reveal the case as more complex than it at first appears or as requiring adjustments in the light of subsidiary theoretical principles. But (these kinds of complications aside) forward-moving theory-invoking prediction is quite a different matter from backward-looking theoryinvoking explanation. An account of how we do the latter cannot just call on ‘our knowledge of the theory’ but must also tell some story about how we generate a range of possible explanations compatible with the theory and how we assess them. The same general kind of point needs to be made in connection with simulationism. Even if we accept a simulationist account of how prediction about others’ thoughts or behaviour are arrived at (e.g. in cases of type 2 above), this does not of itself tell us how backward-looking interpretations and explanations are arrived at. So we should beware of the idea that simulationism is the proposal that mere awareness of another person – of his or her circumstances and behaviour automatically produces in the observer, via some natural sensitivity, a simulation of the other’s mental state.
44
Jane Heal
Simulationism ought not to be the claim that we have some kind of quick route to knowledge of other minds, or that we can empathetically ‘tune in’ to others, or anything of this kind. Perhaps such a thing exists in a few basic cases. But patently other people are often difficult to understand; often we know that we are ignorant of their thoughts and feelings or we have little confidence in our conjectures about what they may be. Simulationism is not the promise of some easy answer to these difficulties.1 Let us now turn to consideration of the distinctively simulationist story about mind-mind links, i.e. about how I might come to some prediction or further belief about another’s thoughts on the basis of knowledge of some subset of her thoughts. To take a very schematic case, suppose that the other believes that p1-pn and is interested in whether or not q. How might I work out her likely opinion on whether or not q? Theory theory of course says that I, so to speak, look up what the theory tells me about what is the likely upshot of the combination of believing p1-pn with an interest in whether q. I look up the relevant axioms about beliefs and interests of that kind and apply them to this particular case; so it is by applying my knowledge about thoughts and their effects that I work out what to expect. Simulationism will say something different. But there are two contrasted ways in which the simulationist story can be told. One story starts with a picture of the mind which is very congenial to the theory theorist and derived to a considerable extent from cognitive science. The mind, on this picture, consists of a number of subsystems which perform various functions. For example there are two stores in which beliefs and desires are kept; there are various processors which produce beliefs and put them in the belief store; these include a sensory analysing system, which takes sensory inputs and transforms them into beliefs; they include also some inference meachanisms which take beliefs and derive other beliefs from them; there is also a practical reasoning system which takes beliefs and desires as input and produces intentions as output; and so forth. Each processing system is designed to accept certain kinds of input; receiving input of the appropriate kind causes it to go through its distinctive evolutions and to produce output of distinctive kinds. These inputs and outputs – sensory states, beliefs, desires, intentions and so 1
Some simulation theorists do postulate a natural tendency to ‘catch’ others’ mental states. For example in normal infants we find very early a disposition to attend to what others attend to, to be frightened if adults in their company are frightened and the like. It is extremely plausible that we do have some such basic pattern of response and that this is central to our ability to under stand other minds. The point I am emphasising, however, is that this does not take us very far. How we build on it to arrive at interpretations in the more complex cases is something about which simulationists have some proposals but not fully worked out ideas. See the papers by Gordon, Goldman and Heal in the Davies and Stone collections.
Understanding Other Minds from the Inside
45
forth – are realised or coded in vehicles which are, in fact, brain states, for example neuronal patterns described at some suitable level of abstraction. And on this picture what really drives the evolution of the inference mechanisms, practical reasoning system and the like are the intrinsic properties of the vehicles, the brain states or neuronal patterns, which are the beliefs, desires etc. Given this view of what goes on in the mind, simulationism is now spelt out in the following way. Suppose, as in the schematic example above, that I wish to work out what the other, is likely to think about whether or not q, given that she believe that p1-pn, when I myself do not share her beliefs. What I do is first construct some ‘pretend’ beliefs that p1-pn. These are items which do not, in my mental architecture, play the role of beliefs; they do not come from my belief store. Nevertheless they are like beliefs in the nature of the vehicle in which they are coded. I now take my inference mechanisms ‘offline’ – that is I detach them from their usual links with my belief store. I feed in the pretend beliefs I have constructed, at the same time making some adjustment to the mechanisms to make them search for q-relevant consequences; I then wait to see what the mechanisms produce as output. If they output a pretend belief that q then I attribute to the other the belief that q. Of course I do not do all this consciously. Nevertheless this is what is going on at the level of the operations of my subpersonal cognitive machinery. Simulation theory presented in these terms is conceived of as an empirical hypothesis. Those who articulate it like this suppose that it has empirical consequences different from those of the theory theory, that we can already see what those consequences are and that we can set about testing them. But I would like to suggest that there are considerable problems with this conceptualisation of the issue. Consider first the fact that we do not have any well-backed-up and detailed view about what kind of functional ‘systems’ are to be found in the brain or of how the various kinds of mental state and process recognised in commonsense are in fact implemented at the sort of level envisaged. Many kinds of architecture are imaginable other than the one sketched above. For example, is it necessary to distinguish between theoretical and practical reasoning in the way proposed? To insist that we should is to make substantive and controversial philosophical assumptions about the relation of belief, desire and value. Another, and for our purposes more important, question is whether we have to take it that ‘inference mechanisms’ operate on beliefs, i.e. the whole complex state including both content and the attitude to it. A different articulation would take it that ‘inference mechanisms’ operate on mental representations minus their attitude determiners on, so to speak, ‘thought radicals’. There is surely some case for thinking that we can reason with representations which we do not believe. How do we explain what
46
Jane Heal
we are doing in arguing by reductio ad absurdum or reasoning hypothetically if every piece of reasoning needs a belief as a starting point? But if we thus reconceptualise the ‘inference mechanisms’ as operating on thought radicals, then simulationism, formulated in terms which presuppose the existence of inference mechanisms operating on beliefs, turns out to involve a false picture of the mind and so to be worthless. Consider also the fact that (even supposing that the original sketch of the architecture of our cognitive machinery is the right one) we have very little idea of what would be involved, neurophysiologically or functionally, in taking a system ‘off-line’. We do not know what features of operation would remain the same and what would be different. It may be that we are seduced by the image of a machine made of cogs, levers and pistons, where we can make sense of things like disengaging the gears, detaching the drive belt and so forth. But is the brain like that in any sense? Who knows? I am not here seeking to make difficulties for the whole project of seeking to understand the mind by breaking up its overall operation into various different functions and looking for the biological structures and processes which subserve those functions. Good luck to the cognitive scientists, psychologists and neuroanatomists and so forth who grapple with these fascinating and difficult tasks. The issue is rather that we do not yet have enough grip in how that project might actually work out in detail to have any confidence that we are working in terms of the right architecture when we talk of ‘pretend beliefs’ and ‘off-line running’. Nor do we have enough understanding of how that proposal could be implemented to see what the talk of ‘off-line’ running would actually amount to. The latter point means that we do not really know how to test simulationism, regarded as this empirical hypothesis. The former means that the simulationism, if it is articulated in terms of this particular architecture, is made hostage to future discoveries in brain science and might, given unfavourable developments there, turn out to be a total mistake. But it seems that the simulation hypothesis has considerable plausibility quite independent of any empirical developments in brain science, a plausibility noted by Kant, Dilthey, Collingwood and others who were not at all in the business of speculative cognitive science or high level neuroanatomy. This suggests that there ought to be a reading of the simulation proposal in which it is articulated in quite different terms, terms which place it nearer the a priori end of the spectrum and on which it is effectively insulated from how things turn out in neuroanatomy and the like. Let us also note that the idea of ‘off-line’ use of inference mechanisms and the like does not offer any particularly congenial setting for the idea of ‘understanding from the inside’. The attraction of the idiom is not at all illuminated
Understanding Other Minds from the Inside
47
by the simulationist story as spelled out above. If I wish to predict how another person will react to some new supposed cholesterol-lowering medication I may try to find out its effects on her by taking a dose myself and observing the results. Or (indulging in some science fiction) I might be able to unhook a part of my circulatory system and run an experiment on that. In either case I could ‘simulate’ in myself the operation of the drug on her. But the fact that it is a bit of my own bodily apparatus which is being run in experimental fashion gives no special insight ‘from the inside’ into the workings of the drug. And we have been told nothing which entitles us to think the case of the mind – i.e. the brain – to be any different. But the idea of simulation did seem to have some resonance with the idea of ‘understanding from the inside’. So again we are led to the thought that there may be an alternative way of conceptualising the idea. So now let me sketch such an alternative. Consider a normal person who is capable of having beliefs about a certain subject matter, i.e. of forming them appropriately and reasoning from them appropriately, among other things. Let us take Charles as an example; he is an investment expert and can form the belief that the base rate will rise on seeing evidence that it will and can make sensible inferences from this, for example to a fall in the value of shares. Now we take it entirely for granted that if Charles is capable of doing these things then he is also capable of reasoning hypothetically about what would happen if base rate were to rise. It is difficult to make any sense of the opposite supposition. Remember that Charles is a normal human being, so that in dealing with most subject matters, cups of tea, rain, buying a house and so forth, Charles can cope with both actual and hypothetical. Suppose that we now try to graft on the supposition that, for example, when faced with sentences beginning ‘Suppose that base rate were to rise ...’ Charles goes deaf, or berserk or in some other way just fails to cope, although he does respond normally to the straight assertion ‘Base rate has risen’. Or suppose we try to add on the idea that Charles can appreciate the need for contingency planning in connection with most kinds of events but never seems to indulge in any kind of contingency planning about base rate rises, although, remember, he copes with great competence when they actually occur. Can we really fill in the details of such a story in a coherent way? I do not say that it is provably impossible that we should do so. We are familiar with the extremely bizarre and disconcerting way in which what are normally treated a unitary abilities can unravel in the case of brain damage and disease (agnosias, aphasias and the like). But the interesting point is that such cases are extremely rare and that our ordinary psychological concepts do not allow for them. Our ordinary concepts are, quite properly I suggest, tailored to the
48
Jane Heal
outward, behavioural contours of the normal case, to the kinds of successful performances and achievements one can regularly expect of persons. They are not tailored to respect or record the structure of the machinery which realises those abilities. In our thinking about other people one fundamental question we can and often do ask is what subject matters they are familiar with, i.e. roughly what concepts they possess and in what kind of detail. Do they understand about tables and chairs? About royalty? About snow? About car engines? About income tax? And how well do they understand about each? If a person is familiar with a subject matter and understands it to some roughly indicated level then we take it for granted that this ability to think about the subject matter will manifest itself just as much in coming to counterfactual beliefs, in considering possibilities and their upshots, as it does in forming and reasoning from categorical beliefs. The ability will show itself also in desires, intentions, emotions, dreams and fantasies. Competence in thinking about a subject matter is a multifaceted ability. It is an error, a distortion of our central psychological notions, to think of concept possession as something which shows up only or centrally in the formation of categorical beliefs. Rather, belief formation is just one facet of an ability which naturally manifests itself also in other kinds of thinking. Note here a further important point, implicit in what has been said already but needing emphasis. A parallelism between certain psychological processes is already presupposed in the everyday conception we have been spelling out, namely a parallelism between, on the one hand, the inferences a person makes with categorical beliefs in virtue of his or her grasp of a subject matter and, on the other, the counterfactual conditional beliefs he or she would form as a result of making suppositions and the like. So Charles infers from ‘Base rate will rise’ to ‘Share prices will fall.’ But it is also the case that if he wonders ‘What if base rate were to rise?’ then he will come to the conditional belief ‘If base rate were to rise then share prices will fall.’ This parallelism must stay more or less in place on pain of our losing our right to describe the content of Charles’ wondering as ‘What if base rate were to rise?’ It cannot be base rate and its possible rise that he is wondering about if he does not at this point come up with the same idea, to figure in the consequent of his conditional belief, as he would come up with in a straight belief to belief inference. The fact that this parallelism exists is what makes viable the whole conceptual structure we use in talking of others’ thoughts, plans, desires, reasonings etc. It is the idea of the multi-faceted ability which is, in effect, the idea that the same content can be identified as playing a role embedded within other contents and as the object of various different attitudes.
Understanding Other Minds from the Inside
49
Someone might here offer a hypothesis about how it is that we have such a multifaceted ability, i.e. about the nature of the systems or devices in which the ability is realised. Perhaps what goes on when I wonder ‘What if p?’ is that I take some inference mechanisms ‘off-line’ and feel in a pretend belief that p?1 But to pursue the line of thought I am proposing we do not need to get embroiled in issues like this at all. The parallelism between thinking about what is taken to be actual (having a belief) and thinking about what is taken to be merely possible (wondering, hypothesising, imagining and the like) exists, whatever its underpinnings turn out to be. And we are entitled to invoke it in our account of thought about other minds. So now back again to simulation and other minds. We can now present the simulation hypothesis like this: ability to think about another’s thoughts, e.g. to reason from the existence of those thoughts to conclusions about the existence of further thoughts, is an extension or redeployment of ability to think about the subject matter of the other’s thoughts. How does this work? Let us take the following way of spelling things out. Let us revert to our schematic example in which the other believes that p1-pn and is interested in whether or not q. I know this and I am interested in whether or not she comes to believe that q. What she will do is wonder ‘In the light of p1-pn is it the case that q?’ i.e. she will direct her thought to answering the question whether q, having in mind the evidence that p1-pn. If the propositions that p1-pn imply that q, and she comes to be aware of them as so doing, then she will come to believe that q, taking this to be a belief to which she is entitled, in the light of the facts (as she sees them) that p1-pn. What will I do? If I share her beliefs I may, in effect, pose myself just the same question, viz. ‘In the light of the facts that p1-pn is it the case that q?’ But if I do not share her beliefs then the question I should address is, rather, ‘If it were that p1-pn would it be that q?’ But in either case the other person and I share a central aim, namely trying to get a sense of the relations of implication or otherwise between p1-pn and q. And we carry out this aim by exercising our ability to think about the subject matter of p1-pn and q. And if comes to seem to me that if it were that p1-pn then it would be that q, then I attribute to the other the belief that q. Let us now reflect on the concepts implicit in the story I have just sketched. I have spoken of us as having ‘a sense of’ some thoughts as implying or being implied by others. Much recent philosophy, influenced perhaps by cognitive science models, tacitly operates with a picture of the progress of 1
See, e.g., the paper by Nichols et al. in Theories of Theories of Mind, ed. Carruthers and Smith, referred to in footnote 2 above.
50
Jane Heal
thought through time, when a person is reasoning and reflecting, as a matter of there being one thought (perhaps quite a complex one) occupying the conscious mind at one instant and of its being entirely replaced by another thought at the succeeding moment. So, for a schematic example, at first I think ‘p, if p then q, q?’ and then this complex thought is swept away and replaced by ‘q’. But this is surely a distortion of our experience as thinkers. A slightly more accurate narrative is one in which I first think ‘p, if p then q, q?’ and then next think ‘Well, clearly q, since p and if p then q’. That is, I judge that q in a light of a sense of it as following from p and if p then q. Relatedly I take this latter belief of mine to be justified by my beliefs that p and that if p then q. So far only beliefs have been considered. But the above is a structure which we find in numerous intrapsychic connections, for example between desires, intentions and emotions (or at least some important aspects of them) and other contentful states. So my fear of something consists, at least in part and in central cases, of my taking it to be dangerous and threatening. But when I so take it, it is in the light of my belief that it may explode or may bite. So my fear, insofar as it is to be identified with taking the feared thing to be dangerous, is experienced by me as justified or appropriate in the light of other thoughts. Similarly I may take a resolution to perform an action in the light of that action seeming to me to be advantageous and to have no drawbacks. Again it is not just that first I think about the advantages and lack of drawbacks and then the next instant these considerations are entirely swept away from my conscious mind and replaced by the thought ‘I’ll do it!’ Rather, the ensuing thought is more like ‘I’ll do it (since it is advantageous and has no drawbacks).’ And what goes for me goes for others, on the account of the use of psychological concepts which I am sketching. We do not think of others primarily or solely as extremely complex biological machines with many physical structures inside interacting in elegant ways; in thinking of a person as a person, these aspects of human existence are not to the fore. Of course there is complexity in others’ psychological states and this complexity is implicated in temporal development which it is quite proper to think of as causal, in some sense of ‘causal’. This is what makes the ‘biological machine with complex innards’ story, and the related functionalist view in philosophy of mind, seem plausible at all. But when we think of persons the complexity we are aware of is unified in a particular way. It is not unified just as ‘the states of the bits of stuff inside that skull’ but rather as ‘the elements of the coherent world view constructed by the person whose body that is’. And the person is unified inasmuch as her mind is unified, i.e. inasmuch as the elements of it are seen as cohering and are brought to bear on one another to suggest new conjectures, to correct miscon-
Understanding Other Minds from the Inside
51
ceptions, to provide mutual support through their rational connection and so forth. A person becomes aware of her world and builds up a picture of it, through perception, memory and reasoning. And that view must be unified in the way sketched. But let us note also that her view will necessarily include, woven in among the rest, many indexical thoughts, defining her beliefs about herself, her placement, role, capacities and so forth. For example they will include beliefs of the form ‘I am in such and such a location’, ‘I am capable of these or those actions’, ‘I occupy such and such a role’, ‘These and those achievements, dangers, disappointments or pleasures are possible for me’ and so forth. These elements may be said to define ‘a point of view’ on the world, in both a literal and a metaphorical sense. So when I attempt in simulationist style to recreate another’s thoughts, insofar as such indexical thoughts are included, then I have, to some greater or lesser extent, attempted to recreate her point of view. The suggestion I would like to pursue now is that it is this complex of ideas which makes the adoption of the idiom of ‘understanding from the inside’ so natural and attractive. There are a variety of strands in this metaphor. The mind of other is ‘inside’ in the sense that (sometimes at least) it is not immediately apparent in behavious what a person thinks and so that it takes reflection to see what her thoughts are. It is also ‘inside’ in that mental events and states are capable of moving the body to spontaneous (i.e. not immediately externally caused) movement. But there is also the fact that, on the view sketched, when I consider the nature of what is ‘inside’ another person, in the senses suggested by these two points, what I find myself postulating is a set of thoughts which represent the world from a point of view. So the ‘inside’ which I find is not mere mechanical or biological complexity. If the inside were of that kind there would not be any question of anything being ‘from’ it. But things can be ‘from the inside’ with a person because what is ‘inside’ is itself outwardly directed. It is an interlocking complex of items with indexical representational content concerning the world around that person. The existence of this kind of outward-directed content is bound up with the person’s ability to respond to changing perceptions and reasoning by modifying and enlarging the world view in rationally intelligible ways. We think of the content as having been build up by the exercise of the person’s cognitive capacities, her perceptual awareness and her abilities to remember and reason. The idea of reason then provides a further strand which enriches the ‘from the inside’ metaphor, inasmuch as in deploying it I represent other people as beings capable of recognising and responding to norms, whose thoughts and behaviour therefore have sense and can be justified, in ways which have no analogue in the explanations provided for the behaviour of inanimate items.
52
Jane Heal
What is the status of all this, you might ask. I have just outlined very roughly a picture which we have of ourselves and others – each of us a rational subject with a point of view, having multifaceted abilities to think effectively about many subject matters and so forth. And this picture is, I have suggested, the one presupposed by the form of simulationism which I have tried to outline and defend. I would also like to suggest the converse, namely that this kind of simulationism is the natural theory of the understanding of other minds for someone who conceives of persons as unified rational subjects. It is clear, then, that a presupposition of rationality, ability to appreciate what follows from what, to respond to reasons by grasping their force, is central to this whole complex of views. But could it be that this presupposition might turn out to be recognisably false? If so either we must say that, contrary to what I have urged, there is no conceptual link between the mental and rationality or we shall have to reconfigure our idea of the mental so as to extrude the rationality assumption. Or might it turn out that mental notions are inextricably bound up with the illusory idea of rationality and so need eliminating altogether? This is too big a topic to address here. But I would like to conclude by offering a few reflections. The idea that we are rational has received some excellent probing and clarification from philosophers; and psychologists have also undertaken fascinating empirical investigations bearing on the actual workings of our inferential processes. The joint upshot of this philosophical and empirical work is that it is quite clear that there are a number of grand and demanding senses of ‘rational’ in which we cannot properly claim to be rational. Such ability as we have is imperfect, limited by the finitude of our memories and by the amount of complexity we can take in. We do not have the time, energy or attention even to do all of the comparatively simple thinking and inferring which would be useful to us, let alone many elaborate reasonings, and letting even further alone the grandiose projects of achieving total consistency or coming to recognise all the logical consequences of what we accept. And, worse, we seem to be prone to systematic errors in elementary reasoning; there seem to be inferential versions of perceptual illusions such as the Mueller-Lyer case, where we are gripped by the conviction that something follows from something else when it does not.1 So our rationality, if it exists, does not amount to anything very grand. But then we do not need anything very grand to defend the picture sketched
1
Two excellent books, the second of which also provides references to much other recent work, are C. Cherniak, Minimal Rationality (Cambridge MA: MIT Press, 1986) and E. Stein, Without Good Reason (Oxford: Oxford University Press, 1996).
Understanding Other Minds from the Inside
53
above, any more than we need to credit ourselves with illusion-resistant eyesight of eagle-like acuity in order to defend the claim that in vision we have a sense which enables us to become aware of the placement and properties of things about us. Sight is reliable enough for us to be able to become aware, when we reflect, of when it is prone to illusion. So we can use it in increasingly subtle and well-focused ways (involving cross-checking, self-critical awareness of possible sources of error, helpfully devised instruments and the like) to progressively improve our grasp on the layout and properties of objects. The central claim we need about rationality is closely analogous. We need to be entitled to the assumption that thinking about a question, deploying all that we know which bears on it, will generally tend to improve our grasp on that issue rather than the contrary. And, as in the case of vision, what makes this central idea defensible is that we are capable of such things as reflecting on our reasoning practices, of recognising mistakes through cross-checking and of turning, where need be, to various aids. And thus we are capable of progressively improving our sense of what follows from what. No empirical evidence currently to hand shows that we are not entitled to the assumption that our basic thinking capacity is fundamentally pointed in the right direction, i.e. in the direction of leading us, when we employ it, to a better grip on things. Indeed the empirical studies which identify our inferential shortcomings are precisely evidence to support the assumption. And how would anyone who did not make it proceed with his or her thinking? What is the practical alternative to making it? There is none. Consider finally something about our relations with other people. It is often taken for granted in the discussions of philosophers and psychologists that the central role of psychological concepts is to enable us to predict inner states in others so that we can, in turn, predict and sometimes influence the behaviour those states bring about. But this is a serious distortion. Our relations to other people do not have the same structure as our relations with inanimate objects, plants or machines. We do not deal with our family members, friends, colleagues or fellow citizens as we do with volcanoes, fields of wheat or kitchen mixers, namely by trying to figure out the nature and layout of their innards so that we can predict and perhaps control them. Prediction and control may (sometimes rightly and sometimes wrongly) be the name of the game for psychiatrists, prison governors or dictators, in some of their dealings with some people. But it is not the name of the game for most distinctively human interactions. A much more central pattern occurs where one person offers to another some articulation of how things strike him or her – a remark, gesture, action or expression – in the course of pursuing some more or less well-defined joint project. Certainly this will be offered in the
54
Jane Heal
expectation (or at least the hope) that it will be identified for what it is. Thus far a prediction will probably be made. And also a prediction will be made that a response of a certain very broad class will be forthcoming. So in a philosophical debate one will expect to get back a philosophical question or observation, in a chess game one will get a chess move, in a game of mud pies one will get an elaboration of the mud pie world, in a courtship one will get a move to deepen the intimacy and so forth. But the specific nature of the response is not predicted. Social life would be utterly boring, completely different from the communicative reality we experience, if it were. What we hope of another with whom we interact is not that he or she will go through some gyrations which we have already planned in detail, but that he or she will make some contributions to moving forward the joint and co-operative enterprise on which we are both, more or less explicitly, engaged. There will be in any particular case many moves which would fit the general bill; which move an individual makes depends upon his or her individual appreciation of the situation, to which he or she brings not only differences in temperament, inventiveness and the like but, nearly always, differences in awareness both of empirical facts and of what follows from what. In a philosophical discussion the parties will probably share a good deal of common ground; but they will not be, psychologically speaking, identical twins. That is why there is a point in discussion; we engage in it so that we can pool our knowledge, insights, inventiveness and so forth. This is one way of combating our finitude, namely by having different of us pursue different lines of thought, since there is typically more labour in discovery than in appreciation of the discovery once made. Division of intellectual labour is not something which comes on the scene merely with large accumulations of knowledge and specialisation in the sciences. It is built right in to the idea of conversation and co-operation in the most everyday activities and plans. The way in which we carry on such activities shows that we presuppose the rationality of others, presuppose that we share standards of what follows from what and what is relevant to what. Our first move, on finding another’s response not immediately intelligible and helpful, is to search round for an interpretation which makes it so. And if others disagree with us about what constitutes good reasoning, making moves which show that at some level they do not share standards about what follows from what, then we seek to put them right in the expectation that they will acknowledge the mistake (or perhaps they will show that it is we who have made the mistake). Let us further note that when a mistake is agreed to have been made we will often look for, and find, a reason why it was made, not just in the sense of a cause or regularity in its making but in the sense of some excuse which reconciles the
Understanding Other Minds from the Inside
55
mistake with the idea that, even in making it, the perpetrator was exercising his or her rationality. This may be done by pointing to the false presuppositions which were accepted, the misleading analogy which was unduly prominent or some similar factor. Few mistakes, whether factual errors or mistakes in reasoning, are just opaquely and blankly completely unintelligible when reflected on. Some shred of justification can nearly always be found. And how well this general orientation to others serves us, how well things work out, on the whole. And how completely lacking we are of any conception of how things could be differently conducted. Empirical studies of our limitations and proneness to error (together with such things as awareness of the differences of our own outlook from those of other cultures and times or Freudian insights into the deeper workings of our motivations and self-conceptions) may all enrich the mixture and make us aware that the intelligibility we seek is not always to be found easily or on the surface. But such things do not prevent us looking for reason and intelligibility or stand in the way of our thinking, in most cases, that we have found it. So I conclude that rationality, in the schematic but still powerful sense sketched, is a very deeply entrenched assumption in our picture of human beings, ourselves and others, and hence that the understanding of other minds which calls upon the simulationist framework is not to be easily dislodged or replaced.
First place of publication: Current Issues in Philosophy of Mind, ed. Anthony O’Hear, Cambridge University Press 1998 rep. with a few very minor corrections from the 1998 text.
56
Christopher S. Hill
CHRISTOPHER S. HILL From Assertion to Belief: The Role of Linguistic Data in the Practice of Belief-Ascription1 Abstract This paper is concerned with the question of how we arrive at knowledge of the propositional attitudes of other agents. I describe a number of methods, but focus on the method that involves arriving at conclusions about the beliefs of others from information about their assertions and acts of assent. I attempt to give a reasonably full characterization of this method. Among other things, I maintain that when it is properly understood, the method is seen to be altogether independent of simulation. Thus, one conclusion of the paper is that simulation is not in any sense a universal method. At best, it is a method that we use in a highly restricted range of situations. Another conclusion is that there are features of the method of inference from assertion and assent that tend to provide support for the theory-theory – that is, for the view that our ascriptive practice involves the implicit use of a body of principles that resembles a scientific theory.
Section I: Introduction I will be concerned in these pages with one of the main methods that we use in arriving at beliefs about the psychological states of other human beings. Specifically, I will be concerned with the method of linguistic inference – that is, with the method of inferring psychological states from linguistic behavior. This method is important for four reasons. First, it appears to be highly reliable. There is, it seems, no better guide to the internal states of others than the descriptions that are provided by their possessors. Second, MLI is extremely powerful, in the sense that it gives us access to psychological states that no other method can reveal. Consider, for example, the belief that Fermat’s Last Theorem is true. Apart from MLI, it appears that we have no method of determining whether other agents have this particular belief. Third, the scope of the method is extremely broad. It enables us to arrive at beliefs about prop1
This paper was written while I was participating in Robert Gordon’s N.E.H. Summer Seminar on Folk Psychology and Simulationism. I am grateful to Professor Gordon and my fellow seminarians, especially James Garson and Nigel Thomas, for encouragement, stimulation, and advice.
From Assertion to Belief
57
ositional attitudes, but also to arrive at beliefs about moods, emotions, images, and bodily sensations. Fourth, it is a method that we use constantly, both in everyday life and in psychological research. Perhaps it is even true that we rely on it more frequently than we rely on any other method for arriving at conclusions about other minds. My first goal is to construct a description of MLI; more precisely, it is to describe a component of MLI that appears to play a particularly fundamental role in our ascriptive practice. My second goal is to characterize, at least in a rough and preliminary way, the scope of our deployment of this component of MLI. In pursuing these goals, I will simply be trying to make explicit certain assumptions that are built into commonsense psychology. The assumptions in question are not deep or hidden. Thus, in a sense, in describing the nature and scope of MLI, I will be concerned only to formulate certain platitudes. I feel, however, that these platitudes have been overlooked in the contemporary literature on knowledge of other minds. Participants in the contemporary discussion tend to be associated with one or the other of two schools of thought. Advocates of simulationism claim that we interpret others by simulating them, that is, by making imitative and representational use of the very faculties that are responsible for the formation of our own beliefs, desires, intentions, etc. Advocates of the theory-theory maintain that we arrive at conclusions about the mental states of others by applying a body of principles that are in a sense similar to the postulates of a scientific theory. As far as I have been able to determine, neither of these groups has given much attention to MLI.1 This oversight is unfortunate. As we noticed, MLI plays a quite important role in our ascriptive practice. No theory of ascription that fails to come to terms with it can reasonably be regarded as complete. Moreover, as we will see in the sequel, MLI provides a basis for settling decisively a couple of key questions in the debate between simulationists and theory-theorists. Thus, by reflecting on MLI, I will argue, we can come to appreciate that simulationists are at least partly wrong: there is at least one dimension of our ascriptive practice that cannot be regarded as involving simulation. We will also see that advocates of the theory-theory are at least partly right: there is at least one dimension of our ascriptive practice that can be said to be similar, albeit in a rough and extended sense, to scientific theorizing.
1
There are illuminating accounts of the debate between theory-theorists andsimulationists in the introductions to Davies and Stone 1995a, and Davies and Stone 1995b.
58
Christopher S. Hill
Section II: The Nature of AB Inference In the interests of definiteness, I will focus in this discussion on what I will call AB inferences. AB inferences differ from other forms of MLI in that their premises are always concerned with acts of assertion or assent and their conclusions are always concerned with beliefs. Thus, for example, I perform an AB inference when, having observed you assert the English sentence “Mark McGwire is the home run king,” I infer that you believe that Mark McGwire is the home run king. What can we say about the form or structure of AB inferences? My suggestion is that in the simplest and most fundamental case, they have a form that is captured by the following schema: (S)
Agent x asserted or assented to S. S means that p. Therefore, agent x believes that p.
Of course, when we compare schema (S) to concrete examples of AB inference, we quickly realize that it involves considerable idealization. For example, it fails to reflect the complexities of AB inferences in which the linguistic data include indexical expressions. Reflection suggests, however, that many such complexities, if not all, can be adequately captured by adding bells and whistles of various sorts to schema (S). Even when it is recognized that schema (S) is to be seen as an idealization, one might have doubts about its adequacy as a representation of the structure of AB inference. It appears to give us license to attribute beliefs whenever we encounter assertions. But assertions can be insincere. Further, assertions can be the products of emotional pressures or self deception, and when they are, they are poor guides to the underlying doxastic states of speakers. Still further, it sometimes happens that a speaker is misusing the words that figure in an assertion. Thus, for example, I may say “Mary is jail bait,” meaning thereby to communicate my belief that Mary is a jailbird. In situations of this sort, the normal connections between assertion and belief are dissolved. Now it is plausible that the mechanisms that underlie AB inference are in some sense sensitive to the distinction between speakers who are sincere and speakers who are not. Equally, it is plausible that they are sensitive to the distinction between speakers who have self knowledge and speakers who are self deceptive. Further, it is plausible that they are sensitive to the distinction between speakers who have a good command of the language and speakers whose mastery of the language is partial or flawed. In view of these facts, it is somewhat tempting to suppose that AB inference does not have the compara-
From Assertion to Belief
59
tively simple form that is represented by schema (S), but rather the form that is captured by schema (S’): (S’)
Agent x asserted or assented to sentence S. Agent x was being sincere. Agent x is not a victim of self deception. Agent x is a fully qualified user of language L. S means that p in L. Therefore, agent x believes that p.
Prima facie, at least, schema (S’) does a much better job than (S) of honoring the complexities that must be taken into account if we are to arrive at reasonable hypotheses about beliefs. I of course wish to grant that the cognitive systems that support AB inference are sensitive to the distinctions we have just been considering. Otherwise we would be misled by linguistic behavior much more often than we actually are. But I wish to distinguish between two ways in which such sensitivities might be incorporated in the cognitive systems in question. Let us say that M is the cognitive module that supports the most basic and most frequently used form of AB inference. On the one hand, it might be the case that, for example, a sensitivity to the distinction between sincerity and insincerity is built into M itself. That is to say, it might be part of M’s job to determine whether there is evidence which indicates whether the current speaker is insincere. On the other hand, it might be that M itself is not equipped to classify or weigh evidence that is pertinent to questions of sincerity. In general, it might be that M proceeds as if normal situations were the only possible situations. On this scenario, the systems responsible for tracking sincerity, self knowledge, and command of the language would be distinct from M. Moreover, they would not be activated unless there was some positive evidence to indicate that the current speaker was insincere, or self deceptive, or incompetent. Once activated, however, they would quickly acquire the status of control mechanisms, overriding or inhibiting the processes in M, and transferring the responsibility for AB inference to other mechanisms that are capable of more complex inferences than M. So we have two hypotheses to consider. According to one, the module that is responsible for AB inferences in normal cases incorporates cognitive engines that are capable of tracking sincerity, self knowledge, and linguistic competence. And according to the other, the module that is responsible for AB inferences in normal cases is functionally independent of the modules that perform the tracking tasks, though it can be over-ridden or inhibited by the latter modules in unusual situations. Now it is reasonably clear that, absent
60
Christopher S. Hill
relevant experimental evidence, the latter hypothesis is the one that should be preferred. For it is much simpler than the alternative. On the former hypothesis, the cognitive systems that are responsible for AB inferences always check for sincerity, self knowledge, and linguistic competence. On the latter hypothesis, they only do so when there is some positive reason to think that one of these characteristics is absent. Thus, since sincerity, etc. are present in normal cases, the former hypothesis implies that the cognitive systems in question devote a good deal of time and energy to tasks that are altogether unnecessary. It seems unlikely that the brain would be guilty of such gross inefficiency. In view of these considerations, it seems reasonable to take the claim that (S) succeeds in capturing the most basic form of AB inference as a working hypothesis, and to set schema (S’) aside. This is, at all events, what I propose to do here. This brings us to another question. Reasoning that conforms to schema (S) must incorporate a premise of the form “S means that p.” What can we say about the cognitive processes that provide us with premises of this form? How do we come to know that they are true? Here we must distinguish between two types of situation. In the normal case, the agent who arrives at such a belief is a fully qualified speaker of L, and is able to deploy his or her practical command of L in coming to hold the belief. In the other sort of case, which represents only an extremely tiny fraction of the whole range of cases, an agent arrives at a belief of the given form by exercising what might be called an external command of language L – that is, by applying a translation procedure that is based on a lexicon and a set of explicitly represented grammatical rules. Because of its statistical salience, I wish to focus here on the first sort of case. So let us ask: What happens when you, a speaker of English, move from a perception of a token of the English sentence S to a belief of the form “S means that p in English”? I suggest that two processes are involved. First, immediately on perceiving the token in question, the language processing mechanisms that support your practical command of English cause you to entertain the thought that p. The operation of the mechanisms does not involve discursive reasoning of any kind, and is in fact entirely sub-personal. The second process is inferential, but the relevant inference is quite simple in structure, being performed by a mechanism that incorporates the following default assumption: “If I come to entertain the thought that p as a result of applying my language processing mechanisms to S, then S means that p in English.” The main idea here is that we do not make use of special-purpose cognitive faculties in forming concrete semantic beliefs, but rather rely primarily on the mechanisms that support our practical ability to grasp the contents of English
From Assertion to Belief
61
sentences. This idea is attractive because of its simplicity, and also because it can be embraced without incurring significant risks. We know that mechanisms of the sort in question exist. Accordingly, the hypothesis that we arrive at concrete semantic beliefs by exploiting these mechanisms involves fewer risky commitments than the hypothesis that we employ an altogether different set of mechanisms. This completes my sketch of the nature of AB reasoning. I have been concerned here only to characterize the simplest and most basic forms of such reasoning; but even though extremely modest in scope, the present investigation appears to support two conclusion that are of some theoretical significance.
Section III: Implications for the Theory-theory/Simulationism Debate First our reflections show conclusively that simulationism is at least partly mistaken. Thus, as we noticed earlier, simulationists maintain that we arrive at beliefs about the mental states of others by making imitative and representional use of the very faculties that are responsible for our own mental states. Now reasoning that has the form of Schema (S) is not in any sense imitative. It is just a particular form of inductive reasoning. But it is clear that our interpretive practice depends heavily on our inferences of form (S). It follows that much of our interpretive practice does not involve simulation, and by the same token, that simulation is at least partly wrong. Second the foregoing line of thought supports the conclusion that advocates of the theory-theory are at least partly right. The theory-theory maintains that when one goes about determining what mental states to attribute to others, one relies on principles that are similar in several respects to the principles of scientific theories. As I see it, this view receives strong support from the claim that we rely on instances of (S) in attributing beliefs. To be sure, (S) itself does not have the form of a theoretical principle, for it does not have the form of a proposition. It can, however, be appropriately regarded as an implicit representation of a theoretical principle. To see this, consider proposition (P): (P) For every proposition q, sentence S, and agent x, if it is the case that x asserts or assents to S, and that S means that q, then, ceteris paribus, x believes that q. It is evident that if a cognitive mechanism is dedicated to producing reasoning that has the form exhibited by schema (S), then that mechanism can reasonably be said to presuppose the truth of (P), and can also reasonably be said to
62
Christopher S. Hill
represent (P) implicitly. Moreover, it is reasonably clear that it is appropriate to describe (P) as a theoretical principle. Like standard cases of theoretical principles, it has a high degree of generality; it employs predicates that stand for unobservables; it can be used in prediction; and it can be regarded as making explicit certain aspects of the meanings of its constituent terms. In view of these considerations, it seems fair to say that when we engage in reasoning that is of form (S), we are implicitly relying on a proposition that is fundamentally akin to propositions that are constituents of scientific theories. Now there is a version of the theory-theory which maintains that we make explicit use of theoretical principles in folk psychological reasoning. Of course, the present line of thought has no tendency to support that form of the view. It is my understanding, however, that a number of advocates of the theory-theory are fully prepared to allow that our reliance on such principles is implicit.1 Thus, as I see it, it is common among theory-theorists to deny that we arrive at folkpsychological conclusions by applying simple, domain-neutral deductive procedures in which theoretical postulates figure as explicit premises, and to maintain instead that we arrive at such conclusions by employing domainspecific procedures that are substantially more powerful (and also less trustworthy) than the rules of deductive logic. It must be acknowledged that (P) is different in an important respect from the paradigmatic laws that we find in physics and chemistry. Thus, (P) purports only to describe a pattern that obtains in normal cases. In other words, “ceteris paribus” is essential to (P); without that clause, (P) would be trivially false. It would, however, be a mistake to think that this feature prevents (P) from counting as a law. As Fodor and others have emphasized (see, e.g., Fodor 1987, pp. 4-5), the special sciences are full of laws that hold only ceteris paribus.
Section IV: The Scope of AB Inference I would like to turn now to consider a couple of questions about the role of AB inference in our ascriptive practice. Is its use pervasive or restricted to contexts of certain particular types? Further, what is the comparative size of its role? That is to say, how does its role compare in size with the roles of other methods of arriving at hypotheses about the doxastic states of others? Now these are empirical questions, and they are therefore questions that are best approached from the perspective of cognitive science. Since the per1
The view that our grasp of folk-psychological laws is implicit figures prominently in the work of Stich and Nichols. See Stich and Nichols 1995a, and Stich and Nichols 1995b.
From Assertion to Belief
63
spective of the present paper is one that is defined by commonsense knowledge and the method of reflection, it follows that we can only hope to arrive at conclusions that are rough, tentative, and incomplete. It seems, however, that there are several achievable conclusions of this sort that are worthy of our attention. It will, I think, be helpful to begin by taking note of several methods of belief ascription that count as alternatives to the method of AB inference. I will mention seven such methods. The first of these is what might be called the method of perceptual interpretation. In applying this method to an agent, we take note of the states of affairs that figure most prominently in the agent’s current perceptual environment, and we then attribute beliefs to the agent to the effect that those states of affairs obtain. Second, there is the method of testimonial interpretation. This method is grounded in our intuitive conviction that agents are inclined to accept the testimony of others unless they have some specific reason to suspect that it is untrustworthy. The third alternative method is the method of inferential interpretation. Folk psychology implies that we are disposed to form new beliefs by applying deductive and inductive inferential procedures to beliefs that we already possess. We are all mindful of this implication, and we are all moved by it to attribute beliefs to others when we know that the beliefs can be reached by obvious inferences from beliefs that the others can be seen to possess on independent grounds. Fourth, there is a method that might be called the method of introspective interpretation. This method consists in moving from a premise of the form agent x is in psychological state P and a second premise of the form agent x has reason to consider whether he or she is in state P to a conclusion of the form agent x believes that he or she is in state P. The fifth alternative method is the method of memorial interpretation. In applying this method, we begin by considering whether there is reason to suppose that an agent has held a certain belief in the past. Assuming that the answer to this question is affirmative, we conclude that the agent continues to hold the belief. The alternative methods that we have considered thus far can easily be subsumed under a generalization. So as to be able to express this generalization, let us agree to say that a doxastic mechanism is a cognitive mechanism that either causes us to form beliefs or to retain beliefs that have been independently formed. Then we can make the following claim: for every doxastic mechanism M, there is a method A of attributing beliefs such that application of A simply leads us to attribute beliefs that agents are likely to have formed or to have retained as a result of M. All five of the alternative methods that we
64
Christopher S. Hill
have considered thus far are paired with doxastic mechanisms in the way that this generalization describes. Unlike the first five, the remaining alternative methods of attribution can be explained without reference to the doxastic mechanisms of attributees. One of them is what might be called the method of inference from doxastic generalizations. Here is an example of how this method works. I know that almost everyone living in the United States who is above the age of twenty believes that Clinton succeeded Bush to the Presidency. Applying this knowledge to a new acquaintance, I infer that it is very likely that he or she holds the given belief. The last alternative method is inference to best explanation. In applying this method, we observe a series of actions and ask ourselves what set of beliefs and desires provides the simplest and most plausible explanation of why the agent was led to exhibit the actions of which the series is composed. Assuming that we can find a set of beliefs and desires that fits these criteria, we attribute the members of the set to the agent. Now these alternative methods are interesting, and they are worthy of careful scrutiny. For one thing, in the case of several of them, it is prima facie plausible that their use involves the use of simulation. Thus, for example, in applying the method of inferential interpretation to you, I may try to determine what inferences you are likely to have performed by using my inferential mechanisms to simulate yours. It would be nice to have some sort of sense of how frequently this sort of thing happens. Here, however, I wish to set aside questions about what we do in applying the alternative methods, and to focus instead on questions about the relationships between the alternative methods and the AB method. There are three points that I would particularly like to make. First, if we compare the role of the method of AB inference to the sum of the roles of all seven of the alternative methods, we quickly come to appreciate that it would be a mistake to claim either dominance or pervasivesness for the former method. There are a great many contexts in which it is easier to rely on one of the alternative methods, and some contexts in which we are forced to rely on one of the alternatives. Thus, for example, if you and I are at dinner, it would be silly for me to rely on the AB method to get at your beliefs about the arrangement of plates and eating utensils on the table. It would be much more efficient to deploy the method of perceptual interpretation. It is also true that, in most cases, anyway, ethologists have no choice but to employ one of the alternative methods in interpreting the behavior of their subjects. In general, it seems that ethologists rely heavily on inference to best explanation in the early stages of an investigation. Once they have used this technique to arrive at hypotheses about the kinds of thoughts that their subjects can enter-
From Assertion to Belief
65
tain, they come to rely increasingly on perceptual interpretation, inferential interpretation, memorial interpretation, and inference from doxastic generalizations.1 On the other hand, and this is my second point, if we compare the role of AB inference to the role of any one of the alternative methods, it is natural to conclude that the former is impressively large. Thus, consider what you know of the doxastic states of your acquaintances. My guess is that in the typical case, your knowledge of an acquaintance is based to a very large extent on what you have learned about him or her in conversation. Indeed, I conjecture that in the typical case, the portion of your knowledge that is due to AB inference is considerably larger than the portions that are severally due to the alternative methods. The third point is that in applying the alternative methods we are sometimes depending indirectly on the AB method. The most obvious examples of this, though by no means the only examples, are afforded by applications of the method of inference from generalizations. Thus, consider the generalization that students at the University of Arkansas tend to believe that the parking facilities at the University are grotesquely inadequate. I accept this generalization, and I have on occasion used it as a ground for attributing beliefs to individual students. But I have never used the parking facilities at the University! Nor have I ever watched students attempting to find parking places. My only evidence for the generalization comes from overhearing student complaints. To summarize: In addition to the method of AB inference, there are several alternative methods that we use in arriving at hypotheses about the beliefs of other agents. It would therefore be wrong to suppose that the AB method is in any way the core or the foundation of our attributive practice. On the other hand, it is not implausible to say that its scope is as large, or at least nearly as large, as that of any other method of attribution. Moreover, in some cases, anyway, its deliverances appear to be presupposed by alternative methods.
Bibliography Cheyney, D. L. and Seyfarth, R. M. 1990: How Monkeys See the World. Chicago: University of Chicago Press Davies, M. and Stone, T. (eds.) 1995a: Folk Psychology. Oxford: Blackwell Publishers
1
Of course, the range of cases in which ethologists feel entirely comfortable in attributing beliefs and other high level psychological states is comparatively small. My remarks in the text are intended to primarily to characterize research on chimpanzees, baboons, and various species of monkeys. See, e.g., de Waal 1982, Strum, 1987, and Cheyney and Seyfarth 1992.
66
Christopher S. Hill
Ltd. Davies, M. and Stone, T. (eds.) 1995b: Mental Simulation. Oxford: Blackwell Publishers Ltd. de Waal, F. B. M. 1982: Chimpanzee Politics. New York: Harper and Row. Fodor, J. A. 1987: Psychosemantics. Cambridge, MA: MIT Press. Stich, S. and Nichols, S. 1995a: Folk Psychology: Simulation or Tacit Theory?” in Davies and Stone 1995a, pp. 122-158. Stich, S. and Nichols, S. 1995b: “Second Thoughts on Simulation,” in Davies and Stone 1995b, pp. 87-108. Strum, S.C. 1987: Almost Human. New York: Random House.
Content, Interpretation, and Consciousness
67
DAVID M. ROSENTHAL Content, Interpretation, and Consciousness Abstract According to Dennett, the facts about consciousness are wholly fixed by the effects consciousness has on other things. But if a mental state's being conscious consists in one's having a higher-order thought about that state, we will in principle have an independent way to fix those facts. Dennett also holds that our speech acts sometimes determine what our thoughts are, since speech acts often outrun in content the thoughts they express. I argue that what thoughts we have is independent of how we express them in speech, and that this is consonant with speech acts’ often seeming to have more fine grained content than the thoughts they express. This model has the advantage, compared with Dennett’s, of accommodating our folk-psychological taxonomy of intentional states and preserving the traditional idea that speech acts express antecedent intentional states. Speech acts doubtless do sometimes have richer content than the thoughts they express, though sometimes verbally expressing a thought simply makes us conscious in a more fine-grained way of what that content is. On the higher-order-thought model, as on Dennett’s, a mental state’s being conscious is, in effect, our spontaneously interpreting ourselves as being in that state. But such spontaneous self-interpretation need not be the last word on what content our thoughts have. Even though the content of speech acts sometimes outrun that of the thoughts they express, we can explain why the two seem always to be exactly the same. Even when a speech act is richer in content than the thought it expresses, the well-entrenched pragmatic equivalence between saying something and saying that one thinks that thing ensures that one will be conscious of one’s thought as having the richer content of the speech act that expresses it. We are conscious of our thoughts as having the content that our speech acts would capture.
I. First-Person Operationalism and Higher-Order Thoughts We are all familiar with situations in which memory distorts some current experience. I may see a person I don’t know at all, but my memory of an old friend causes me to misperceive that person as my friend; the conscious experience that results is of seeing the friend. Perhaps, for example, my friend wears glasses and the person now before me doesn’t. Although I see a person with-
68
David M. Rosenthal
out glasses, my memory of the friend intrudes and I seem, so far as consciousness is concerned, to see a person wearing glasses. Folk psychology accommodates two distinct explanations of such cases. My memory of the earlier experience might contaminate the current visual information before it even reaches consciousness; if so, I in effect hallucinate the glasses. But perhaps, instead, I begin by consciously seeing the person as having no glasses, but the memory then immediately revises the experience by adding the glasses and also overwrites any current memory of the new visual experience without glasses. These are the two scenarios that Dan Dennett and Marcel Kinsbourne labeled Stalinesque and Orwellian, respectively, famously arguing that the distinction between them is spurious. When it comes to consciousness, they urge, we cannot distinguish between appearance and reality. So, if the two scenarios are indistinguishable to consciousness, they are indistinguishable in reality, as well. Since consciousness cannot fix the time of contamination as being before or after the new visual information reaches consciousness, the two hypotheses differ only verbally. Folk psychology, by allowing for distinct scenarios, misleads us into thinking that this kind of case can occur in two different ways.1 According to Dennett and Kinsbourne, “there is no reality of consciousness independent of the effects of various vehicles of content on subsequent action (and hence, of course, on memory)” (CE 132). So “there are no fixed facts about the stream of consciousness independent of particular probes” (CE 138; cf. 275). Dennett usefully calls this view first-person operationalism (CE 132), since it holds that the facts about consciousness are wholly fixed by the effects consciousness has on other things. Because theory outstrips observation, a satisfactory theory can often settle questions that won’t yield to observation alone. So even if the Stalinesque and Orwellian scenarios are indistinguishable to consciousness itself, perhaps a reasonable theory of consciousness will, in principle at least, show us how to tell which scenario any particular case conforms to.
1
Daniel C. Dennett and Marcel Kinsbourne, “Time and the Observer: The Where and When of Consciousness in the Brain,” The Behavioral and Brain Sciences 15, 2 (June 1992): 183-201.; open peer commentary, 201-234, and Dennett and Kinsbourne’s Author’s Response, “Escape from the Cartesian Theater,” 234-247; and Dennett, Consciousness Explained, Boston: Little, Brown and Company, 1991 (henceforth CE), chs. 5 and 6.
Content, Interpretation, and Consciousness
69
I have argued elsewhere1 that we can do just this on a theory according to which a mental state is conscious just in case it is accompanied by a higher-order thought (HOT) to the effect that one is in that state. We do not regard as conscious any mental state of which we are wholly unaware. So we must in some way be conscious of every conscious state, and having a thought about a state is one way of being conscious of it. Intuitively, it seems that the way we are conscious of our conscious states is direct. We can explain this intuition by hypothesizing that we remain unaware of any inferences or other antecedent factors that might lead to HOTs or explain their occurrence; HOTs seem to arise spontaneously.2 Indeed, HOTs need not themselves be conscious, and typically won’t be.3 On this theory, a mental state becomes conscious at the onset of the relevant HOT. So whether my unrevised visual sensation reaches consciousness depends solely on whether the contamination occurs before or after the onset of some HOT. Since HOTs are determinate states, their exact moment of occurrence is determinate, whether or not we can discover it in practice. At the same time, however, the HOT model explains why we should feel a certain reluctance to classify particular cases as being Stalinesque or Orwellian. Suppose a sensation occurs and very quickly becomes conscious. But then, almost immediately, the sensation changes, and a moment after that one becomes conscious of the change. Is this revision Stalinesque or Orwellian? That depends on whether we focus on the sensation in its original or changed form. The case is Stalinesque relative to the changed sensation, since that change occurred before the sensation became conscious in its new form. But relative 1
“Multiple Drafts and Higher-Order Thoughts,” Philosophy and Phenomenological Research, LIII, 4 (December 1993): 911-918; “Multiple Drafts and Facts of the Matter,” in Conscious Experience, ed. Thomas Metzinger, Exeter, UK: Imprint Academic, 1995, pp. 275-290; and "First-Person Operationalism and Mental Taxonomy," Philosophical Topics, 22, 1 and 2 (Spring and Fall 1995), pp. 319-349.
2
Though HOTs are noninferential from a first-person point of view, they might still result sometimes from inferences we are unaware of. We need not suppose that the way we are conscious of our conscious states actually is unmediated to explain our intuitive sense that it is.
3
They will be conscious only when we are conscious of target states in the deliberate, attentive way we call introspection. For development of this HOT model, see “A Theory of Consciousness,” in The Nature of Consciousness: Philosophical Debates, ed. Ned Block, Owen Flanagan, and G•ven G•zeldere, Cambridge, Massachusetts: MIT Press, 1997, pp. 729-753; “Thinking that One Thinks,” in Consciousness: Psychological and Philosophical Essays, ed. Martin Davies and Glyn W. Humphreys, Oxford: Basil Blackwell, 1993, pp. 197-223; “State Consciousness and Transitive Consciousness,” Consciousness and Cognition, 2, 4 (December 1993): 355-363; and “Two Concepts of Consciousness,” Philosophical Studies 49, 3 (May 1986): 329-359.
70
David M. Rosenthal
to its original form the case is Orwellian, since the sensation was conscious in that original form before it changed. The case looks Stalinesque if we regard the revised sensation as a new, distinct state, and Orwellian if we see the revised sensation as just a latter stage of the original state. But the choice between these two descriptions is in most cases unprincipled, since it will typically rely on artificially precise identity conditions for mental states. Whether such a case is Stalinesque or Orwellian hinges on arbitrary questions about taxonomizing our mental states themselves. Still, this does not show that it’s not determinate when states with specific content properties become conscious, nor that the facts about when revision occurs and when states become conscious are exhausted by how things appear to consciousness.
II. Thinking, Speech, and Probes “[T]here are,” according to Dennett, “no fixed facts about the stream of consciousness independent of particular probes.” One of the most intriguing applications of this challenging claim has to do with the connection between thinking and speech. When we speak, we express the thoughts we have. Our speech acts, moreover, reflect the content of those intentional states. It is natural to hold that this correspondence of content is exact; whenever we say anything, the speaker’s meaning of our speech act is the same as the content of the intentional state we express. We all have experienced how putting our thoughts into words can appear to tighten up those very thoughts. It’s usually assumed that this happens because the process of finding suitable words for, one’s thoughts clarifies the thoughts themselves. On that account, the clarifying speech act does not actually outstrip, in respect of content, the antecedent intentional state it expresses. It’s just that fixing on the right words results in a new intentional state, whose content is more fine grained than one’s original thought. So goes our ordinary, folk-psychological description of these cases, and so it seems to us from a first-person point of view. We have a robust first-person sense that our speech acts exactly match in content the antecedent intentional states they express. Whether or not the match is exact, it seems that way to us. We will see below that folk psychology and our first-person sense of things overestimate the exactness of this match of content. But it is worth considering certain ostensible counterexamples to the idea that our first-person impression is always that an exact match obtains. We do sometimes discover as we say
Content, Interpretation, and Consciousness
71
something that what we’re saying does not really reflect our thoughts after all. We may have changed our mind, or even find that we never actually thought that thing at all, but said it only conversationally or from habit. But when this type of thing happens, our first-person sense is not of really saying the thing in question, but something like “as if” saying. We don’t sense ourselves saying the thing with full illocutionary force, but simply producing the relevant utterance. In Wilfrid Sellars’ apt metaphor,5 the utterance is produced parrotingly, as with mere recitations, from causes that are tangential to what one thinks. We do sense in these cases a divergence between our speech and what we think, but the speech productions do not seem to us, from a first-person point of view, to be full-fledged illocutionary acts.6 As we saw, an exact match in content between full-fledged illocutionary acts and the thoughts they express can accommodate cases in which our speaking seems to clarify our thoughts. But these cases seem to fit equally well with another interpretation, which itself lends support to Dennett’s First-Person Operationalism. Perhaps putting our thoughts into words clarifies our thinking not because it results in our having clearer thoughts, but because speech acts themselves actually fix the content our thoughts have. Speech acts, on this view, are one sort of probe that determine the facts of consciousness. This is the “Pandemonium Model” of speech production that Dennett develops in chapter 8 of Consciousness Explained,7 on which there are no determinate intentional states prior to the occurrence of verbal expressions. It is not simply that our choice of words often influences the content of our thoughts.8 Rather, many forces occur in the intentional arena and compete for expression. The speech act that ultimately wins out in effect results from the vector product of those forces, rather than from some single, antecedently existing state with the relevant determinate content.
5
First used in a published letter to Roderick Chisholm: Chisholm, Roderick M., and Wilfrid Sellars, “Intentionality and the Mental,” Minnesota Studies in the Philosophy of Science II, ed. Herbert Feigl, Michael Scriven and Grover Maxwell, Minneapolis: University of Minnesota Press, 1958, II, 507-539, p. 524.
6
Insincere speech and the speech productions of actors’ playing a part are also like this. See my “Intentionality,” Midwest Studies in Philosophy, X (1986), pp. 151-184, sec. v. Thus J. L. Austin notes that if I insincerely say ‘I promise’, I don’t strictly speaking promise, but only say I do (“Other Minds,” in Philosophical Papers, 2nd and 3rd editions, Oxford: Oxford University Press, 1970, 76-116, pp. 101-103 [1st edition, 44-84, p. 69-71]). Similarly, Frege remarks that “stage assertion is only sham assertion” (Gottlob Frege, “Thoughts,” in Logical Investigations, tr. P. T. Geach and R. H. Stoothoff, New Haven: Yale University Press, 1977, 1-30, p. 8).
7
Esp. 245 and 247; see also ch. 10, sec. 5, esp. 315.
8
Which Dennett suggests also happens (CE 247).
72
David M. Rosenthal
The folk-psychological view that speech acts always mirror the content of the intentional states they express reflects our subjective impression of these situations. It always seems, from a first-person point of view, that what we say exactly matches the content of some intentional state we are in. But Dennett’s Pandemonium Model also does justice to this subjective impression. After all, if speech acts do fix the content of the intentional states they express, we will sense a perfect fit between them. In addition to squaring with our subjective impressions, these two models both capture important aspects of the connection between thinking and speech. But the aspects they capture are different. One aspect has to do with what it is for a speech act to be meaningful. Two things are needed. One is that the sentence uttered have semantic meaning. But even so, no utterance is meaningful if it’s a mere recitation without underlying thought, if, as Sellars put it, it is produced parrotingly. We can, of course, distinguish tolerably well between meaningful and parroting speech production, independent of any appeal to theory. But if we want to explain what that difference consists in, we must appeal to the idea that meaningful speech acts express intentional states, whereas parroting utterances do not. The folk-psychological picture of the relation between speech and thought reflects this explanation. Nonparroting speech productions express antecedent intentional states, and expressing an intentional state means that the semantic meaning of the speech act matches to some suitable degree the content of that intentional state. Such an explanation is unavailable on the Pandemonium Model. If speech acts fix the content of our intentional states, it cannot be that expressing an antecedent intentional state with roughly the same content is what makes a speech act nonparroting. One might reply that the Pandemonium Model can, after all, explain the difference between parroting and nonparroting speech productions. Speech is nonparroting if it results from the forces competing in the intentional arena for expression; otherwise it is parroting.1 Whether this reply works depends on just what those intentional forces are that the Pandemonium Model posits. If they are merely subpersonal events that resist folk-psychological taxonomy as full-fledged intentional states, the reply fails, since parroting utterances also result from the interaction of subpersonal events that resist folk-psychological taxonomy. But perhaps the Pandemonium Model actually posits fullfledged intentional states operating in the intentional arena. Speech acts, then, simply settle which of these states wins out in the competition for expression, rather than 1
I owe this idea to Tim Kenyon, in conversation.
Content, Interpretation, and Consciousness
73
converting subpersonal events of content fixation1 into genuine intentional states. Then the model can explain speech as nonparroting if it results from such intentional states. But the Pandemonium Model so construed does not differ relevantly from the folk-psychological picture, which also posits antecedent intentional states, and regards nonparroting speech as that which results from such states. Any satisfactory explanation of how parroting and nonparroting utterances differ must invoke the intentional states posited by the folk-psychological model. There is good reason, in any case, to construe Dennett’s Pandemonium Model as positing only subpersonal events of content fixation, rather than full-fledged intentional states. Thus he writes: We replace the division into discrete contentful states – eliefs, meta-beliefs, and so forth – with a process that serves, over time, to ensure a good fit between an entity’s internal information-bearing events and the entity’s capacity to express (some of) the information in those events in speech.2
The forces whose interaction results in speech acts are properly speaking proto-intentional, and precipitate into intentional states only when they issue in speech acts or in reactions to some relevant probe. Moreover, just which speech acts and intentional states these subpersonal forces issue in depends to some extent on factors irrelevant to the intentional content at hand. Events of content fixation determine speech performances more in the manner of instructions in aleatory music than fully developed scores. Nonetheless, the folk-psychological model arguably exaggerates the match in content between speech act and intentional state, and the Pandemonium Model provides a useful corrective. If the content of an intentional state corresponds exactly to the semantic meaning of the speech act that expresses it, thinking must itself exhibit a language-like structure; perhaps thinking even takes place in a language of thought whose syntax and semantics echo those of overt natural language.3 Such a picture, however, arguably underestimates the 1
Dennett speaks throughout CE of such nonconscious, subpersonal events that subserve mental states taxonomized folk psychologically. Because they occur nonconsciously, “their onsets do not mark the onset of consciousness of their content” (113; emphasis Dennett’s). And he holds that, unlike mental states taxonomized folk psychologically, “content-fixations . . . are [each] precisely locatable in both space and time” (113).
2
CE 319, emphasis Dennett’s. Cf. “The Message is: There is no Medium,” Philosophy and Phenomenological Research LIII, 4 (December 1993): 919-931, pp. 930-31.
3
Most forcefully and impressively defended by Jerry A. Fodor in, e.g., The Language of Thought, New York: Thomas Y. Crowell, 1975; “Propositional Attitudes,” The Monist, LXI, 4 (October 1978): 501-523; “Why There Still Has to Be a Language of Thought,” Appendix in Psychosemantics: The Problem of Meaning in the Philosophy of Mind (Cambridge, Mass.: MIT Press/Bradford Books, 1987); “A Theory of Content,” in A Theory of Content and Other
74
David M. Rosenthal
extent to which what we think and how we think it are affected by the way we express our thoughts in speech. The Pandemonium Model seeks to capture the effect our putting words to our thoughts has on the thoughts themselves. Is there a view that preserves the virtues of both models, allowing us to explain the difference between parroting and nonparroting speech without tempting us to adopt an unqualified language of thought? Arguably yes. Perhaps meaningful, nonparroting speech acts always express intentional states with antecedently fixed content, but the speech acts are nonetheless often richer and more fine grained in content than the intentional states expressed. The content of the speech acts rules out more possibilities and invokes more distinctions than the less refined content of the intentional states those speech acts express. How would this work? Which speech acts we perform is of course largely determined by the intentional states they express. But speech acts express those states in words that have, to some extent, an independent semantic life.1 So, although the content of speech acts derives mainly from the intentional states expressed, the words used in expressing that content often go beyond it, making for speech acts whose semantic meaning is correspondingly enriched and refined. Our thoughts, by contrast, seldom need to respect the finegrained semantic distinctions inherent in natural language. Any intentional state can typically be expressed equally well by a range of speech acts whose semantic meanings are not exactly the same. So it’s reasonable to regard the content of our intentional states as neutral among the distinct semantic meanings of those various possible speech acts. Any particular choice of words, then, produces a speech act with semantic meaning more fine-grained than the content of the corresponding intentional state. Some match of content is required for a speech act to count as expressing an intentional state, but the match need not be exact. What words we choose to express an intentional state may itself sometimes be due to some other intentional state. In these cases we might regard the resulting speech act as expressing both intentional states: that which led to some such speech act’s being performed and also that which influenced the particular choice of words. The speech act would be more fine grained in content than either intentional state, but might not outstrip the two combined. Essays, Cambridge, Mass.: MIT Press/Bradford Books, 1990; and The Elm and the Expert: Mentalese and its Semantics, Cambridge, Mass.: MIT Press/Bradford Books, 1994. 1
Partly for reasons developed in compelling detail by Tyler Burge, “Individualism and the Mental,” Midwest Studies in Philosophy IV (1979): 73-121, “Individualism and Psychology,” The Philosophical Review XCV, 1 (January 1986): 3-45, and elsewhere.
Content, Interpretation, and Consciousness
75
But choices of words are doubtless often arbitrary, resulting from no mental factor at all. Mere habit or word pattern might determine what words we use; indeed, it is likely that this is what typically happens. In such cases, our speech act does outstrip in content any relevant antecedent intentional states. And, since the content of intentional states can be less fine grained than the meaning of the speech acts that express those states, there is less temptation on this view to suppose that thinking fully mirrors the syntactic and semantic properties of natural language. Indeed, it may sometimes be difficult to capture our more coarse-grained thoughts in words, especially with the intentional states we ascribe to nonlinguistic animals. Dennett therefore concludes that the “‘’thought’ [of such an animal] might be inexpressible (in human language) for the simple reason that expression in a human language cuts too fine.” But Dennett concedes that “we may nevertheless exhaustively describe what we can’t express.”1 But if we can exhaustively describe the content of a thought, we can use those very words to frame a sentence that would, however awkwardly, express that thought. If the content of some thought elides various distinctions inherent in human language, suitable disjunctions can provide neutrality in respect of those distinctions. Even when no straightforward sentence could express the content of thought, some complex compound should succeed.2 A speech act expresses an intentional state only if ist content matches that of the intentional state, but the match need not be exact. How close must that match be? No precise answer is possible, but neither should we expect precision. For one thing, it is unclear just how finely we can differentiate intentional states and speech acts in respect of content. In addition, the distinction between parroting and nonparroting speech productions itself doubtless admits of an intermediate gray area. Nor, finally, should we expect to be able to specify with precision just which speech acts express which intentional states. Some reasonably close match of content is required even though the
1
Kinds of Minds: Toward an Understanding of Consciousness, New York: Basic Books, 1996, p. 42. See also his “How to do Other Things with Words,” in Thought and Language, ed. John Preston, Cambridge: Cambridge University Press, 1997, pp. 219-235.
2
It is possible that many human thoughts have content that makes them better suited to being expressed by specific nonverbal actions than by speech; doubtless this is true of the thoughts of nonlinguistic animals. One might have a thought, e.g., that is expressed by one’s getting in out of the rain. Deliberate cases of such an action are not merely automatic; some intentional state with suitable content are among their causes. It may not be clear in these cases just what that content is--whether, e.g., to get out of the rain, to seek shelter, to stop getting wet, or what. But it would be surprising if that content could not be captured, in some gerrymandered way, in human language.
76
David M. Rosenthal
precise degree of correspondence eludes specification. It is an advantage of the present model that it does not aim for such precision.
III. Consciousness and Verbally Expressing Our Thoughts Let’s call this third picture of the relation between thinking and speech the refinement model. Arguably, it avoids the disadvantages of both the folk-psychological and Pandemonium models, by allowing us to explain the difference between parroting and nonparroting speech performances while circumventing the temptation to hypothesize a full-fledged language of thought. The Refinement Model shares with the Pandemonium Model the recognition that speech often outstrips our thoughts in content. But unlike the Pandemonium Model, the Refinement Model preserves both our folk-psychological taxonomy and the traditional view that speech acts express antecedent intentional states. Nonetheless, the Refinement Model appears to face a difficulty that both the Pandemonium and folk-psychological models avoid. As noted earlier, whenever we express our thoughts in words, our subjective impression is that what we say exactly matches the content of the intentional state we express. It never seems, from a first-person point of view, that our speech act is richer or more fine grained in content than our antecedent intentional state. The folk-psychological and Pandemonium models both reflect this, since both posit an exact correspondence of content between speech act and intentional state, differing only about whether the intentional state fixes the content of the speech act or the other way around. Either way we would seem to sense a perfect fit between them. Folk psychology generally trades, of course, on such conformity to our subjective impressions. And in this case the Pandemonium Model follows suit. Can we defend the Refinement Model despite its departure from our subjective impressions about thinking and speech? A slight detour here will be useful. As noted earlier, putting our thoughts into words sometimes seems to clarify those very thoughts. By itself, however, the clarifying effect of verbally expressing our thoughts does not tell against an exact match of content between thought and speech. It could simply be that the process of finding words to express our thoughts forces us to clarify those very thoughts. As we mentally hear ourselves say what it is that we think, we find it confused or unclear, and so revise on the fly what we think. But by adjusting our words as we go, we get the right thought out. The content of the
Content, Interpretation, and Consciousness
77
resulting speech act could still match exactly that of the suitably revised thought. There is, however, another way in which putting words to our thoughts might have a clarifying effect on those thoughts. Rather than leading us to clarify the thoughts themselves, it could result instead in our becoming clearer about just what those thoughts are. We would know better, having spoken, just what it was that we had been thinking all along. We discover what we think only as we say it. Dennett takes this kind of case to support the Pandemonium Model, on which the content of our thoughts is fixed only when we speak (CE 245); hence his striking epigram from E. M. Forster: “How do I know what I think until I see what I say?” (CE 193) But our discovering what we think only as we say it does not, by itself, support the Pandemonium Model, since it does not show that our thoughts lacked fixed content until we spoke. It could instead simply be that our thoughts often are not conscious in the relevant way until we express them in words. Even if the speech act results from an antecedent intentional state whose content exactly matches that of the speech act, perhaps it’s only as we speak that we become conscious of the intentional state as having that content. These considerations help with the problem the Refinement Model faced. On that model, our speech acts often outstrip the intentional states they express in respect of content. But that is never how it seems from a firstperson point of view. Our subjective impression is always that our speech acts exactly match in content the intentional states they express. But these first-person impressions may not accurately reveal the content of our intentional states. It could be that our speech acts do often outstrip our intentional states in content, but that we are nonetheless conscious of those intentional states as having the richer content. We might subjectively sense an exact match in content not because such a match obtains, but because that is how we are conscious of our verbally expressed thoughts. Because we are conscious of such a thought as expressed by a particular speech act, we in effect read back the content of the speech act onto the intentional state it expresses. We interpret the intentional state as having the content exhibited by the speech act. The idea that we interpret our thoughts in the light of our speech acts fits well with the spirit of Dennett’s Pandemonium Model. To some extent at least, we rely on the same considerations others do in determining what we think. As Dennett notes, in cases when we discover what we think only as we say it, “we are ... in the same boat as our external ... interpreters, encountering a bit of text and putting the best reading on it that we can find” (CE 245).
78
David M. Rosenthal
Hence Dennett’s heterophenomenological method, on which our theorizing about mind seeks to do justice to the verbal reports people make about their mental states.1 The Pandemonium Model, however, takes these self-interpretations to be the last word about the content of our thoughts.2 This squares with Dennett’s First-Person Operationalism, which rejects the idea that there can be a difference between how things seem and how they seem to seem (CE 132). Intentional states are subjective states; they are a matter of how things seem to us. And how things seem to us in virtue of our being in some intentional state is a matter of the content that state has. So there is no difference, on First-Person Operationalism, between the content one’s intentional states seem to one to have and the content they actually have. If it seems to one that an intentional state has a certain content, that’s the content it has. This approach saves an explanatory step. Why, if our speech acts outstrip in content the intentional states they express, do the intentional states subjectively seem to us to have the richer content of the speech acts? The Pandemonium Model avoids having to answer that question by positing that the intentional states really do have that richer content. But even if the content of our speech acts does outstrip that of the thoughts they express, we can readily explain why those thoughts and speech acts seem, subjectively, to have the same content. As remarked earlier, a speech act counts as expressing a thought only if it has roughly the same content.3 But we also 1
CE 72-85; also ch. 4, passim, and Dennett’s “Two Approaches to Mental Images,” in his Brainstorms, Cambridge, Mass.: MIT Press/Bradford Books, 1978, pp. 174-189.
2
The heterophenomenological method relies on the reports people make of their mental states; such reports in effect constitute self-interpretations. Dennett also regards cases in which we learn what we think by seeing what we say as involving a kind of self-interpretation (CE 245). But there self-interpretation occurs by way of speech acts that verbally express our intentional states, rather than explicitly reporting them. Seeing verbal expressions of intentional states as self-interpretations as well as reports suggests that Dennett may be tacitly assimilating reports to verbal expressions. And that, in turn, may help explain why he regards self-interpretation as the last word about the content of our thoughts. Even if speech acts can differ in content somewhat from the intentional states they express, one might plausibly regard such expressions as better than any other evidence could be about the content of those states. So, assimilating reports of intentional states to their verbal expressions will make such self-interpreting reports themselves seem to be decisive about the content of the states those reports are about. Assimilating the reporting of intentional states to their verbal expression also encourages the idea that mental states are all conscious, since the content of every mental state would itself then affirm, as reports do, that one is in that state. This idea may, moreover, be implicit in first-person operationalism, on which a state’s mental properties are determined by how that state appears to consciousness.
3
It is arguable that, in addition, the intentional state must be causally implicated in producing the speech act. See my “Intentionality,” Midwest Studies in Philosophy, X (1986), pp. 151-184.
Content, Interpretation, and Consciousness
79
sometimes convey our thoughts by actually reporting them – by saying literally that we have those thoughts. Suppose I think that it’s raining. I verbally express that thought by saying, simply, that it’s raining, whereas I report that very same thought by saying, instead, that I think that it’s raining. Indisputably, two such speech acts differ semantically, since they have distinct truth conditions. Still, they are easily conflated, since, with minor qualifications about degree of conviction that won’t matter here, we can appropriately say one thing whenever we can appropriately say the other. This equivalence of performance conditions, moreover, is second nature for us; any time I actually do say it’s raining, I might as easily have said that I think that it’s raining, and conversely. Suppose, then, I think that p and I express my thought with the somewhat richer, more refined statement that p’. The richer statement that p’ is, then, performance conditionally equivalent to the statement that I think that p’. And because this equivalence is second nature, I might as easily have said that I think that p’. But if I had made the higher-order remark that I think that p’, my statement would have expressed a thought that has roughly the same content: the HOT that I think that p’. And, since I might just as easily have made that statement, I must have had the HOT that the statement would have expressed. The upshot is that, whenever I say that p’, I have a HOT that I think that p’. And that HOT determines how I am conscious of the thought my speech act expresses. So, whatever the actual content of my thought, I am conscious of that thought as having the content; that p’. We can explain why, even though my speech act may be somewhat richer and more refined in content than the thought it expresses, I am conscious of the thought as having the richer, more refined content.1 But a difficulty looms for this explanation. If first-order conscious thoughts can be less fine grained than the speech acts that express them, why
1
These considerations explain, more generally, why it is that, all verbally expressed cognitive intentional states are conscious. Whenever I express my thought that p by saying that p, I could as easily have said that I think that p. So I must have had the HOT I think that p, and on the HOT model my thought that p will accordingly be conscious. It is unlikely that any other model can explain why verbally expressed cognitive states are conscious. For more, see my “Why Are Verbally Expressed Thoughts Conscious?”, forthcoming in a collection of my essays tentatively entitled Consciousness and Mind, Oxford: Clarendon Press. In “Consciousness and its Expression," Midwest Studies in Philosophy XXII (1998), forthcoming there and in Consciousness and Mind, I show how these considerations also explain why affective states, unlike cognitive states, are often verbally expressed without being conscious.
80
David M. Rosenthal
can’t the same happen with HOTs as well?1 And if our verbal report of a first-order thought might outstrip in content the HOT that report expresses, the HOT need not, after all, reflect the richer content of that report. And then we wouldn’t be conscious of the first-order thought as having that richer content. Perhaps this is all so. But recall that we are not trying to show that speech acts invariably have more fine-grained content than the thoughts they express, but only that this may sometimes happen, despite our subjective sense that it never does. And we have is no reason to think it ever does actually happen with HOTs. Indeed, it is highly unlikely that it ever happens there. We think that first-order thoughts have less fine-grained content than their verbal expressions because it seems clear that what we think could have been expressed equally well by distinct, semantically nonequivalent speech acts. But that’s not the case with the HOTs that accompany our verbally expressed intentional states. Suppose I verbally express my first-order thought that p by the more fine-grained remark that p’. How I express my HOT about my less fine-grained, first-order thought that p is now dictated by the performanceconditional equivalence between saying the more fine-grained p’ and saying that I think that p’. So I will now express that HOT only by saying that I think that p’. Verbally expressing my thoughts constrains the way I am conscious of them.2
IV. The Refinement Model and First-Person Operationalism On the Refinement Model, speech acts often outstrip in content the intentional states they express. To explain how this squares with our subjective impressions, we must distinguish the content those states have from the content they seem to us to have. But the content of intentional states is a matter of 1
This echoes a challenge Dennett raises for the HOT model, that that the Pandemonium Model should apply at higher levels as well; “the second-order state,” he claims, “comes to be created by the very process of framing the report” (315, emphasis Dennett’s). Also: “The emergence of the [verbal] expression [of a higher-order thought] is precisely what creates or fixes the content of higher-order thought expressed” (315).
2
Dennett notes with approval Elizabeth Anscombe's argument that it is “wrong to claim that we know what our intentions are; rather we just can say what our intentions are” (CE 315, n. 10; see G. E. M. Anscombe, Intention, second edition, Oxford: Basil Blackwell, 1963). The present explanation reflects this primacy of verbal expressions. On that explanation, the verbal expression of a thought constrains, by the performance-conditional equivalence, the content of our HOT about the target first-order thought, and hence how we are conscious of that target thought.
Content, Interpretation, and Consciousness
81
how things seem to us. So this explanation conflicts with Dennett’s First-Person Operationalism, which denies any difference between things’ seeming a certain way and their seeming to seem a certain way.1 But there is reason to reject that denial. Consider the game Dennett describes of “Hide the Thimble,” which dramatizes how we can look straight at an object we're trying to find and yet fail consciously to register it (CE 334). This kind of case invites us to distinguish our seeing something consciously from our seeing it without being conscious of seeing it; things may seem a certain way even though they don’t consciously seem that way. The striking and subjectively surprising limits on parafoveal resolution that Dennett cites take us even farther. Parafoveal vision can produce only low resolution sensations of most of the Warhol Marilyns,2 but it seems subjectively that we are aware of them all in a clear and focused way. What it’s like for one to have a particular conscious sensation is a function of how one is conscious of that sensation. So the best explanation of this case is that we have blurry parafoveal sensations of most Marilyns, but the way we are aware of those blurry sensations represents them as having high resolution.3 What our sensations of the Marilyns is like for us is a function of how we’re conscious of
1
The hierarchy of levels ends up, he claims, in our “having to postulate differences that are systematically undiscoverable by any means, from the inside or the outside,” distinctions that are “systematically indiscernible in nature” (CE 319).
2
CE 354. See CE 53-4 for Dennett's striking illustration of the limits on parafoveal vision in attempting to discern the color of playing cards seen parafoveally at arm’s length.
3
Similarly, in the thimble case; there is a conscious sensation in our visual field corresponding to the location of the thimble, since no subjective gap occurs there, but we are not conscious of that sensation as a sensation of a thimble. We may also be conscious of sensations in ways that leave out aspects of their qualitative character, as when a throbbing pain is conscious only as painful and not as throbbing, or a sensation of red is conscious as red but not in respect of any particular shade. So, even though visual sensations appear to us to be “ultimately homogeneous,” in Wilfrid Sellars' useful phrase, it may well be that their sensory qualities are actually composed of many pixels representing specific characteristics; their ultimately homogeneous appearance may be due only to the way our HOTs represent collections of such pixels. We are conscious of our sensations in a way that smooths them out, so to speak, and elides the details of their particulate, bit-map nature. (Wilfrid Sellars, “Philosophy and the Scientific Image of Man,” in Frontiers of Science and Philosophy, ed. Robert G. Colodny, Pittsburgh: University of Pittsburgh Press, 1962, pp. 35-78; reprinted in Science, Perception and Reality, 1-40, p. 36.) For more on sensations' diverging from the way we are conscious of of them and the way HOTs function in that connection, see “State Consciousness and What It's Like," forthcoming in Consciousness and Mind, and "Consciousness and Metacognition,” in Metarepresentation: Proceedings of the Tenth Vancouver Cognitive Science Conference, ed. Daniel Sperber, New York: Oxford University Press, forthcoming.
82
David M. Rosenthal
those sensations.1 And, if the way we're conscious of our sensations sometimes goes beyond their mental properties by in effect refining them or touching them up, it is highly likely that the same happens with our intentional states, as well.2 On the Pandemonium Model, our intentional states come to have determinate content only when expressed by a speech act, or fixed by some other sort of probe. This applies equally, Dennett argues, to the higher-order states in virtue of which we are conscious of our conscious states; those higher-order states themselves “[come] to be created by the very process of framing the report.”3 Such “heterophenomenological reports” are the verbal pronouncements people make about their own mental states, to which Dennett’s heterophenomenological method seeks to do justice. Since framing those report fixes the higher-order content, Dennett holds that sincere heterophenomenological reports are constitutive of what it’s like for the subject at that time and, hence, constitutive of that subject's consciousness. It is those heterophenomenological reports, rather than any antecedently occurring intentional states, that fix the contents of consciousness. This conclusion might seem tempting because we are seldom conscious of the higher-order states in virtue of which we are, in turn, conscious of our conscious mental states. So our first-person impression will be that heterophenomenological reports alone determine how we are conscious of those states. But this first-person appearance is unreliable. We are in many intentional states of which we are not conscious, and we should expect that we typically remain unaware of the HOTs in virtue of which we are conscious of our conscious states. Such HOTs would be conscious only if we had third-order thoughts about them, and we can safely assume that that seldom happens.
1
Consider the striking results John Grimes reports, in which subjects fail consciously to notice dramatic changes of color and shape in a salient object if the change occurs during a saccade (John Grimes, “On the Failure to Detect Changes in Scenes across Saccades,” Perception, ed. Kathleen Akins, New York: Oxford University Press, 1996, pp. 89-110). Our subjective sense that conscious experience is continuously responsive to changes in what we see outstrips the extent to which we actually track changes in our visual sensations. This lends support to the conclusion, argued for by Dennett (CE ch. 11), that our subjective sense of great detail throughout our visual field is also illusory. The way we are conscious of our sensations goes beyond those sensations themselves.
2
Because what it's like to be in conscious states is sometimes informationally richer than the states themselves and that additional informational content is occurrent, we must posit occurrent states of higher-order awareness of the states, and not just dispositions to be aware of them.
3
CE 315, emphasis Dennett's. Also: “The emergence of the [verbal] expression is precisely what creates or fixes the content of higher-order thought expressed” (CE 315).
Content, Interpretation, and Consciousness
83
Moreover, we are occasionally conscious of our HOTs, wholly independently of whether we express them in heterophenomenological reports. Focusing introspectively on some particular state makes us aware not only of that state but also of the reflective, higher-order state of being aware of the lower-order target. The only reason to hold that heterophenomenological reports, rather than the HOTs they express, fix the contents of consciousness is Dennett’s rejection of the folk-psychological model, on which speech acts generally express antecedent intentional states. HOTs not only determine what it’s like for us to be in various mental states; they also are, in effect, subjectively spontaneous interpretations of the mental states we are in.1 Indeed, there is an important connection between consciousness and our seemingly spontaneous self-interpretations. When we spontaneously interpret ourselves as being in certain states, we are conscious of ourselves as being in those states. What it’s like for us to be in particular states is a function of how we spontaneously interpret those states. In the absence of any such spontaneous self-interpretations, our mental states simply aren’t conscious; there is nothing it’s like for us to be in them. First-Person Operationalism is also interpretationist, but it imposes a special constraint all its own. Mental states are conscious, on First-Person Operationalism, only when accompanied by intersubjectively accessible probes. Because we have no access to the mental states of others apart from these probes, such probes provide our sole basis for interpreting what states others are in. So tying the consciousness of mental states to the occurrence of probes means that mental states are conscious only when they are available for interpretation by others. First-Person Operationalism holds that the consciousness of mental states is a function of interpretation, but only third-person interpretations matter.2 But this third-person constraint is gratuitous. The self-interpretations that HOTs provide are by themselves sufficient for one’s mental states to be conscious. Indeed, one might well wonder how third-person interpretations could have any bearing at all on the consciousness of mental states. The key is the denial by First-Person Operationalism of any difference between mental states and their being conscious--between how things seem and how they seem to seem. Given that denial, an interpretation of somebody as being in a particular state serves indifferently as reason to think that the person is in that state and 1
Subjectively spontaneous because we are unaware of any antecedent factor that might explain their occurrence. Because HOTs are not themselves typically conscious, we will seldom be explicitly aware of engaging in the self-interpretations these HOTs embody.
2
And such third-person interpretations cannot distinguish between a state’s never coming to be conscious and its coming to be conscious too briefly to affect memory.
84
David M. Rosenthal
that the state is conscious. First-Person Operationalism restricts itself to third-person interpretations because it denies the distinction between mental states and their being conscious. Rejecting First-Person Operationalism allows us to distinguish being conscious of our mental states from the states we are conscious of. And that opens the way to explaining the consciousness of mental states along lines different from those invoked in explaining other mental properties. In particular, it may well be that mental states have determinate content even though the way we are conscious of those states is a function of the subjectively spontaneous self-interpretations embodied in HOTs. So, too, for the HOTs themselves; what it's like for us to be in various mental states is just a matter of the self-interpretations our HOTs embody, but those HOTs can themselves be determinate in content.1 Interpretationism can be true about what it’s like for us to be in mental states without thereby holding for their content or other mental properties.2 Rep. from Dennett’s Philosophy: A Comprehensive Assessment, edited by Don Ross, David L. Thompson, and Andrew Brook, MIT Press, 2000.
1
As well as determinate as to time of occurrence, which makes it determinate when our mental states come to be conscious. The HOT model may sometimes count a state as conscious even though there is nothing it's like for one to be in that state, e.g., in some cases where we find it hard to decide between Stalinesque and Orwellian explanations. The occurrence of a HOT would be decisive in any event, regardless of first-person impressions; it is standard for well-established theories to resolve problem cases when the empirical input, including first-person impressions, does not suffice.
2
Work on this paper began at the Center for Interdisciplinary Research (ZiF), University of Bielefeld, Germany. I am grateful to the Center for congenial and stimulating surroundings and to a PSC-CUNY Research Award for supporting my stay there. I am also grateful for helpful reactions to an earlier draft from the participants in the November 1998 Memorial University of Newfoundland Conference on Dennett and from members of the CUNY Cognitive Science Symposium.
Thought as Language: A Metaphor Too Far
85
JAY L. GARFIELD Thought as Language: A Metaphor Too Far Abstract Language has often served both as a metaphor for thought. It is highly plausible that language serves as an epistemic entre into thought and that language structures adult human thought to a considerable degree. The language metaphor is, however, uncritically extended as a literal model of thought. This paper criticizes this extension, arguing that thought is not literally implemented in language and distinguishing legitimate from illegitimate uses of language as a device for understanding thought.
How does the philosophical problem about mental processes and states and about behaviourism arise? – The first step is the one that altogether escapes notice. We talk about processes and states and leave their nature undecided. Sometime, perhaps, we shall know more about them – we think. But that is just what commits us to a particular way of looking at the matter. For we have a definite concept of what it means to learn to know a process better. (The decisive move in the conjuring trick has been made, and it was the very one that we thought quite innocent.) – And now the analogy which was to make us understand our thoughts falls to pieces. So we have to deny the yet uncomprehended process in the unexplored medium. And now it looks as if we had denied mental processes. And naturally we don’t want to deny them. (Wittgenstein, Philosophical Investigations) Yet maybe there are patterns we can detect in the motion of the sand, eddies and flows within its ceaseless flux and ebb. Not skeletons or truth, but feelings, encoded bumps and ridges that might make letters, then a word, then a sentence or maybe more, a secret Braille of touch and faith. A quiet urgency of form. Or they might be no more than they seem. (James Bradley, Wrack, p 9)
1. Introduction: Origins, Motivations and some Puzzles I have been arguing for years that the intentionality of language is conceptually and epistemically prior to that of thought and that propositional attitudes can only be understood on the analogy of overt utterances. I don’t wish to retract any of that now, but I must confess that this commitment has led to my having awakened at times to find strange bedfellows. Depending on their other
86
Jay L. Garfield
dalliances, these have included members of the LOT (language of thought) club, instrumentalists, eliminativists and indicator-semanticists of various stripes. So, in the interests of greater selectivity regarding the conceptual company I keep, I want to circumscribe these commitments, to provide a critique – in the Kantian sense of that term – of the metaphor of language as applied to an understanding of thought. That such a critique is necessary should have been obvious from the beginning. For, as Wilfrid Sellars was fond of saying, “All metaphors limp.” And as a tool for understanding thought, language cannot help but be used metaphorically. After all, language is, in its literal sense, an artefact of human communication. It is literally at home in speech, and, pace Derrida, by extension in writing and in the development of artificial languages. (One might even argue that languages are best understood as abstracta, but I prefer to leave these metaphysical disputes to the side for now.) Even a hard-core Chomskyan–on any charitable interpretation of that doctrine – when she says that language is “in the brain” or that the mind contains a “language organ” indicates by those expressions the mechanisms subserving the production and comprehension of language, and not any language itself. (Cowie 1997) And so, when we characterise thoughts as inner speech, or when we speak of a belief that p as standing to p as does the assertion that p we trade on a metaphor, and we must be prepared to say what features of the metaphor we take seriously, and what features play no role. Failure to handle metaphors with care can get theorists in trouble. Clark Hull’s commitment to a physical forces as a metaphor for the determinants of behaviour blinded him to the utility of representations; Tolman’s commitment to the map prevented him from considering linguistic representation; the von Neumann metaphor for cognition inhibited the development of non-classical models of cognition, etc... Bad metaphors generate bad questions and sterile debates. How best to measure ether drag? Which tree pruning algorithm do humans use to play chess? On the other hand, metaphors are indispensable tools of theory, and this particular metaphor has been particularly fecund in cognitive science and in the philosophy of mind. It is an important fact about thought that it involves the propositional attitudes, and an important thing about the attitudes that they, like assertions and other linguistic acts, take propositions as their objects. It is an important fact that thought, like language, is both intensional and intentional. It is also a striking fact about human cognition that it is both specialised for and highly structured by, language. And for too many reasons to rehearse here, I believe that we only come to be able to attribute thoughts to ourselves and others, and only able to enter into the full range of human thought, through language.
Thought as Language: A Metaphor Too Far
87
It is through reflections on topics such as these, initiated by Frege and brought to full flower by Sellars, that we – here I include the vast majority of contemporary philosophers of mind and cognitive scientists – have come to take language for granted as a model of thought. That does not mean, of course, that there is any unanimity on what that taking-for-granted amounts to. For some that means that we think in natural languages (Carruthers 1996, Davies 1996); for others that there is a language of thought awaiting discovery by cognitive scientists (Fodor 1975, 1997, Devitt 1996). Others assiduously seek to produce that code or at least to identify the more primitive icons that might transduce it (Millikan 1993, Dretske 1995). For still others, the failure to find such brain writing entails the very non-existence of thoughts themselves (Churchland 1989). This is not an exhaustive survey of the conceptual space this contention generates, of course. At this point I only want to indicate some of the consequences of seduction by this siren metaphor and the consequent pre-reflective attribution of spurious properties to that to which the metaphor is meant to give us access. For when a metaphor becomes too entrenched it can become invisible, and that is when seduction becomes the greatest danger.
2. How the Realism vs Instrumentalism Debate Got Stirred In The language metaphor has become entangled with debates about intentional realism. This tangle is unfortunate, and makes it hard to discuss either the structure of thought or the philosophy of cognitive science with any clarity. So we will begin with a bit of unravelling. One obvious way to cash out the language story is that adopted by Fodor and his followers–to argue for a LOT hypothesis. If there were a LOT, and hence a set of LOT tokens to be found in the brain and a set of processes defined over them constituting cognition, this would be a fine vindication of a serious realism about the propositional attitudes and related psychological phenomena. These would be the very real entities about which such a realist might dream. At the same time it would cash out the language language in the most literal imaginable way. The converse, of course, does not follow. Converses never do. It does not follow, that is, that intentional realism entails the LOT hypothesis. Nor does the mere theoretical utility of language language in psychology entail that strong empirical hypothesis. But just as converses never follow, they seem, in cognitive science always, albeit illicitly, made more plausible. And so it is in the present case. Proponents of the LOT hypothesis often argue that if there is no LOT then we must give up realism about the attitudes, in virtue of the
88
Jay L. Garfield
absence of anything in the brain or mind that could be both a real entity and a propositional attitude. The premise that would turn this unconvincing enthymeme into a pretty good methodological argument, of course, is this: No psychological phenomenon besides tokens in a LOT could underwrite intentional realism. And this premise is hotly contested. (See Fodor 1997, Dennett 1987, Garfield 1997, Garson 1997.) I note this only to set it aside, as this essay is not directed at the LOT hypothesis. But the LOT debate is only an instance of the way that the language metaphor has structured debates about realism. More generally, we can note that eliminativists and instrumentalists defend these positions with regard to the attitudes by urging that when we look at the brain, or at plausible models of sub-personal cognitive processes, we find no physical or functional states that look sentential or representational of propositional content. Intentional realists, on the other hand, are either content merely to promise such tokens or provide concrete, though typically peripheral or primitive examples of such tokens complete with mini-psychosemantics for them. This is not the place to adjudicate any instance of this debate. I only note that however vigorously it is prosecuted, each relevant pair of antagonists shares a commitment to the claim I am interested in interrogating, viz., that intentional realism stands or falls depending on whether or not we find sentence-like sub-personal tokens in our cognitive psychology or neuroscience. (Fodor 1997, Stich 1983, Millikan 1993) We should, of course, be dubious of any such claim. For it suggests that the only appropriate level of analysis at which we might find the attitudes, or representations more generally, is the sub-personal. Many phenomena, however, appear at a personal level of description and analysis, fail to reappear at lower explanatory levels, but nonetheless must be counted as real for relevant sciences and in ordinary epistemic and social life. Kinship relations, fiscal status, citizenship, etc are the most obvious examples. If representational phenomena, including the propositional attitudes are among this class–as I have argued elsewhere (Garfield 1988) that they must be–there is no reason to think that questions about the depth to which the language metaphor can be pushed bears upon questions about intentional realism. We can take the PAs seriously even if nothing remotely PA-like or even representational appears at any sub-personal level. Dennett is eloquent on this point (See also Garfield 1988, esp pp 130-156.): But I do not believe that [the hypothesis] is obvious. Those who think it is obvious, or inevitable, that such a theory will prove true... are confusing two different empirical claims. The first is that intentional stance description yields an objective, real pattern in the world.... The second is that this
Thought as Language: A Metaphor Too Far
89
real pattern is produced by another real pattern roughly isomorphic to it ... Doubting the existence of the second real pattern is not doubting the existence of the first. (“True Believers,” Intentional Stance, p 34) Now, I think that realism is the only respectable attitude in the philosophy of science, including the philosophy of cognitive science. I won’t argue for that here, nor will this commitment play a role in the subsequent argument. But I wish to emphasise the consistency of my considerable circumscription of a metaphor often taken to underpin such realism with as robust an intentional realism as anyone could want. I now turn more directly to the metaphor.
3. What Languages are Like, and Why Thoughts Aren’t Like THAT Getting clear about the utility and limits of the language metaphor requires first that we are absolutely clear about what it would be to take the language hypothesis of thought literally. That would be to adopt the full-blown language of thought hypothesis. We know what languages look like, more or less. Most of us even speak at least one, and linguistic theory tells us quite a bit about the less obvious but nonetheless important structural features of human natural languages. I focus on human natural languages here, rather than the more formal artificial languages of logicians, mathematicians and computer scientists simply because human natural languages are the only languages with which we are familiar with sufficient expressive power to capture all that we would want to capture in a theory of thought. Some LOT theorists, of course, even suggest that we literally think in our native natural languages. (Davies 1996, Carruthers 1996) But even those who posit a more universal language of thought argue that it must have all of the conceptual and grammatical resources of a human language, and must be processed by mechanisms sensitive to those resources and defined syntactically. I will therefore use the term “language” quite self-consciously to refer to an articulated symbol system that shares at least most of the structural features of a human natural language. A language then, has a syntax, including a primitive vocabulary, a recursive set of formation rules for generating wffs from atoms and an algorithm for generating canonical syntactic representations of its wffs. Any symbol system with less structure than that is at best a code, and has no hope at achieving the systematicity compositionality and power of a real language. A language has a semantics. In the relevant sense, this means not simply that there is a mapping from its strings to some arbitrary domain or other, such as the natural numbers, but that that mapping is nontrivial in an ecological sense (and of course to specify this sense would be roughly to solve all outstanding problems in
90
Jay L. Garfield
philosophy) and cognitively available to the language user. To put the point bluntly, if we do not know what we are talking about, or what our interlocutors are talking about, however mellifluous and ornate the noises or gestures we produce, we are not speaking a language. A language also has what is usually called a “phonology,” though I prefer the more general “morphology” in order to comprise non-spoken languages as well, and so as not to beg any trivial questions in the present case. The important point here for our purposes is simply that there must be a way to get from physical to syntactic tokens and types in the language. This seemingly obvious demand is often overlooked by those who argue that the brain/mind must, in order to make sense of functionalist physicalism, be a syntactic engine. It must be a physical engine, of course, but the physical and syntactic levels must be kept distinct in cognitive psychology and the philosophy of mind just as they must in linguistic theory. And it in no way follows from the fact that cognitive operations must be characterised physically in order to explain embodied mind that they must be characterised syntactically. Finally, a language is characterised as well by a pragmatics: Languages are used for a variety of speech acts, which in turn have a multitude of felicity conditions. Not to distinguish between questioning, asserting and commanding, between describing and commending, between greeting and explaining or between literal and metaphorical description is to fail to speak a language with full competence. None of these four requirements is controversial, and none of what I have just said is news to anyone who has thought seriously about language in the framework of cognitive science. But we should now ask whether it is plausible, justified or theoretically fruitful to say that thought satisfies any or all of these requirements. I shall argue that it is at least a highly dubious claim to say that it does, and that the intuitions and commitments that have led so many to assume that it must can be saved without supposing that it does. Suppose I say of my daughter that she believes that wallabies are friendly. I have ascribed her a thought. Perhaps it is occurrent, as she approaches a wallaby. Perhaps dispositional. After all, this attribution is one that I would now defend despite the fact that she is probably asleep at this very moment. In either case I thereby indicate an attitude towards a proposition. Perhaps even a complex dispositional attitude towards a sentence, or an equivalence class of sentences in bona fide languages, such as English, Tibetan or Chinese. But there is no warranted inference from that fact to the presence of any syntactically structured object in her mind or brain that would be a member of such an equivalence class. Consider a few possibilities: Connectionism might be true. (Smolensky 1990) Non-linear dynamic models of thought might be
Thought as Language: A Metaphor Too Far
91
true. (Van Gelder 1995, Garson 1997) Or it might be that attitudes toward simply consist either in complex dispositions or in some yet-to-be characterised occurrent states with specifiable causal properties which properties explain our standing in propositional attitudes, but nothing like a syntax as it is understood with regard to languages. (Garfield 1988, Dennett 1987) States of our immune systems, digestive systems and respiratory systems are like that. Why not of our central nervous systems? Remember, if taken literally, the syntactic requirement is quite stiff, and there is to date not a shred of empirical evidence for a cognitive/neural syntax. Dennett remarks on this point: When cognitive scientists have talked about representations, they typically have committed themselves to some view about syntax... This heavy commitment ... is made solely on the basis of aprioristic reasoning, for... while there has been plenty of interesting speculation, there is almost no empirical evidence yet that tends to confirm any substantive hypothesis about the nature of the supposed syntax of mental representation. [“Styles of Mental Representation,” IS, pp 215-216]
What should we say about psychosemantics? Suppose that we can provide a realistic, naturalistic account of mental representation without a mapping from sub-personal states to propositions or their constituents. I will argue for this possibility below. But think for a moment about what our sub-personal states are for: They must guide behaviour, process perceptual information on line, subserve personal cognitive states, and generally maintain our functional coupling with our environments. We can certainly imagine accomplishing this by means of a mechanism that involves structured states sustaining a propositional interpretation. After all, that is how our societies, great and small, accomplish much the same task, through the use of language as a coordinating device. (See Millikan 1993.) But this is a useful strategy in the social domain precisely because of the need to co-ordinate distinct organisms, each with clearly demarcated and relatively opaque interface boundaries. That need is absent in the mind, and there are plenty of ways to regulate the behaviour of systems in environments that do not involve articulated interpreted states. (As I argued in reply to Millikan in 1989.) Thermostats, of course, are the standard examples. But if we examine the homeostatic mechanisms that regulate our body temperatures, posture, that co-ordinate limb orientation in locomotion, maintain perceptual constancy, etc, we see any number of autopoetic and coupling mechanisms that proceed without the aid of interpreted or interpretable discrete states, and without any interpreter to read them. A morphology is necessary only to the extent that it is important for a system to map physical tokens to syntactic types for the purpose of subsequent formal processing and interpretation. Again, it is obvious that in the case
92
Jay L. Garfield
of language, where we must map a wide range of physically disparate tokens into syntactic equivalent classes, given a variety of speech patterns, fonts, etc, a morphological theory is a sine qua non. But when we theorise about the mind and seek underlying processes to explain and to naturalise intentionality and intentional processes, we seek physical states. For our brains, nervous systems, perceptual and motor apparatus are one and all physical. To posit a mapping from a physical to an intermediate syntactic description as a necessary mediator of mentality is at least gratuitous, particularly given the absence of any positive evidence for such a mapping from cognitive neuroscience. To point out that there is no pragmatics of thought seems like a cheap shot. And it is. (But not a totally unnecessary one, as Millikan 1993 makes clear.) The only reason I take it here is to drive home the point that our language-talk about thought is always metaphorical. To take it seriously would entail trying to make sense of a pragmatic theory as well. The point of these reminders is not to provide a knock-down argument against the LOT hypothesis. Nor is it to argue that we can dispense with a representational theory of mind. Rather it is to point out that in our use of language as a paradigm of representation, we must be cautious about presuming that the many the features distinctive of language, per se, should be sought in cognitive processes.
4. Intentionality and Causality Among the principal motivations for adopting the language metaphor is that language provides our best epistemic access to the phenomenon of intentionality so definitive of thought. Arguably we come to our conception of a thought that wallabies are friendly, and its logical and semantic properties on the model of our assertions that wallabies are friendly. (Sellars 1963, Garfield 1988) And indeed, given the demand that cognitive science naturalise intentionality in the cognitive domain we have witnessed over the past few decades a spate of proposals for the analysis of the intentionality of thought in terms of non-intentional phenomena. (Millikan op. cit., Dretske op. cit., Fodor op. cit., Devitt op. cit.) None has been spectacularly successful, to say the least. Again, I think a bit of caution in the move from intentionality at the personal level to the posit of intentional states and processes at the sub-personal level is in order, and a relaxation of this expectation can indeed prove theoretically liberating. Language achieves its intentionality through being rule-governed, in ways articulated in detail by Wittgenstein and Sellars, among others. Its meaningfulness depends upon the observation of regularities in usage that are enforced
Thought as Language: A Metaphor Too Far
93
by and reflect norms established through consensus. So much is beyond serious dispute. The greatest challenges to programs in indicator semantics, information-theoretic semantics, proper function theories, and other such approaches to psychosemantics derive from the difficulty or impossibility of providing a substitute foundation for that intentionality. Without exploring these programs in detail in this venue, let me just say that there is something right-headed and something wrong-headed in these approaches. Here is what is right-headed: There must be some physical, non-intentional explanation of how it is that we manage to get ourselves into intentional states. The alternatives are dualism or black magic. Moreover, that explanation undoubtedly adverts to sub-personal abilities to track events and objects in our environment, to control movements, and to maintain internal dynamics so as to facilitate such perceptual-motor behaviour in the future. And all of that requires that there be states of our nervous systems that are rather tightly coupled to specific properties of our environments. But here is what is wrong-headed: Noon whistles and noon trains are reliably coupled, but not semantically related. One, however, can be used to track the other, just in virtue of reliable coupling, and someone wishing to meet the noon train just might use the whistle for that purpose. Tree rings and tree ages are well-coupled, but not semantically related, and are useless to anyone except botanists. And these failures of semantic relation simply reflect the lack of a conventional or norm-governed practice for using one member of a locked pair to mean the other in the context of a communicative system. Just so with cognitive states and processes that indicate, or carry information in Dretske’s sense, or that have proper functions related to organism-environment interface in Milikan’s sense. One can grant virtually all of the naturalistic side of Dretske’s and/or Milikan’s programme, thereby granting that there are sub-personal, cognitively crucial processes that carry information about the environment, about action, and about other cognitive states and processes, and processes and states that have proper functions related to tracking the environment, guiding action, effecting cognition, etc, and still be nowhere near granting that any of these states are in fact meaningful in the linguistically relevant sense, or that any analysis of meaning can be built upon them as a foundation. The claim here is not that psychosemantics is impossible, but that it is beside the point. It is beside the point precisely because the enterprise of semantics has its point at a level of description where we have a language, whose expressions have meanings determined by and determining the use of its tokens. Absent such a phenomenon, we can fruitfully discuss important causal regularities, functional roles, and information carried by subpersonal states, and can take such talk
94
Jay L. Garfield
perfectly literally. But to import fully semantic talk to that level is at best metaphorical, and can at worst mislead.
5. Level-Headed Head-Level Realism about Representation My parsimonious refusal, without powerful empirical evidence, to grant representational status to subpersonal states should not be regarded as hostility to mental representation, or to the propositional attitudes. Far from it. I cannot imagine an account of human cognition proceeding without adverting to the propositional attitudes, or a psychology that does not take seriously the fact that organisms, including not only humans but sophisticated vertebrates engage in complex mental representation. Most of my professional philosophical life has been devoted to the defence of such a view, and I retract not one bit of that defence. (Garfield 1983, 1988, 1989, 1990, 1996, 1997) The question that must be faced, however, concerns the level of description at which such mental representation can properly be said to occur. And this is one important spot at which the language metaphor threatens to lead us astray. Let us follow the garden path for a moment. We really do have propositional attitudes. So some of our psychological states are directed on propositions and are properly characterised as sharing content with linguistic items. These attitudes are enabled by subpersonal cognitive processes and states which covary systematically with our environment, and are systematically related to one another. So those states, in virtue of their – albeit simpler – functional roles must also be contentful, and so can be characterised intentionally, and so as having propositional content, and the rest follows. But compare the following not-so-tempting, but structurally similar garden path: Our written sentences have propositional content, and they are enabled by words, which have lexical roles connected with objects in the environment, and whose occurrences are related systematically to one another. So these words must also be contentful, and perhaps each even has propositional content. Pretty bad so far. But let’s continue. Each of these words comprises letters, which comprise lines, curves, serifs. These, of course, must have content, since they underlie the contentful words ... You get the point. And so back to the mind: Realism about representation at the personal level can be underwritten with a commitment to a sound proper function or indicator account of subpersonal mental states which stops well short of full propositional characterisation, and more fundamentally – or perhaps even only by – a purely causal story about continuous biological or fundamental psychological processes coupled to the environment. That is, while it is quite plausible that
Thought as Language: A Metaphor Too Far
95
our representational states – prominently including the propositional attitudes – and the transitions between them are explained by states which can be described as indicative of stimuli or responses with which they covary, or which can be characterised as having cognitive proper functions, the story about such indication and function can be a purely causal and not a semantic story. And that causal story can be explanatory with respect to representation without urging that the intentional character of the representations they explain is thereby exhausted. And such a two-stage account of cognition can respect realism at both levels–the subpersonal and the personal.
6. Not Instrumentalism; Not Behaviourism; Not Eliminativism Nor am I any kind of instrumentalist, behaviourist or eliminativist in virtue of this denial of the linguistic character of fundamental cognitive processes. Each of these is a silly doctrine. To develop a detailed critique of any of them here would take us far afield. (I develop detailed attacks on these views in 1988 and in 1997.) Let me just indicate in the most general way why, by withholding the linguistic metaphor from subpersonal levels of description, I am not committing any of these metatheoretic sins. Each of these three views, in its own way, requires us to deny the degree of reality to propositional attitudes, or to mental representations quite generally that we would ascribe to physical states, to dispositions, or to behaviour. One might think that the view I am here suggesting is committed to some such doctrine in virtue of my denying what many other realists assert – that there are inner subpersonal cognitive tokens which can fruitfully be characterised as having representational content, and which are the objects of formal cognitive operations. To be sure, were one to be committed to the existence of such states as the extension of the language metaphor to that level of analysis would suggest, one would be committed to a robust realism. But withholding commitment to such a set of phenomena in no way commits one to any kind of anti-realism. For realism or one of its contraries is always a view about a specific set of entities. And entities posited or observed at one level of description do not necessarily reappear at more fundamental levels of analysis or explanation. (See Garfield 1988, Hadrcastle 1995 and Dennett 1987 for more on this point in the philosophy of science.) Budget deficits don’t reappear as physical objects; social networks don’t re-emerge as tissues or organs. This is by now old news and these are becoming shopworn analogies. And the realism I endorse – and the only one relevant to the possibility or necessity of an intentional psychology – is a realism with regard to representations,
96
Jay L. Garfield
including prominently those that have propositional objects. These representations, to which the language metaphor is quite appropriate, are indispensable posits of any scientific or commonsense view of human beings described at the level of description at which much of our action, thought and perception is explained. It is impossible to imagine proceeding without them or seriously denying their reality. And asserting their reality at the personal level – as well as their propositional content – in no way entails asserting the reality of similarly contentful phenomena at the more fundamental neural or functional psychological levels of description of the systems that enable them.
7. Three Spurious Debates Dissolved What do we gain by abandoning the language metaphor at subpersonal levels of cognitive theory? Better to ask what we lose. I think we lose a lot of debates, which, while providing content for contemporary journals and reading material for tenure committees, in the end tell us little about how the mind works. I am not sure how many spurious debates in the foundations of cognitive science dissolve when this metaphor is withdrawn from their underpinnings, but there are three that come to mind immediately: the debate about whether thought occurs in natural language vs the language of thought; the simulation theory/theory theory debate; and debates concerning whether innate knowledge such as linguistic knowledge or folk theory knowledge is tacitly or explicitly represented. Let me say a bit about each of these. Partisans of the LOT hypothesis argue that the vehicle of thought is an internal code with all of the expressive power of a natural language, over which formal operations constituting the primitive operations of cognition are defined. Most prominent among these partisans, of course, are Fodor (op cit) and Devitt (op cit), though their versions of the hypothesis differ from each other in ways with which we need not be here concerned. Others, such as Carruthers (op cit) and Davies (op cit), argue that we think in natural language – that is, that our thoughts are in fact represented in internal tokens of strings of the languages we speak. Now, though as I argue below, and as I have recently argued in much greater detail (1997a) there is something to this latter suggestion, here I simply point out that we are led into this debate only by the extension we have been criticising of the metaphor of language as a way of imagining thought, and the illicit importation of a syntactic image as the consequence of that extension. Insofar as there is no reason to believe that the fundamental vehicle of thought is codelike at all (again, see 1997a) this debate about the nature of that code becomes spurious. Clark (1992) notes that this
Thought as Language: A Metaphor Too Far
97
fixation on the syntactic character of the language in terms of which we imagine thought also leads us to ignore the importance of dynamic cognitive structures as the substrata of representation: [S]tandard ideas about explicit representation... rely too heavily on the superficial structural notion of explicitness – a notion which is illegitimately building in the idea of visibility to inspection by a human agent as the mark of immediate useability. Thus we find ourselves reflecting on the idea of a word on a page as a paradigm of explicitness. But that case, if we are not aware of the processing dimension, will mislead. [382]
The debate between proponents of the mental simulation theory and the theory-theory of our competence in folk psychological attribution and explanation has been a thriving industry of late. It appears to be a contest between two theories of the structure of our knowledge about minds. Theory-theorists argue that our minds contain data structures that encode a theory of folk psychology which we consult in attributing mental states of our fellows, or in predicting or explaining their behaviour. Simulation theorists, on the other hand, argue that our minds contain devices for processing a representation of ourselves in a situation similar to that in which behaviour must be predicted or explained, or attitudes attributed, whose output is the relevant prediction, explanation, or attribution. I have argued elsewhere (1997b) that the distinction between these two positions is in danger of crumbling in virtue of the fact that neither can appeal to an account of mental representation sufficiently determinate to distinguish between explicit representation of a theory with attendant implicit representation of a simulation mechanism and explicit representation of a simulation mechanism with attendant implicit representation of a theory. But now we can point to why this is so. The supposition that there is a real difference between the content of these two formulations of theory, as opposed to mere notational variation, is grounded in the supposition that there is a mental code in which one of the other is explicitly written whose syntax and morphology, when deciphered, will reveal one rather than the other. But when we abandon the code fixation, the apparent difference melts. Hence we have an explanation of the gradual fizzling of this debate over the last year. There appears to be all the difference in the world between tacit and explicit knowledge representation. And this appearance seems to be confirmed by the plethora of ways proffered of drawing this distinction, and by the auxiliary degrees of explicitness that get tossed in from time to time, such as potential explicitness, implicitness, etc. Once we get this set of distinctions mobilised, we can get off and running with a set of debates regarding whether
98
Jay L. Garfield
this or that kind of knowledge is represented tacitly, explicitly, implicitly, potentially explicitly or perhaps potentially tacitly explicitly. If I am right, however (as I argue in 1997b), most or all of these debates are empty of content. For the metaphor that motivates them all is the metaphor of the printed page, in particular the letter of recommendation, reading explicitly, “her handwriting is neat and she discharges her duties to her own complete satisfaction,” while reading implicitly – between the lines as it were – well, you know. The problem, though, for the use of this metaphor in thinking about thought, is that there are no lines in thought between or on which anything can be written. What, after all, is the difference between an organism’s containing an actual occurrent representation whose actuality simply is the disposition of a system to produce a particular discrete token or behaviour, its containing a potentially explicit representation which is that token or behaviour, potentially explicit because the token or behaviour in question is not now occurrent, and its having a bit of tacit knowledge exemplified in its ability to produce that token or occurrence in the right circumstances? A thousand scholastics gathered on the head of a pin could not sort this ontological muddle out. Simply because there is none. To the extent that we can talk about explicit thought we are talking about occurrent person-level attitudes available for verbalisation or for the occurrent guiding of action. But by that standard, all the rest is equally implicit, tacit, or otherwise. And by that standard, what is explicit is what is least interesting to cognitive science. Only if we are seduced by the inner code view do the debates in question get going. This last point about what it takes for thought to be properly explicit leads us to a consideration of what we can fruitfully say about the relation between language and thought.
8. What Language Brings to Thought I have been in this discussion resolutely deflationary with respect to the pretensions of language as a model of thought. But this line must be tempered by a recognition of the powerful and important role that language does play in the thought of the linguistically competent. For it is in the case of thinkers such as us that it makes the most complete sense to say that thought is linguistic. This is no accident. Human natural language is a tool of enormous expressive power and utility in guiding and co-ordinating behaviour. Language makes possible the representation of far more content than can be represented without it. As such, the development of linguistic competence and the entry into collective cognitive and epistemic life it makes possible open an entirely
Thought as Language: A Metaphor Too Far
99
new range of possibilities to the mind, a range of possibilities of which it is more than willing to take advantage. The abstract, the remote, the merely possible, and even the impossible come within our ken when we acquire language. Indeed it can truly be said that only with language acquisition does fully human cognition emerge. But there is more to the transformation than this: When we acquire language we acquire a range of concepts that can structure thought and a range of thoughts that can be directed upon linguistic objects, and hence, through these linguistic objects upon their contents. And this, of course, is why language works so well as a model of thought: Our introspectible thought – the thought we can self-ascribe at the personal level and the thought we ascribe to our fellows – is thoroughly linguistic in character and is introspected, reported and known through the conceptual resources derived from language. But there is no reason at all to suppose that its being linguistic consists in its being constituted by sets of cerebral linguistic tokens either in natural language or mentalese. After all, when we attribute attitudes to ourselves or others we neither observe nor hypotheses about any such tokens. How could we? And did we not talk thought talk before anyone dreamed of such inner tokens, let alone observed them? So the principle of classification of cognitive states cannot respect such syntactic boundaries. Or if it does, we must conclude that we at best hazard the wildest guesses concerning our own or others’ intentional states. No. We classify, introspectively attribute to ourselves and – whether confidently or diffidently – attribute states to others by reference to how those states would find expression were we to give them voice. Cognitive states such as ours can be expressed, and it is via their expression or potentially expression that they are known and described. But it does not follow at all that the underlying physical or computational states that would be so expressed, and that in part cause our utterances share in any way the form of their public expression. To say that they do even when we explicitly think “in” or “of” words is to confuse the form of the object of thought with the form of the vehicle of thought. Even here, where the language metaphor is most apt, to push it too far is both tempting and potentially dangerous. Even more so when we hypothesise about more fundamental cognitive processes, or the cognitive processes of our infralingual friends. That is not to say that speculation regarding the psychology of infralingual children and animals is idle or impossible, any more than it is to say that speculation regarding the computational or causal structure of subpersonal thought it idle. It is only to say that when we use attitude talk to characterise such thought, it must always be taken as a façon de parler: Animal representation systems as Sellars and I (1988) have called them are always in an important
100
Jay L. Garfield
sense sub-representational, and must be understood in terms of iconicity or causal linkage, as, for instance, Millikan would have it in her less semantic-reductive moments. Absent language, the language metaphor has little to recommend it. While it is true that thought, like language, is systematically connected to perception and behaviour, and has the function of co-ordinating an organism with its environment by encoding relevant information, the remainder of the potential content of the metaphor is at best unmotivated and at worst irrelevant or misleading. Once language is in the picture, thought is indeed – not surprisingly – strikingly linguistic in character. But for all that, its linguistic character is apparent in the sortals by means of which we know it and by its object, not necessarily in its underlying formal or biological mechanisms. Once again, theory launched in response to the metaphor is likely to land far beyond its evidential support.
9. Thinking Clearly; Speaking Carefully I have not been urging the abandonment of language as a model of all thought, as an epistemic entré to thought, as a vehicle for characterizing the content of thought, as providing a correct analysis of the content of much thought, or even as a useful metaphor for thinking about thought. Rather I have been urging caution in the extension of that metaphor beyond its ability to illuminate and into territory where it is likely to mislead, and where, like Kant’s dove, it brings us into transcendental dialectic. Nor have I been urging any abandonment of the propositional attitudes or a retreat from realism about thought processes defined over propositional attitudes. I urge us to be realistic with respect to our deliberations, musings and beliefs as well as about our assertions, questions and orders. And finally, I urge a serious realism with regard to subpersonal cognitive states. We can fruitfully treat these states as indicative, as causally relevant to perception, cognition and action, as implicated as necessary conditions of propositional attitudes. But such robust realism does not entail that such subpersonal states ought to be treated as fully representational, as having content in the sense either that public linguistic tokens or that the personal states of language users do. The extension of the language of content to thought at these levels can at times be a useful metaphor, as when we use “means” to mean “means,” sliding from indicative sign to semiotic device via homonymy. But when that slide turns from trope to theoretical commitment the metaphor has gone too far, the “decisive move in the conjuring trick” has been made, and we lose our ability
Thought as Language: A Metaphor Too Far
101
to see that “the eddies and flows within the ceaseless flux ... may be no more than they seem.”
References Bradley, James, (1999). Wrack. New York: Henry Holt. Carruthers, P. (1996). Language, Thought and Consciousness. Cambridge: Cambridge University Press. Churchland, P. (1989). A Neurocomputational Perspective. Cambridge: MIT Press. Clark, Andy. (1993) Associative Engines. Cambridge: MIT Press. Davies, M. (1996). “Cognitive Architecture and Psychological Theory.” University of Tasmania Philosophy Seminar. Dennett, D. (1987). The Intentional Stance. Cambridge: MIT Press. Devitt, M. (1996). Coming to Our Senses: A Realistic Program for Semantic Localsim. New York: Cambridge University Press. Dretske, F. (1995). Naturalizing the Mind. Cambridge: MIT Press. Fodor, J. (1975). The Language of Thought. New York: Crowell. Fodor, J. (1994). The Elm and the Expert: Mentalese and its Semantics. Cambridge: MIT Press. Fodor, J. (1997). “Connectionism and Systematicity.” Cognition 62, 109-119. Garfield, J. (1983). “Propositional Attitudes and the Ontology of the Mental,” Cognition and Brain Theory 6, pp 319-331. Garfield, J. (1989a). “The Myth of Jones and the Mirror of Nature: Reflections on Introspection,” Philosophy and Phenomenological Research 50, pp 1-26. Garfield, J. (1989b). “Mentalese and Mental Se: Keeping Language and Thought Distinct,” Wesleyan University Conference on Mind, Meaning and Nature. Garfield, J. (1996). Casting Out Demons and Exorcising Zombies: Exposing Neocartesian Myths in Frank Jackson’s Philosophy of Mind,” in Dowe, Nichols and Shotton, eds., Australian Philosophers, Hobart: Pyrrho Press. Garfield, J. (1997a). “Mentalese Not Spoken Here: Computation, Cognition and Causation,“ Philosophical Psychology 10, pp. 413-435. Garfield, J. (1997b). “Explicit/Tacit: Is this Distinction any better than Analytic/Synthetic, Scheme/Content, Procedural/Declarative and their Antique Ilk?” Adelaide Concerence on Cognitive Simulation Theory, 1997. Garson, J. (1997). “Syntax in a Dynamic Brian,” Synthese 110, 291-305. Harcastle, V. (1995). Locating Consciousness. Philadelphia: John Benjamins. Millikan, R. (1993). White Queen Psychology and Other Essays for Alice. Cambridge: MIT Press. Sellars, W. (1963). “Empiricism and the Philosophy of Mind” in Science Perception and Reality. New York: Routledge and Kegan Paul. Smolensky, P. (1990). “Tensor Product Variable Binding and the Representation of Structure in Connectionist Systems,” Artificial Intelligence 46, 159-216. Stich, S. (1983). From Folk Psychology to Cognitive Science: The Case Against Belief. Cambridge: Cambridge University Press. Van Gelder, T. (1995). Mind as motion : Explorations in the Dynamics of Cognition. Cambridge: MIT Press.
102
Robert M. Gordon
ROBERT M. GORDON Sellars’s Ryleans Revisited Abstract It is often said that the simulation vs. theory debate must be resolved empirically. But part of the needed empirical work, perhaps the decisive part, is “armchair” testing against philosophical touchstones such as intentionality, opacity, and Moore’s paradox. I will assay Sellars' myth of “our Rylean ancestors,” frequently cited as the prototypical “theory” theory. Sellars’ laudable aim was to show how “privileged access” could be preserved without making first person ascriptions of mental predicates incorrigible or dependent on “immediate experience.” He attempted this by portraying these ascriptions as essentially theoretical: We are so conditioned that, when situational and behavioral evidence indicates certain theoretical states and episodes, we make the corresponding self-ascriptions, without having to consider the evidence. Sellars’ approach fails, however, because self-reports are actually coordinated only with verbal behavior – typically, the outward-looking “expression” of a state in the non-mental ‘Rylean’ language. (Such coordination is illustrated, in the case of belief ascription, by Moore’s paradox, and more generally, by what I call “ascent routines.”) Thus our self-reports could not be a product of theory-based training – nor directly theory-based, as some psychologists suggest. Given this coordination, the only way to achieve Sellars’ aim – without sacrificing nivocality – is to suppose ascriptions to others to be “third person self-ascriptions” based on the other’s situation and behavior, as the simulation theory maintains. Sensitive to nonverbal as well as verbal behavior, they prevail over first person ascriptions. This leads to a certain conception of people: peepholes through which the world reappears, possibly transformed.
Wilfrid Sellars’s essay, “Empiricism and the Philosophy of Mind,”1 introduced, although it did not exactly endorse, what many philosophers consider the first defense of functionalism in the philosophy of mind and the original 1
In The Foundations of Science and the Concepts of Psychoanalysis, Minnesota Studies in the Philosophy of Science, Vol. I, ed. by H. Feigl and M. Scriven (Minneapolis: University of Minnesota Press, 1956), pp. 253-329. Reprinted in Sellars, W., Empiricism and the Philosophy of Mind, with an introduction by Richard Rorty, and a study guide by Robert Brandom (Cambridge, Mass: Harvard University Press, 1997), 13-118. Page references are to the original publication.
Sellars’s Ryleans Revisited
103
“theory” theory of commonsense psychology.1 I will be posing a problem: If Sellars’s account were true, then we would not talk the way we do about our own mental states. This problem confronts any otherwise plausible version of the theory theory.
Part I. Sellars’s Myth The overall purpose of Sellars’s essay was to move philosophers from Ryle’s alternative to Cartesianism to a more plausible alternative. More precisely: from the prevailing behaviorist understanding of mental language, which denied that mental events were privately accessible episodes that caused behavior, to a broader empiricism that captured what was valid in the classical Cartesian view: the intuitions that mental language does indeed refer to events or episodes that cause behavior, that these events are in some sense “inner,” and that we have privileged access to them. What could not be salvaged from Cartesianism were the ideas that this access is by inner perception and that it yields infallible knowledge. Those ideas, which belonged to the Myth of the Given, were to be overthrown. To achieve this aim, Sellars proposes to replace the Rylean dispositional reconstruction of mental language with a theory-like reconstruction. Mental language does indeed refer to unobserved inner episodes - but unobserved and inner only in the way that the submolecular particles and events of physical theory are unobserved and inner - that is, theoretical, and therefore not definable in observational terms. He proposes this, not as an accurate analysis of actual mental language, but merely as yielding new insight into certain philosophical problems. To exhibit the strengths of the theory account, Sellars invented a developmental myth: one that portrays our distant ancestors as initially talking about people in the way Ryle represented us as talking, then learning to talk about inner episodes and states as theoretical causes of behavior. Sellars’s early humans already have a language, but one that enables them to speak only of “public properties of public objects” (309). He supposes this public language to include whatever logical resources, such as the subjunctive conditional, would be needed for dispositional concepts. What the language would not equip its native speakers to do was to talk about an inner mental life. Sellars assumes that, as a consequence of this deficiency, the speakers would have had a greatly impoverished capacity to explain human behavior. 1
It was Brad Cohen who persuaded me that Sellars did not subscribe to these views himself.
104
Robert M. Gordon
His student, Paul Churchland, describes the limitations such a language would impose: Their explanatory resources for explaining human behaviour are limited to a few purely dispositional terms, all of which can be operationally defined (like ‘is soluble’)... For this reason, Sellars refers to these people, pejoratively, as ‘our Rylean ancestors’ ... They can explain some human behaviours, but only very few.1
Because Sellars thought these people so compromised, he supposed them to be in need of someone to teach them how human speech and action could be causally explained. That innovator, whom he dubs Jones, speculates that overt utterances and “intelligent” nonverbal behavior are caused by certain inner episodes – episodes that resembled utterances in having semantic properties such as reference. These episodes, which he calls “thoughts,” were supposed to be actual occurrences that cause behavior, not mere dispositions to behavior. They were hidden in the way theoretical properties are, yet grounded in public experience, like theoretical entities in the physical sciences, and eventually open to confirmation or disconfirmation. I will return later to the more interesting elaborations of Jones’s theory.
Part II. My own visit Now I offer a brief myth of my own. I too paid a visit to a prehistoric human cohort with a vocabulary limited to public properties of public objects, and with no conception of an inner mental life. Despite this deficiency, I was surprised to find that their capacity to explain human behavior did not seem greatly diminished. I accompanied them in their daily activities and made something of a pest of myself, constantly testing their ability to explain behavior. Here is an excerpt from my journal: We are tracking across the savanna, lots of excited talk. After ten minutes, the leader makes a sharp turn, and we all follow. I ask one of them, Why did she turn this way? My Rylean does not seem at all puzzled by the question. He replies: This is the way to the mounds. Testing further, I ask, But why is she leading you to the mounds? The answer: They are the termite mounds! I persist: What is special about termite mounds? A few others overhear me. They turn around in surprise. That is where we 1
In S. Guttenplan, ed., A Companion to the Philosophy of Mind, Oxford: Blackwell, 1994), p. 308.
Sellars’s Ryleans Revisited
105
can get termites! For you, our honored guest! Rudely I ask: What is so special about termites? They look at one another and break out in general laughter. The leader walks over to me with a solicitous smile and says, It is the way they taste, of course. Nothing in the whole wide world - I include the forest and yonder mountain as well as the savanna nothing has a more pleasant taste! That is why we are laughing. Ryleans they were not. Although they never referred to mental states or episodes, they gave appropriate explanations of action - causal explanations, it would appear, in terms of the reasons for which the actions were performed though strictly in terms of public properties of public objects, or at least what they took to be public properties. This included how pleasant or good termites taste; how scary, disgusting, soothing, attractive or repulsive certain items in the environment, or indeed certain possible actions, were. Yet they knew nothing of mind – not even that the pleasantness of the taste of termites is in the mind of the taster. I suppose you could say they lacked the sophistication to purge nature of qualities we tend to regard as mere projections of the mind. For them the mental is spread out over the world, coloring objects and situations with emotional and motivational charges, and not yet bottled into minds. Because they were always looking outward to the world, never inward to the mind of the agent, I will call them the Outlookers. Unlike Sellars’s Rylean ancestors, these people seemed in no need of a Jones to teach them how to explain behavior. Indeed, their explanations resembled those we give most of the time. True, we might say “She turned this way because she thinks it is the way to the mounds.” We do this under special conditions, for example, if we ourselves don’t agree that it is the way. I will not take the time to develop the point, but where the choice is between “because p” and “because she thinks (or: believes) that p,” it is the former, the factive form, that is the standard or default form of explanation. It is the form that is used unless one has some reason not to use it. The relatively noncommittal “because she thinks” form accordingly carries a conversational implicature to the effect that one does have some reason not to make the stronger commitment implicit in the factive form. We can easily account for the difference between Sellars’s Ryleans and my Outlookers. Sellars’s rhetorical purpose was to show how philosophers of a Rylean persuasion can with good reason speak of inner episodes while preserving their self-respect as empiricists. That is why his Rylean ancestors carry a much heavier burden than the restriction to public language. They are restricted to a much more austere idiom, which eschews not only causal explanations of human action in terms of mental states and episodes, but also causal
106
Robert M. Gordon
explanations of human action in terms of reasons like, “it is the way to the mounds,” which make no mention of mental states and events. My Outlookers have a much lighter load to carry. Sellars’s effort to move from behaviorism to a theory that acknowledges the role of the mental as in some respect an “inner” cause of behavior certainly represented a major step in the right direction. It was a serious attempt to make respectable some of Descartes’ better intuitions. Other notable efforts were made in the 1960's to answer Wittgenstein and Ryle with a theory theory that spoke of inner causes of behavior: for example, by Putnam and by Chihara and Fodor. But, focusing, as Wittgenstein and Ryle had, on philosophically problematical language, these critics neglected explanations couched in unproblematical terms, such as, “She turned this way because it is the way to the mounds.” The upshot is that they wrote as if psychological competence the capacity for anticipation, explanation, and social coordination of behavior - begins and ends with the possession and use of mental concepts and the principles that link them. The tendency in the years since the publication of Sellars’s essay has been to narrow the focus even further to the concepts of belief and desire. Many use the term belief-desire psychology to name the source of human psychological competence. The success of the Outlookers is bound to be fragile. We have to imagine them to be a small, homogeneous group without dissension, or we will find them frequently perplexed by one another. They are not replicas of one another. If one of them misses a meal and is ravenous, the others who are not hungry at the time may have no way of understanding his behavior. And if one is privy to information another lacks, the two may find one another’s behavior beyond explanation. That is why we, at least, find it useful to speak also, when the need arises, of the mental causes of action. But it is important to appreciate that mental causes are in general a second best, invoked when there is reason not to locate the explanans out in the world. Not only the direction of the mounds, but also the taste of termites, and the pleasantness of that taste, resides in the world, unless one must for some special reason locate the cause in the agent. The simulation theory then goes further and says that there is no radical departure from this focus on the world even when we do have to introduce mental causes such as thoughts. As Jane Heal says, the capacity to think about the thoughts of others is just an extension of the capacity to think about the objects of these thoughts, their subject matter, “together with some extra sophistication.” That is, “the capacity to think about thoughts must be seen as an extension of the capacity to think about their objects.”
Sellars’s Ryleans Revisited
107
For example: In the famous false belief test of Wimmer and Perner, you see Maxi put his chocolate away in hiding place A. While he is outside, someone transfers the chocolate to hiding place B. Where will Maxi look for his chocolate? If you are simulating, you don’t shift from thinking about where the chocolate is to thinking about that other, more complex object, Maxi. You think about the chocolate, but in a special way. Locating yourself mentally with Maxi while he’s out of doors, you blank out what’s going on inside. Then you come back in with him and look for the chocolate—but without benefit of the crucial information about the move from A to B. According to the simulation theory, then, even when dealing with the problem cases where the explanans is not out there in the world, we go on explaining or predicting in terms of facts, but at the 4-year-old stage we can mentally relocate ourselves and adjust the facts accordingly. For an analogy, think of those little binocular telescopes we sometimes see at tourist sites. People line up to put in their coin and look into the scope. You put your coin in, look into the eyepiece, and if all goes right, the shutter opens and, lo and behold, there’s the other rim of the Grand Canyon, or the fumaroles of Kilauea. So too, for the simulation theory, at least as Heal and I understand it, to find out what goes on “inside” a person (in the relevant sense), we look in, and, if all goes right, we see, in some cases, the very scene before us, though from a different physical perspective, and possibly with changes in the emotional and motivational colorations of things, and possibly serious mutations and transformations of the very things themselves. This is in contrast to another way to understand looking into the telescope and seeing what is there inside it: taking it apart and examine it, or using some sort of non-invasive imaging, or speculating about the internal systems that allow it to do what it does - the internal structures and mechanisms that account for the optical properties of the telescope and its response to an inserted coin. What the telescope analogy is meant to suggest among other things is that when you look inside people psychologically, what you see there is nothing but the world – the world with its emotional and motivational qualities, its affordances, and its various modal properties. The point, I think, is similar to Hume’s, about entering most intimately into what he calls himself, always stumbling on some particular perception or other, and never observing any thing but the perception. This brings me to the topic of how one looks into oneself and says what’s in there. The answer I want to defend will be, of course, that one does this by looking out to the world and talking about what’s out there.
108
Robert M. Gordon
Part III. Self-ascription Theory theories have had particular trouble with accounting for the apparent differences between third person and first person ascription: the third person dependence on situational and behavioral evidence and the first person independence of evidence, together with the epistemologically privileged character of first person ascription. The extreme position sometimes taken is that there are no real differences. One simply applies the theory to oneself on the basis of situational and behavioral evidence. It is just that the range and amount of evidence available in one’s own case is typically far greater than what is available to others. Further, it has been argued that our expertise concerning our own psychological states, based on past applications of folk theory to ourselves, may lead to the illusion that our knowledge of our own states is “perceptually immediate, noninferential, direct.”1 Other theory theorists, perhaps most notably Sellars, have found this implausible and have searched for an alternative that grants at least some degree of first person authority. The main point of his myth, Sellars tells us, is that we need not give up the privacy of inner episodes such as thoughts in order to acknowledge the essential intersubjectivity of language and languagelearning. For the reporting role of these concepts – the fact that each of us has a privileged access to his thoughts – ... is built on and presupposes this intersubjective status. Sellars’s novel idea was that privileged access to the truth about one’s own thoughts does not require an experience or awareness of one’s thoughts. What alternative is possible? Sellars describes the process in terms of operant conditioning. Once the Ryleans have mastered Jones’s theory and can apply it to others, they are so conditioned that, when situational and behavioral evidence indicate certain theoretical states and episodes, they are disposed to make the corresponding self-ascriptions, using the language of the theory, without having to consider the evidence. In Sellars’s story, Jones brings this about, roughly, by applauding utterances by Dick of “I am thinking that p” when the behavioral evidence strongly supports the theoretical statement “Dick is thinking that p”; and by frowning on utterances of “I am thinking that p”, when the evidence does not support this theoretical statement. Our ancestors begin to speak of the 1
Alison Gopnik, “How we know our own minds: The illusion of first-person knowledge of intentionality,” Behavioral and Brain Sciences 16:1 (March 1993), Pages 1-14.
Sellars’s Ryleans Revisited
109
privileged access each of us has to his own thoughts. What began as a language with a purely theoretical use has gained a reporting role. Once trained, Dick should not have to make theoretical inferences from his own behavior. If the theory is right and if the trainer applies it correctly, the inner state that caused the behavior that was reinforced will indeed be one that satisfies the functional role assigned to “thinking that p” in Jones’ theory. And in the future that state will cause Dick’s self-ascription, “I am thinking that p” - provided, of course, that Dick wants to report what he is thinking. A great virtue of this account is that it shows how mental predicates may have exactly the same sense whether ascribed by another on the basis of evidence or self-ascribed by the subject independently of evidence. The predicate is univocal simply because the evidence-based third person sense is the only sense it has. The first person use makes no independent contribution to meaning. I am much in sympathy with Sellars’s idea that there is privileged access to the truth about one’s own thoughts and that it does not require an experience or awareness of one’s thoughts. It is certainly more plausible phenomenologically than the view that, like everyone else, we have to examine our behavior, even if we can do it in a “perceptually immediate” way. I also find it better than the views of those simulationists who bank on an inner sense dedicated to the perception of mental processes. Yet Sellars’s account will not do. If something like it were true, then we would not talk the way we do about our own mental states. The problem is easily illustrated for the ascription of beliefs and thoughts. Consider the following sentences: It is raining, but I don’t believe it is. I believe it is raining, but it is not. As G.E. Moore pointed out, to assert either of these conjunctions would be self-stultifying, as the asserting it conversationally defeats what is being asserted. Yet neither sentence is formally inconsistent. In fact, there have been occasions on which I wasn’t aware that it was raining and also occasions on which I mistakenly believed it was raining. This is Moore’s paradox: Despite the formal consistency of what is uttered, the utterances have what has come to be called Moorean or pragmatic inconsistency. The same problem occurs with reports that one is currently thinking that p, as distinguished from reports of a standing belief that p; thus, with reports of what Ryle and Sellars call episodes rather than states. Consider, “It is raining, but I am not thinking it is,” and, “I am thinking that it is raining, but it is not.”
110
Robert M. Gordon
Unless “thinking” is merely entertaining a thought, you will hear these as pragmatically inconsistent. Now suppose you have received Jones’s theory-based training in the use of “I believe” in self-reports. Now, you use the “I believe that p” form only when the behavioral evidence strongly supports the theoretical statement that you have the belief in question. Consider, “It is raining, but I don’t believe it is.” The first clause is a weather report, the second clause is a self-report. If we accept Sellars’s account, there should be no reason to expect the two always to correspond. The weather report requires only training in the public language. The self-report, however, is guided by the special theory-based training you received for giving such reports. Such training is based on your behavior, verbal and non-verbal. Part of the behavioral evidence, namely your asserting, “It is raining,” may be outweighed by other behavioral evidence that strongly indicates you do not believe that it is raining. So there are liable to be occasions on which you might justifiably and without inconsistency say, for example, “It is raining, but I don’t believe it is.” Likewise, there should be no appearance of inconsistency if you say, “I believe it is raining, but it is not.” For the two clauses concern two quite different topics. As I said earlier, if something like Sellars’s account were true, then we would not talk the way we do about our own mental states. This is not just a problem for Sellars’s account of how a theoretically based self-ascription is possible. The same problem exists for Gopnik’s view that we examine our own behavior in the light of the theory, and accordingly report that we believe or are thinking, say, that it is raining. Rather than canvass other possible variants of a theory theory account, I will shortly offer an argument that I believe covers any variant that any philosopher or psychologist would seriously want to put forward. First I will illustrate what I mean by coordinating ascription with expression. Here is a question I invite you take a moment to answer: Is the following something you believe? “It is raining outside.” I doubt that in answering the question you examined your recent behavior in the light of a theory. And I doubt that you introspectively searched for a telltale feeling or other experiential mark of belief. You probably just reinterpreted the question as, “Is it raining outside?” Maybe, like me, to answer the question you try to recall your last glimpse of the outside world a few minutes or hours ago. If your answer to this meteorological question was affirmative, then you were prepared to say, “Yes, I do believe that it is raining outside.” If your answer was negative, then you were prepared to say, “No, I don’t believe that it is raining outside,” or more strongly, “No, I believe that it is not raining
Sellars’s Ryleans Revisited
111
outside.” If you had no answer to the lower-order question, then you might have said, “No, I don’t believe one way or the other, I just don’t know.” (There appear to be constraints that rule out, among other things, taking excessive time to retrieve or derive the answer, or actually taking out that map again.) In a couple of recent papers1 I argued that the self-ascription of mental states, more precisely, any self-ascription that is not based on evidence, is accomplished by the use of what I call ascent routines. These are procedures routines requiring no reasoning - that allow one to answer a question about oneself, and specifically about one’s mental states, by answering a question that is not about oneself, nor about mental states at all: an outward-looking question. They answer a question about a mental state that is about x by answering another question pitched at a lower semantic level, a question directly about x. That is why I call them ascent routines. The general idea is, roughly, that to ascribe a belief, desire, intention, or other mental state to ourselves in a reasonably reliable way, all we need is the ability to express the belief, desire, intention, and so forth. For example, when I tell the clerk, “I want two of these and three of those,” it is very likely that I do want possession of the items I point to. To attain reliability with regard to what I want, I don’t look inward, I look outward at the display case. In the two papers in which I discuss ascent routines, I described the ascent routines for just a few mental states, perhaps most convincingly for belief. So I am not certain that for every mental state there is an ascent routine, and that the use of this routine underlies all self-ascriptions of that state that are not evidence-based. But there is reason for optimism if you consider the following to be pragmatically inconsistent: That is frightening! But I am not at all frightened by it. That is absolutely delightful! But I am not at all pleased by it. My toe hurts! But I am in no pain. The taste of termites is just fabulous! But I don’t like the way they taste. In each case, the self-ascription must be so guided as to avoid the possibility of any of these utterances - except in a special disclaimer context like the one in which I am mentioning them. The general idea is that we coordinate one type of verbal behavior, selfreports of a mental state or episode, with another, the outward-looking “ex1
“Simulation Without Introspection or Inference From Me to You,” in M. Davies & T. Stone, eds., Mental Simulation: Evaluations and Applications (Oxford: Blackwell, 1995); and, “‘Radical’ Simulationism,” in P. Carruthers and P. Smith, eds., Theories of Theories of Mind (Cambridge University Press, 1996)
112
Robert M. Gordon
pression” of the state or episode. A self-report should give others a glimpse into a subject looking out onto the world. Hence it is tied to speaking about the world, including speaking expressively about the world, finding something scary, or finding a part of the body painful. The only way for a theory theorist to avoid Moore’s paradox is to concede that our self-ascriptions, at least regarding what we believe or are thinking, are coordinated in the way I suggest with verbal expressions of the state – the outward-looking “expression” of a belief or thought in the public non-mental language1- neglecting all nonverbal evidence and any other verbal evidence. Of course, this accommodation to Moore’s paradox does not come cheap. It leaves the theory theorist with a dilemma. If on the one hand she foregoes the theory in the case of first person ascriptions but does not forego the theory in the case of third person ascriptions, she gives up the univocality of first person and third person. The meaning of “x believes it is raining” will differ between first person and third person uses - a problem Sellars’s proposal had avoided by assuming that there is only a third person theoretical meaning. If on the other hand she decides to forego the theory for the third person as well as the first person, then of course she gives up the theory theory. Suppose we accept my conclusion that self-ascription is coordinated just with self-expression, and not, as the various theory theory accounts would have it, also with nonverbal evidence and verbal evidence other than expression. How can we also allow that third person ascription is and should be based on all of these kinds of behavior? Wouldn’t we be giving up univocality? Not at all, if my version of the simulation theory is right. According to my version, it does not matter whether we are ascribing mental states to ourselves or to another: In either case we answer the question about belief by directing our attention to the subject matter of the belief: meteorological conditions, in the case of the belief that it is raining. Whether in the role of another or not, we find out whether the subject has a belief that p simply by surveying the world to determine if it is the case that p, and then we turn the answer to this over to an ascent routine. The ascent routine insures that the coordination is simply with the verbal expression of belief, and not with any other sort of behavior. And yet, if I am making the ascription within the scope of a simulation – say, I see someone about to step out the door carrying an umbrella over his head – I take this nonverbal evidence very much into consideration when I first simulate him. There is a two-step process in ascribing a mental state to another: First simulate, using any available evidence; then self-ascribe, using an ascent routine that connects only with verbal expression. 1
I am leaving aside the special case in which the content of the belief concerns mental states.
Sellars’s Ryleans Revisited
113
I agree with Sellars that even the first person use of mental predicates is essentially an intersubjective achievement. But Sellars locates the intersubjectivity in the wrong place. We don’t need other people to tell us what we think, as in the old behaviorist joke, “You are fine, how am I?” Selfreports borrow from the intersubjectivity of the public language itself, the language of “public properties of public objects.” What is essential to regulating the first person use of believe and think, as in “I believe it is raining,” is intersubjective training in making assertions about the world, such as, “It is raining.” To regulate our first person reports, we do not require any new intersubjective training - none, that is, beyond training in the appropriate ascent routines, which teach us how to piggyback on the public language.1 I said that my argument covers any version of the theory theory likely to be put forward. That is because no one would want to endorse a theory theory that denies the relevance of nonverbal behavior. But if the theory theorist allows that nonverbal behavior is relevant to ascription, and thus that first person ascription is not uniquely tied to the verbal expression of mental states, then she invites utterances like, It is raining, but I don’t believe it is. That is scary! But I am not at all scared by it. The taste of termites is fabulous! But I don’t like the taste. Thus, were any version of the theory theory that admits the relevance of nonverbal behavior true, we would talk about our own mental states in a way that in fact we do not. It is generally agreed that the issues between the simulation theory and the theory theory are fundamentally empirical and therefore to be settled on empirical grounds. The usual assumption, however, is that the relevant empirical investigations lie within the domains of psychology, particularly developmental and cognitive psychology, and neuroscience. I interpret “empirical” more broadly. Part of the needed empirical work, perhaps even the decisive part, is “armchair” testing of a sort that calls more on the skills of philosophers than on those of experimentalists: the testing of competing views against philosophical touchstones such as intentionality, referential opacity, the peculiarities of self-awareness, the apparent asymmetry of first person and third person attribution, and Moore’s paradox. These are features that have been 1
One might think this is true only of propositional attitudes, and not, for example, of selfreports of sensations and feelings. One might think there must be special intersubjective training for, “I am in pain,” or “I have a toothache.” Is not it when the child puts her hand on her cheek and winces and cries that we say, “You have a toothache”? Yes, indeed, but this need only piggyback on pain expression, “That hurts!” or, “Tooth hurts!”
114
Robert M. Gordon
thought to set mental state concepts off from nonmental concepts, and thus to constitute test cases for resolution of the mind-body problem. They are also test cases for resolving the question of what our psychological competence consists in. An adequate account of psychological competence must at least be consistent with the existence of these phenomena; an ideal account will predict the existence of these phenomena. What I have argued is that no plausible version of the theory theory is consistent with, much less predicts, Moore’s paradox and, more generally, the tight linkage between our “inward-looking” self-ascriptions and our “outward-looking” reports about the world.
Self-Ascription and Simulation Theory
115
LOUISE RÖSKA-HARDY Self-Ascription and Simulation Theory1 Abstract This paper examines the two leading simulation approaches to mental selfascription, Alvin Goldman’s introspectionist account and Robert Gordon’s nonintrospectionist, “ascent routine” account, with a view to determining their adequacy as accounts of our ordinary self-ascriptions of mental states. I begin by reviewing the features of everyday mental state ascriptions and argue that an adequate account of mental state attribution must be able to account for the salient features of those mental attributions we make by using the sentences of a language we know (section 1). By way of introducing the simulation accounts, I outline the tenets of the ‘Theory’-Theory of mental state ascription and sketch the simulationists’ objections to it (section 2). The specific proposals of Alvin Goldman (section 3) and Robert Gordon to ascent routine simulation approach (section 4) are then examined in detail. I argue that both Goldman’s and Gordon’s approaches to mental self-ascription have serious shortcomings. However, the difficulties facing their respective positions suggest *investigating a third approach to the self-ascription of mental states.
The ability to characterize what we want, believe, fear, intend, hope and feel is fundamental to the way we anticipate and explain everyday thoughts and actions. It underwrites those self-ascriptions of beliefs, desires, intentions, emotions and other conscious, occurrent ‘mental states’ which we take to express self-knowledge. It also subserves the ascriptions of mental states to others that we make in the course of explaining, justifying and predicting their thoughts and actions. The ability to characterize oneself and others in terms of mental states is thus a central feature of the psychological competence of normal adults. It is also vital to our conception of ourselves as minded agents in the world. For in ascribing the full panoply of mental states, we place ourselves and those to whom we ascribe them within the common sense psychological framework, in which we speak of persons and actions.
1
This paper has benefited from discussions with the participants and visiting lecturers at the National Endowment for the Humanities Summer Seminar on „Folk Psychology vs. Mental Simulation: How Minds Understand Minds“, directed by Robert Gordon at the University of Missouri - St. Louis in 1999. I thank the NEH for its support.
116
Louise Röska-Hardy
Our everyday ability to ascribe mental states raises a number of questions, which have important philosophical ramifications. First of all, how do we arrive at everyday self-ascriptions of mental states? Secondly, is the prior possession of mental state concepts necessary for mental self-ascriptions? And, thirdly, if mental state concepts are necessary, what is the nature of the mental concepts we employ in everyday ascriptions of mental states? These questions are at the center of the debate between the ‘Theory’-Theory (TT) and the Simulation Theory (ST).1 The theory-theorists claim that our mental state ascriptions result from deploying theoretical knowledge, while the simulationists ground mental state ascriptions in our own cognitive and practical capacities, untutored by theory. Until recently the TT-ST debate has mainly focused on the ascription of mental states to others, or third-person ascription. However, it is clear that any approach which claims to identify the basis of mental state ascription must also account for the self-ascription of mental states, or first-person ascription. The TT presents a unified account of first-person and third-person mental ascription; both self-ascription and other-ascription result from applying theoretical knowledge and inference. By contrast, the simulation approach does not entail a specific account of mental self-ascription, although it presupposes that one can be given. In this paper I propose to examine the two leading simulation accounts of mental self-ascription, Alvin Goldman’s introspectionist account and Robert Gordon’s non-introspectionist, “ascent routine” account, with a view to determining their adequacy as accounts of our ordinary self-ascriptions of mental states, i.e. those self-attributions of mental states we make by means of language. For, typically, we attribute mental states to ourselves and to others by employing the resources of a natural language we know, i.e. by using a sentence to ascribe a mental state.2 Such ascriptions are characteristic of the way we ordinarily attribute mental states and they are essential to the way we justify, explain and predict thought and action in everyday situations. Any approach to mental attribution that claims to explain our ordinary, everyday attributions of mental states must be able to account for the attributions we make using the sentences of a natural language. For it is these ascriptions that play a constitutive role in our everyday, “folk-psychological” explanations and predictions of others’ behavior, on the one hand, and provide the candidates for our claims to self-knowledge, on 1
The collections edited by Martin Davies and Tony Stone, Folk Psychology (1995a) and Mental Simulation (1995b), and the collection edited by Peter Carruthers and P. K. Smith, Theories of Theories of Mind (1996) offer a survey of the debate.
2
The claim is that ascriptions of mental states are usually linguistically formulated, not that they are explicitly verbalized, i.e. an ascription need not issue in a linguistic utterance.
Self-Ascription and Simulation Theory
117
the other. Therefore, if an account of mental attribution is to prove adequate, it must be able to accommodate the ascriptions of mental states we make by means of language. With this in mind, I begin the discussion of self-ascription and simulation theory by enumerating the salient features of our everyday self-ascriptions of mental states in order to determine the desiderata of an adequate account (section 1). By way of setting the stage for the simulation approach(es) to mental self-ascription, I briefly review the ‘Theory’-Theory account of mental state ascription and sketch the simulationists’ objections to it (section 2). The basic idea of ST is then introduced against the backdrop of the TT and the proposals of Alvin Goldman (section 3) and of Robert Gordon (section 4) are examined in detail. I shall argue that both Goldman’s and Gordon’s approaches to the self-ascription of mental states have serious shortcomings. However, their failings point to a third approach.
1. Self-Ascriptions In the course of explaining and anticipating thought and action, we characterize both ourselves and others in mental terms, e.g. as ‘believing that she’ll take the position’, ‘wanting to have escargots’, ‘hoping it doesn’t rain’, or ‘suffering from a headache’. We say things like “I believe she’ll take the position”, “He hopes it doesn’t rain”, “John wants to have escargots” or “I’m suffering from a headache”. In doing so, we use the sentences of a language we know to ascribe mental states to ourselves and to others. We usually employ a sentence of the form ‚ ‘S ψ[p]’, where S denotes the person to whom the state is ascribed, \ is a psychological matrix verb, representing the mental state type, and p characterizes the content of the mental state. The referent of the subject term S determines whether a mental state ascription is a self-ascription or an other-ascription, since the psychological verbs, which characterize mental state types, and the various specifications of content can equally be employed to ascribe a mental state to oneself and to someone else. Paradigmatic cases of mental self-ascription exhibit the first-person singular pronoun ‘I’ as the subject term. For although an ascriber can use her proper name, a demonstrative expression (“this speaker”) or a description in making a self-ascription, the use of such expressions presupposes that the ascriber “knows who” she is, e.g. that she is identical with the bearer of the proper name, that she is the referent of the demonstrative expression or that she is the object picked out by the description. In order for others to understand an ascription without ‘I’ as a self-ascription, they must be aware of the
118
Louise Röska-Hardy
identity of the self-ascriber, e.g. they must possess an identifying description or rely on contextual cues. However, when ‘I’ is used to make a self-ascription, neither the ascriber nor the audience need possess identifying knowledge regarding the ascriber. Competence in the English language is all that is required. This suggests that first-person singular ascriptions of mental states (in the present tense), i.e. self-ascriptions which exhibit ‘I’ as the subject term, constitute the basic cases of mental self-ascription. If this is correct, it implies that the notion of a first-person point of view, expressed by the explicit use of the indexical expression ‘I’, is an essential feature of self-ascriptions of mental states. The first-person singular pronoun ‘I’ is part of the internally structured system of English personal pronouns, a system which is articulated along the dimensions of number, gender, case and animacy. Importantly, mastering the system of personal pronouns requires mastering not only the individual pronouns, but also the interrelations among them, e.g. the convertibility of ‘I’ > ‘you’, ‘you’ > ‘I’, ‘I’ > ‘she’, etc., as well as their relations to other categories of referring terms such as proper names and descriptions. Like all personal pronouns, ‘I’ is an index in the sense of Peirce; it directly designates whomever produces a token of the pronoun type.1 However, ‘I’ differs from the other personal pronouns in that its reference on an occasion of use is determined solely by the rule which governs its meaning, while the reference of other pronouns is determined not only by their meaning, but also by the speaker’s intentions. Furthermore, ‘I’, in contrast to other pronouns, is not semantically equivalent to any other referring expression. In essence, for competent speakers of English a use of ‘I’ indexes its producer as the agent and author of the utterance.2 In making a mental state ascription, an ascriber characterizes a mental state token as being of a particular type, e.g. ‘believing that she’ll take the position’ or ‘wanting to have escargots’, in contradistinction to other mental state types. In the case of sensations and perceptions, the modality must be specified. In the case of beliefs, desires and other “propositional attitudes”, one must characterize both the mental state type \ and the content p. Plausibly, the phrases ‘hoping that it is raining’ and ‘believing that it is raining’ characterize different mental states, because the mental state types are different, while ‘believing that
1
The characteristic feature of an indexical on Peirce’s view is that on an occassion of use it stands in a „relation of contiguity“ to its object, e.g. as a rap on the door to a caller (C. S. Peirce, Collected Papers, Boston: Harvard University Press (1935): 2.274-2.308; 3.359-3.364).
2
The arguments for these claims are developed and defended in Röska-Hardy (2000a) and (2000b).
Self-Ascription and Simulation Theory
119
it is raining’ and ‘believing that the sun is shining’ are different mental states, because their contents differ, even though the mental type is the same. In view of the fact that mental state type and content can vary independently of one another, they may require different accounts. I shall put questions about content specification to one side in this paper and focus on the ascription of mental states, since characterizing the specific content p of mental states is a problem for both the TT and the ST.1 In ascribing mental states to ourselves and to others, we the use the same verbs and verb complements, the same words. Prima facie, the (complex) predicates employed to characterize mental states are not ambiguous between first-person uses and uses in other grammatical persons – so-called secondand third-person uses. Rather, the meanings of mental state predicates appear to be univocal, despite the fact that the grounds for first-person and thirdperson ascriptions of mental states vary; arguably, there is an asymmetry between the first-person and the third-person concerning the criteria for ascription, e.g. the observation of behavior and speech and the role of inference (Strawson, 1959). If mental predicates are univocal, then each predicate expresses a single concept, whether employed in the first-person or in other grammatical persons. The contention that mental state predicates are univocal and mental state concepts unitary is supported by the observation that firstperson and third-person mental ascriptions are symmetric with respect to truth or falsity, despite their epistemic asymmetry (Tugendhat, 1979/1989: IV). A first-person ascription of a mental state \[p] and a third-person ascription of that mental state \[p] to the self-ascriber, e.g. ‘I \[p]’ and ‘She \[p]’ predicated of me, will be true or false under the same conditions. When we ascribe sensations and propositional attitudes in everyday contexts, we combine psychological verbs with subject terms in different grammatical persons to produce sentences that are ostensibly of subject-predicate form. Logically, the sentences we use to ascribe mental states appear to be cases of predication or judgement. Third-person ascriptions of mental states are routinely construed as descriptive statements with truth-evaluable content, unless the existence of the mental is denied altogether. There are good reasons for construing first-person ascriptions of mental states as cognitively significant, as well, i.e. as capable of being true or false, notwithstanding Wittgenstein’s expressivist remarks on the use of ‘I’ as subject. In the first place, there are important parallels between first-person and third-person mental ascriptions. These extend beyond the grammatical similar1
The problem of content is discussed with respect to TT in Heal: 1996 and more generally in Röska-Hardy (forthcoming).
120
Louise Röska-Hardy
ities in surface structure to their logical behavior, e.g. regarding the substitution of terms, the application of existential generalization and entailment relations.1 These parallels strongly suggest that first-person ascriptions of mental states are cases of predication just like third-person mental ascriptions. Moreover, we need compelling reasons to deny them this status. For it would be odd, as Richard Moran points out, if others could make cognitively substantial third-person claims about my mental states, while I myself am barred from making any (Moran, 1997: 145). In addition, if we take self-knowledge to be a cognitive achievement, the first-person ascriptions of mental states like beliefs or desires which are candidates for self-knowledge must be construed as cognitively significant. In the second place, there is a difference between being in a mental state and being able to characterize which mental state one is in. Simply being in a mental state does not suffice for ascribing the state in question to oneself. Small children presumably have various mental states, but they must first learn to characterize and distinguish mental state types before they can self-ascribe them. Note, however, that a gap persists even after one has acquired conceptual facility, for competent adults’ self-ascriptions of mental states, especially intentional states, seem open to error, even if this is the exception, rather than the rule. These points imply that our mental self-ascriptions like our otherascriptions of mental states are cases of predication, classification or judgement, which are cognitively significant. Making a judgement necessarily involves the application of concepts. So, if mental self-ascriptions are predications or judgements, then mental concepts are required in order to arrive at self-ascriptions of mental states. One must be conceptually equipped in order to self-ascribe mental states. These considerations imply that an adequate account of mental self-ascription will provide an account of the necessary mental state concepts. Finally, merely producing a sentence token of the form ‘I y[p]’ is not sufficient for genuine mental self-ascription. We could train a young child or even a parrot to preface the sentences he produces with the formula ‘I believe: ___’, e.g. “I believe: crackers taste good”. However, prefacing sentence tokens with the words ‘I believe’ requires neither a grasp of the use of ‘I’ as the firstperson singular pronoun, nor a grasp of a predicate as mental, e.g. in contrast to non-mental. That is to say, it does not imply any understanding of mental or self-ascription at all. As P. F. Strawson has argued, the mastery of firstperson mental ascription requires the capacity to make third-person mental ascriptions (and vice versa) (Strawson, 1959: III: 99-100). In Strawson’s view, 1
Cf. Röska-Hardy (2000b): 56-57.
Self-Ascription and Simulation Theory
121
it is a necessary condition of mental self-ascription that one be able to otherascribe mental states as well; the self-ascription and the other-ascription of mental predicates are interdependent. This view is seconded in Gareth Evans’ claim that mastery of the content of a self-ascription like ‘I believe that p’ requires not only possession of the mental concept expressed by the predicate ‘believes that p’, but also that one conceive of the concept as capable of being instantiated by someone other than oneself (1982: 226). Thus, plausibly, genuine, comprehending mental self-ascription requires both a mastery of the systematic relations between the points of view articulated in the system of personal pronouns and a grasp of the systematicity of mental predicates, e.g. as expressed in Evans’ Generality Constraint applied to mental concepts (1982: 104, chap. 6). This suggests that an understanding of mental concepts as mental is necessary in order to make genuine self-ascriptions of mental states. As a consequence, an adequate account of mental self-ascription should support an account of the mastery of the mental concepts which figure in self-ascriptions of mental states. The upshot of these considerations is that in a mental self-ascription a mental state is predicated of the individual indicated by ‘I’. Such self-ascriptions are cognitively significant first-person predications or judgements, i.e. evaluable as true or false, and, consequently, they involve the application of mental concepts. Moreover, genuine, comprehending first-person mental ascriptions require a mastery of the way our self-ascriptions and other-ascriptions of mental states interlock as well as an understanding of the ascribed state concepts as mental. If these observations are correct, an adequate account of mental self-ascription will satisfy three conditions. First, it will explain how we arrive at self-ascriptions of mental states in a way that respects the connections between first-person and third-person mental ascriptions, e.g. the univocality of mental predicates and the unitary nature of mental concepts. Second, it will provide an account of the mental state concepts and mental state predicates which figure in our everyday mental ascriptions. Third, it will sustain an account of the mastery of mental concepts which applies to first-person and third-person mental ascriptions and allows one to distinguish between genuine, comprehending self-ascriptions and mere “prefacing”. I submit that these three conditions constitute the desiderata for an adequate account of our everyday ability to ascribe mental states to ourselves.
122
Louise Röska-Hardy
2. ‘Theory’-Theory vs. Simulation According to the “Theory”-Theory (TT), we ascribe mental states to ourselves in essentially the same way that we ascribe them to others – by deploying theoretical knowledge. We inferentially apply a folk-psychological theory of mind, which consists in a set of generalizations or laws for the deployment of mental concepts. As ascribers we need not be able to state the theory we use. Indeed, most theory-theorists take knowledge of the theory to be “tacit” or “implicit” or they construe it as sub-personal. On the TT account, ascriptions of mental states result from applying this tacit theory of mind to ourselves and to others. Depending on the particular version of TT, the theory involved is conceived as a set of generalizations or laws for the deployment of mental concepts (Lewis, 1970, 1972; Churchland, 1988), as a special purpose body of knowledge in a mental module (Fodor, 1992; Leslie, 1987), or as a theory analogous to any other scientific theory (Churchland, 1988; Gopnik, 1993; Perner, 1991). However, all versions of TT maintain that mental state ascription involves theory-mediated inference. According to TT, we ascribe mental states by inferring their occurrence from observation of behavior and environmental events, i.e. by recognizing their causal-explanatory role in accordance with the theoretical generalizations comprising the theory.1 Versions of TT differ as to whether the theory in question is acquired through a process of scientific theorizing (Gopnik 1993; Perner 1991), through learning, or whether it is innate and matures through a process of ontogenetic development (Fodor, 1987; Leslie, 1987). Despite these differences, all theory theorists claim that mental state ascriptions result from deploying theoretical knowledge, e.g. sets of principles, psychological laws or “folk” platitudes. The theoretical knowledge account of TT entails that an ascriber must possess the concept of a particular mental state (type) in order to ascribe the mental state to herself or to another person. At some level she must grasp the mental concept as it is defined by the generalizations of the theory in order to make a corresponding mental state ascription. Consequently, concept possession is a sine qua non of mental state ascription on the TT account. Moreover, it has been argued that the representation of mental state types, such as ‘I 1
The TT can be developed in close analogy to theories in the natural sciences, e.g. in terms of nomological generalizations and postulated unobservables; Churchland (1988); Fodor (1987). It can be construed along the lines of analytical functionalism; Lewis, (1970) or representational functionalism; Fodor (1987); Lycan (1981). Or, it can be developed in nativist or modularist fashion in terms of the internally represented knowledge structures or modules postulated by cognitive science, Leslie (1987); Fodor (1983), (1992). For further discussion of these positions, cf. Röska-Hardy (forthcoming).
Self-Ascription and Simulation Theory
123
believe that p’ and ‘I believe that q’, is a defining feature of the processes of mental state ascription postulated by TT.1 This means that an ascriber must have the corresponding facility with the content of mental concepts and the meaning of mental predicates in order to use these in making ascriptions. On the TT account of mental concepts, concept possession involves theoretical generalizations about causal or inferential relations and mental state representations. Mental states such as beliefs and desires are defined functionally in terms of their (causal) relations to events in the environment, to other mental states and to overt behavior; these relations are the subject of the theory’s generalizations. In this view, the concept of a mental state is the concept of a state apt to cause or be caused by certain events or certain types of behavior, i.e. the concept of a particular functional role. Spelling out the nature or content of a mental state type on the TT account will thus involve generalizations that make reference to dispositions, causal interactions or subjunctive considerations. According to TT, these generalizations – specifically, the causal-functional role the terms play in the generalizations of the folk-psychological theory – not only determine the content of mental concepts, but the meanings of the mental state predicates employed in ascriptions, as well (D. Lewis, 1966, 1970, 1972; P. Churchland, 1970, 1981, 1988, 1989). Consequently, an ascriber must master the “theoretical” concepts of the theory in order to employ expressions for mental states. Even though the ascriber’s grasp of the concepts may be “tacit” or “implicit”, it is incumbent on the TT to explain how ascribers can acquire facility with mental state concepts so defined. If it should turn out that ascribers cannot acquire and deploy such concepts, this would count against the adequacy of the TT as an account of everyday mental state ascription, for the TT claims that knowledge of the folkpsychological theory of mind grounds mental state ascriptions. An important consequence of the TT construal of mental concepts and mental terms is that they are defined from what is essentially a third-person or observer point of view. The content of mental concepts and the meanings of mental terms are specified in terms of logical and epistemological relations between external stimuli, mental states, and behavior, i.e. they are defined on the basis of inference and observation of external stimuli and behavior. They are not essentially linked to the first-person point of view. In this respect the mental concepts as defined by TT do not differ in kind from non-mental 1
Martin Davies (1994): 117 argues that the representation of mental state types is a defining characteristic of TT as opposed to ST, for within a simulation of the other the simulated mental state types are not represented. On this construal, the TT processes of mental state ascription operate over representations of mental states like ‚I believe that p‘, while the simulation processes simply operate with ‘that p’ or ‘that q’.
124
Louise Röska-Hardy
concepts in theories about the natural world. The third-person approach of TT to mental ascription further entails that there is no principled asymmetry between ascriptions of mental states to oneself and ascriptions to others, for we ascribe mental states both to ourselves and to others by means of theorymediated inference. In effect, TT assimilates mental concepts to non-mental concepts, which can be characterized without reference to a point of view. Thus, according to the TT, the processes of mental self-ascription parallel that of other-ascription. Self-ascriptions of mental states are based on the same theoretical knowledge as ascriptions to others. Some TT models of self-ascription claim that we use exactly the same processes (Gopnik, 1993: 94). Others construe mental self-ascription as theory-laden perception of “theoretical” entities such as beliefs and desires; self-ascriptions of mental states are thus construed as cases of theory-dependent, non-inferential perceptual knowledge. Still others maintain that certain mental states are such that having a particular mental state directly issues in the belief that one has it; in this view self-ascriptions of mental states are the direct and automatic products of the states which are self-ascribed (Shoemaker, 1993). However, all theory-theorists concur that ascribers deploy mental concepts as defined by the folk-psychological theory of mind in making mental ascriptions. The advocates of Simulation Theory, Alvin Goldman (1989), Robert Gordon (1986) and Jane Heal (1986), deny that ascriptions of mental states rely on theoretical knowledge and inference – “tacit” or otherwise. They reject the idea that we ascribe mental states to others by deploying a theory and offer accounts of mental state ascription, which do not invoke theoretical knowledge, in terms of simulation. The simulationists have challenged the adequacy of the prevailing TT view of mental state ascription on several fronts. For example, they point out that the few candidate folk-psychological generalizations which have been explicitly formulated are beset with problems of vagueness due to attendant ceteris paribus clauses (Goldman 1989, Gordon 1992) as well as with problems of application concerning how and when the TT generalizations are to be applied (Gordon, 1986, 1995; Heal, 1986, 1996a). The simulationists take issue with the theory-theorists’ claim that the ascription of mental states involves the application of theoretical generalizations and theoretical reasoning from environmental and behavioral evidence in purely descriptive terms, i.e. without reference to a point of view. In the case of mental self-ascription, the simulationists argue that the TT account is especially problematic, because it is unclear how ascribers can use the proposed generalizations to generate specific self-ascriptions of mental states. For example, Goldman points out that the TT’s individuation of mental states or definition of mental state expressions in terms of causal-functional or
Self-Ascription and Simulation Theory
125
inferential relations has the consequence that an ascriber must recognize the dispositional or subjunctive properties of a mental state token in order to selfascribe the mental state (1993a: § 3; forthcoming). As he notes, in a particular case one may be ignorant of the factors that such a specification involves, e.g. the causes or effects of a particular mental state token. Moreover, one may not be able to determine the functional or inferential role of a mental state token, as postulated by TT, due to the complexity of the network of relevant factors. Goldman suggests that the functionalist specification of mental states may lead to a combinatorial explosion of determining factors (Goldman, 1993a: §3). Finally, on the TT’s theoretical inference or theory-laden recognition account of mental self-ascription, it is not obvious how one is able to self-ascribe occurrent mental states (Goldman, 1993a, forthcoming). For, as Goldman points out, we often self-ascribe mental states without recourse to environmental or behavioral evidence, e.g. when we wake up with a headache or suddenly think of cooking shrimp for supper. In view of these challenges, there is reason to doubt that the TT account of the self-ascription of mental states solely in terms of third-person knowledge and theoretical reasoning is adequate. Despite the shortcomings of the TT account, displacing the ‘Theory’Theory of mental state ascription is no small undertaking, for the TT has three evident attractions: (1) it provides a unified account of how we make mental state ascriptions both to ourselves and to others; (2) it offers an account of the content of mental concepts and the meanings of mental terms; and (3) it provides an account of the mastery of mental state concepts. In other words, the TT addresses the three desiderata for an adequate account of mental selfascription (section 1). By contrast, the simulation approach has been developed as a competing answer to the specific question of how we ascribe mental states to others. According to the ST, one uses one’s own mental resources to ascribe mental states by pretending or imagining oneself to be in the other’s position and then generating the thoughts or actions ascribed to the other within a simulation of the other. Although all simulationists take the firstperson point of view to be essential to mental states and their ascription, they have different views on the nature of simulation, e.g. on whether it involves analogical inference (Goldman), “ascent routines” (Gordon) or “co-cognition” and rationality assumptions (Heal, 1996b, 1998). They also differ over whether simulation requires the prior possession of mental state concepts (Goldman, 1989; Heal, 1986, 1995:44) or not (Gordon, 1995) The simulationists are also divided over whether simulation is a heuristic, which is usually, although not exclusively, employed in making third-person ascriptions, or whether it is necessary feature of the way we ascribe mental states to others. In conse-
126
Louise Röska-Hardy
quence, the exact status of the simulation approach is also a matter of debate, e.g. whether simulation is an empirical theory to be confirmed by future investigation (Goldman, 1993b), “a priori” in a qualified sense (Heal, 1998) or something in between (cf. Gordon, this volume: part III). In contrast to the TT, the basic ST approach to third-person mental ascription does not dictate a specific approach to first-person mental ascription. It is up to each version of ST to develop an account of self-ascription commensurate with the particular view of simulation advocated. In order to compete with the prevailing TT as a general account of mental state ascription, the simulation approaches must develop (1) an account of mental self-ascription, which interlocks with the preferred simulation account of other-ascription, (2) an account of the content of mental concepts and the meaning of mental predicates, and (3) an account of our mastery of mental concepts which coheres with the preferred simulation account of third-person mental ascription. Thus, whether ST offers a more adequate account of our ordinary mental state ascriptions than TT depends on how well the particular simulation approach satisfies the desiderata. Among the proponents of simulation, Alvin Goldman and Robert Gordon have proposed two distinctive accounts of mental self-ascription in conjunction with their respective simulation approaches. Although each presents a simulationist alternative to the TT, they differ on central issues. Goldman combines an introspectionist approach to first-person mental ascription with his simulation approach to third-person mental ascription, while Gordon utilizes the notion of an ‘ascent routine’ to obviate the need for introspection in mental self-ascription and in simulation. Let us examine their proposals in turn.
3. Goldman’s Introspection-Simulation Approach On Alvin Goldman’s introspection-based ST, one ascribes mental states to others by using one’s own cognitive and inferential mechanisms to match or replicate those of the other person (1989, 1993, forthcoming). For example, to determine what decision another person will make, an ascriber pretends to have the beliefs and desires that she takes the other to have and feeds these into her own practical-reasoning system, which generates further mental states as outputs. The “decision” output of the simulation is then ascribed to the other person. Importantly, in a simulation the ascriber’s cognitive system operates over pretend or surrogate states; it is run “off-line” and consequently does not issue in action. Instead, the output of the simulation process provides the basis
Self-Ascription and Simulation Theory
127
for ascribing a decision or mental state to the other. In contrast to the TT, Goldman’s approach does not require that an ascriber have any knowledge of the system employed or of its properties. The ascriber must only “possess” such a mechanism and be able to “feed in” appropriate inputs and to construe its outputs (1992). Simulation is consequently “process-driven”, not “theorydriven” in Goldman’s opinion (1989). His simulation approach is premissed on the assumption that pretend mental states and non-pretend mental states are homologous as well as on the assumption that ascribers and ascribees are similar with respect to practical and cognitive mental processes (forthcoming). On this construal of ST, simulation terminates in an analogical inference from oneself to the other; the ascriber infers the mental state of the other from her recognition of the pretend state issuing from her simulation of the other (1992, forthcoming, MS.). For Goldman the end product of a simulation is a judgement or a belief, which involves classifying an (occurrent) mental state as a token of a particular mental state type and thus requires mental concepts (1993a, 1993b, forthcoming). As Goldman remarks, “A “type” presumably just is a concept” (MS., 5). His version of simulation assumes a prior understanding of mental state concepts on the part of the ascriber. Indeed, the possession of mental concepts is integral to Goldman’s introspectively-based ST. First of all, an ascriber must possess mental concepts in order to identify the initial inputs for simulation, e.g. she must identify the relevant mental states of the other. Secondly, the ascriber must accordingly have conceptual facility with mental concepts in order to conceptualize or classify her mental state tokens as being of a particular mental state type. In the case of a propositional attitude the ascriber also subsumes a content p under a mental state concept \ such as belief, desire, etc. Thirdly, while running the simulation, the ascriber may need to “tag” the simulated states of the other person as pretend or simulated states or to “isolate” or “quarantine” those of her beliefs and desires, which are not shared by the other (MS.). Finally, mental concepts are required at the end of the simulation process, because one ascribes the output of the simulation routine, e.g. a mental state or a decision, to the other by analogical inference. In ascribing a mental state to another, one moves from a pretend state, “flagged” for the other person, to a judgement about the other’s actual mental state. Thus, Goldman’s ST requires the possession of mental state concepts from start to finish. An ascriber must recognize or detect her own mental states in order to ascribe mental states either to herself or to others. Goldman clearly recognizes the need for an account of mental state concepts as a desideratum of an adequate account of mental self-ascription. For in
128
Louise Röska-Hardy
his view, “the detection or attribution of a mental state, whether in oneself or in another, is a conceptual act: it consists in classifying the target state as of one type or another” (MS.: 5). However, he doubts that the simulation approach has the resources to supply such an account. He favors a first-person approach to mental concepts and proposes an account of mental concepts in terms of introspection or self-monitoring, which he take to be a basic component of our grasp of mental concepts (1989: 183, 1993a, 1993b). In his view we have a first-person, introspective understanding of mental state concepts which rests on direct, non-inferential access to our conscious mental states. The introspectionist, direct-monitoring approach, advocated by Goldman, construes introspection as inner sense or higher-order perception, not as higherorder thought (contra Rosenthal 1992, 1997). The introspectively accessible characteristics of consciousness are consequently only potentially identifiable or discriminable (forthcoming). According to Goldman, introspection takes conscious states as inputs and has as outputs beliefs, which have propositional content. He suggests that the self-classification of mental state tokens is analogous to the self-classification of vision, e.g. one uses current available information about a token to “match” it “in non-cognitive fashion” with a representation or pattern stored in memory (cf. 1993a: 16-17, 1993b). Thus, his account of mental self-ascription posits recognitional criteria, which are directly introspectible. According to Goldman, the fact that ascribers can generally recognize or identify mental states without explicit behavioral or functional information to indicate that the recognitional criteria involved in self-ascription have a monitorable or introspectible character (1993a, 1997). He conjectures that mental state types may be phenomenologically “distinctive” (1993a). By this he means that they may have introspectively accessible qualitative, phenomenological characteristics, i.e. qualia. But independently of whether the introspectible characteristics are qualitative in nature, Goldman hypothesizes that some sort of intrinsic (non-relational) and categorial (non-dispositional) properties of mental states are available to the cognitive system (1993a: 87). He suggests that subjects can introspect or directly detect these intrinsic (qualitative) properties and distinguish them from one another, i.e. that such properties are epistemically identifiable. In his view, these internally detectable properties are definitive of a particular mental state type and thus form the primary basis for our understanding of mental state concepts. Notably, Goldman advances this hypothesis not only with respect to sensations, e.g. pains and other sensings, but also with respect to propositional attitude types like belief and desire. He writes:
Self-Ascription and Simulation Theory
129
“My hypothesis is that mental-state concepts such as desire, belief, and so forth are understood at least in part in terms of non-dispositional characteristics of conscious experience, characteristics that can be introspected by the subject of the experience. The most natural account of people’s special access to their mental states is an account that posits some sorts of nondispositional characteristics that are introspectively available to the subject and distinguishable from one another. My hypothesis is that such characteristics underlie the grasp of mental-state concepts” (forthcoming, sect. 3).
Goldman further conjectures that acquiring mental state concepts involves linking complex conceptual structures, e.g. categorial representations, to “the right sorts of conscious characteristics – latching the belief concept onto belief episodes, the desire concept onto desire episodes, and so forth” (forthcoming, sect. 3; cf. 1993a: §2). Although Goldman maintains that introspection is necessary for possessing mental concepts, he denies that it suffices for their possession. As he states, “Property detection requires appropriate conceptual facility, and this facility is not guaranteed by introspective accessibility” (forthcoming, sect. 3). It follows that introspective access to the intrinsic properties of conscious states must be augmented in order to provide an account of the mental concepts we employ in everyday mental ascriptions and an account of our conceptual mastery. What sort of supplementation is required? Goldman has suggested that something like the functionalists’ characterizations, stripped of all scientific and nomological claims, e.g. references to laws or generalizations, might be pressed into service here (1989, 1993a: 23). Specifically, he suggests that an account of mental concepts might take the form of a dual-representation which incorporates both a qualitative representation and an independent, nonqualitative – possibly functionalist – representation (1993a: 23, 101-2). However, Goldman’s trenchant criticisms of the classical functionalist and the representational functionalist characterizations of mental concepts put this suggestion in doubt (1993a: 23, passim). To summarize, the ability to self-ascribe mental states is crucial on Goldman’s introspection-simulation account. This ability is primarily explained by an ascriber’s capacity to recognize or directly detect the intrinsic, categorial properties of sensory and propositional attitude mental state tokens. The nondispositional, non-relational properties of mental state tokens, directly detected or represented in consciousness, provide the recognitional criteria for mental self-ascription (cf. 1993a: 21). In self-ascribing a mental state, one internally detects a current mental state and classifies it as a token of a particular mental state type. This requires both property recognition and the possession of mental concepts on Goldman’s account. Thus, one arrives at self-
130
Louise Röska-Hardy
ascriptions of mental states on the basis of introspectively accessible evidence. Accordingly, the internally detectable properties of mental state tokens, recognized in introspection, underlie our grasp of mental concepts. Acquiring mental concepts involves “associating” or “latching” mental words or conceptual structures, e.g. categorial representations, onto the introspectively accessible properties of mental state tokens. Goldman’s insistence that fully grasping mental concepts involves linking mental words and conceptual structures to introspectible internal properties raises a number of important questions. First, if introspective detection of intrinsic internal properties is necessary in order to possess mental state concepts and, at the same time, mental concepts are necessary in order to classify mental state tokens as being of one type or another, how do we acquire mental state types or concepts in the first place? Upon pain of circularity, we cannot claim that ab initio we recognize an occurrent mental state or a mental episode as an instance of the mental state type or concept \ on the basis of property . How does one initially identify a mental state token as having a certain nonrelational, categorial property ? What does one initially recognize these properties of mental states as? Since the hypothesized recognitional properties of mental states are only necessary, and not sufficient for mental concepts, a story about how we acquire the full-blown concepts used in everyday mental ascriptions must be supplied. Moreover, it must combine with the introspectionist core of Goldman’s account to elucidate the content of mental state concepts and our mastery of such concepts. The suggestion that we identify intrinsic properties by something like private ostension would be congenial to Goldman’s introspectionist approach. However, this suggestion is problematic for reasons adduced by Ludwig Wittgenstein, who argues that private ostension cannot set a standard of use for a mental term (1953: §§ 243-315). Second, Goldman’s contention that the intrinsic properties of mental state tokens are epistemically identifiable and hence form the evidential basis for mental self-ascription raises a further question. How does one re-identify, i.e. re-cognize, a property of a mental state type \ as ? Identifying a property as would also seem to imply the ability to discriminate it from other properties. How is this to be achieved within an introspectionist approach? This is not a point about justification, although some might wish to press the question of justification here. As a proponent of reliabilism in epistemology, Goldman must only show that such identifications are reliable, not that they are cases of justified true belief (1989: VII). Nonetheless, if introspective access is necessary for the possession of mental state concepts and mental self-ascriptions rest in part on the recognition of internal, intrinsic properties, then Goldman needs to
Self-Ascription and Simulation Theory
131
spell out the conditions for the identification and re-identification of the intrinsic properties of mental states that his account presupposes. Third, the suggestion that propositional attitude states have introspectible intrinsic properties requires further articulation. While this claim is plausible for sensory states like pain, mental states like belief do not usually seem to have phenomenal character. Prima facie, propositional attitude states do not exhibit defining introspectible non-relational features which distinguish them from one another. Which non-relational properties qualify a state as a belief that p or a desire that p in contrast to a hope that p? In contrast to sensations, propositional attitude states like belief seem to be characterized less by qualitative ‘feel’, than by systematic relations to other mental states, perception and action. (However, even sensory states like pain exhibit relations to environmental events and overt behavior, e.g. pains result from physical injuries and give rise to certain behaviors.) The attempt to analyze propositional attitude state concepts primarily in terms of internally detectable non-relational properties, while ignoring any relational properties, seems to exclude an important aspect of our everyday conception of such mental states. In addition, the complexity of propositional attitude states, exemplified by the independence of mental attitude \ and content p, is not readily accommodated on the introspectible properties hypothesis. Surely, Goldman’s claim about introspectibly accessible properties, especially the phenomenological hypothesis, is only meant to apply to the specification of mental state types, not to the specification of content. However, even with this restriction the question of how the introspectionist thesis accommodates different attitudes towards the same content must be answered. Thus, at the very least, Goldman’s hypothesis that propositional attitude state concepts are defined in the first instance by non-relational, categorial properties requires elaboration. Finally, if mental concepts are fundamentally construed in terms of directly introspectible properties, how does one come to associate mental expresssions, e.g. psychological verbs, with a categorial representation on the basis of introspectively accessible properties? The claim that one does it by direct detection of the internal properties of mental states runs up against the Wittgensteinian objections of the “private language argument” (1953: §§243-315). While it is not clear that these objections are decisive, they do require an answer. Since Goldman denies that self-ascription takes place at the level of subpersonal systems that access neural properties (1993a: 20), he is obliged to provide an account of our use of mental words or “verbal labels which are recognizably mental” at the conscious, personal level. It is questionable whether this can be done using the introspective methodology on which he bases his account. For
132
Louise Röska-Hardy
example, it is difficult to see how the “latching” process gets started using only the resources available within one’s first-person, introspectible access to occurrent mental states. On the contrary, associating “verbal labels” like ‘belief’ or ‘pain’ with internal states seems to require reference to something other than introspectively accessible, non-relational properties, e.g. such as environmental circumstances or overt behaviors. If our everyday mental concepts have no relation to such factors, it is unclear what reason we have to think that others have introspectively accessible mental states. Assuming for the sake of argument that the analogical inference in simulation from oneself to the other is not problematic, it remains the case that on Goldman’s account the introspectively accessible properties of one’s mental states stand in no relation to one’s environmental circumstances, perceptual situation or behavior. This raises a question about the introspection-simulation account of third-person ascription. How does an ascriber identify the initial inputs for a simulation routine, i.e. the mental states of the person to be simulated, if “external” factors have no relation to the contents of mental state concepts? It seems that an adequate account of mental state concepts and their mastery must take such “external” factors into consideration, even if it takes internal, non-relational properties to be definitive of mental state types. Since Goldman only claims that introspective access is necessary for mental concepts, not that it is sufficient, much depends on how he augments the introspective core of his account. However, it is clear that resources beyond first-person introspective access are needed to provide an adequate account of the content of everyday mental state concepts and their mastery. Thus, despite the fact that Goldman’s introspection-simulation approach addresses all three desiderata, it does not provide a satisfactory account due to the difficulties engendered by grounding the account of mental self-ascription on introspective access to mental states tokens.
4. Gordon’s ‘Ascent Routine’ Approach Robert Gordon introduces the notion of an ‘ascent routine’ to circumvent the problems created by the assumptions that mental self-ascription is based on introspective access and that simulation requires an analogical inference from oneself to the other (1995, 1996). He elaborates his simulation approach to third-person ascription of mental states in tandem with an account of mental self-ascription in terms of ‘ascent routines’. In contrast to Goldman, Gordon denies that simulation requires prior possession of mental state concepts. He also disputes Goldman’s claim that simulation relies on introspective access to
Self-Ascription and Simulation Theory
133
one’s own mental states and rejects the assumption that simulation involves an implicit analogical inference (1992, 1995). By introducing the notion of an ‘ascent routine’, Gordon develops a version of simulation theory that eschews both introspection and analogical inference in accounting for our ascriptions of mental states to ourselves and to others (1986, 1995, 1996, this volume). Nonetheless, he takes the simulation approach to have direct implications for an account of mental state concepts as well as for their mastery. Gordon’s version of simulation does not require an inference from oneself to another. Rather, in simulating another person, the ascriber “recenters her cognitive map” on the other so that the first-person pronoun ‘I’ refers “exclusively” to the individual on whom the ascriber’s egocentric map has been recentered (1995: 60). After this “egocentric shift” the ascriber ceases to be the referent of her uses of ‘I’; in her imagination the person being simulated becomes the referent of ‘I’ (1995: 55; 1992). In imaginative transformation the referent of the first-person singular pronoun changes, but the personal pronoun’s ties to perceptions, memories, actions and emotions, which are “keyed to [one’s] egocentric map”, remain constant, according to Gordon (1992: 32). He describes this recentering as “imaginatively transforming ourselves into other ‘first persons’”, as “identifying with” an individual within simulation or as “becoming” the other. Hence an egocentric shift on the part of the ascriber lies at the core of simulation, not introspection coupled with analogical inference. After the imaginative transformation into the other, a simulator directly ascribes the belief or decision generated within the scope of simulation to the other via an ascent routine, thus obviating the need for an analogical inference from herself to the other (1995: 63). Gordon emphasizes, “[O]nce a personal transformation has been accomplished, there is no remaining task of mentally transferring a state from one person to another, no question of comparing [the other] to myself. For insofar as I have recentered my egocentric map on [the other], I am not considering what [I] RMG would do, think, want and feel in the situation. Within the context of the simulation, RMG is out of the picture altogether (1995b: 56, original italics).
Thus, third-person mental ascription is, in essence, a case of mental “self”ascription to oneself-as-the-other within the context of simulation after imaginative transform or recentering. This is because embedding an ascent routine within a simulation of another allows one to ascribe mental states to the other directly by ascent routine. The ascent routine is the key to other-ascription and self-ascription of mental states in Gordon’s version of ST. Here, as in Goldman’s approach, mental ascription is inherently first-personal.
134
Louise Röska-Hardy
Gordon characterizes an ascent routine as a procedure which allows one “to get the answer to a question about oneself, and specifically about one’s mental states, by answering a question that is not about oneself, not about mental states at all” (1996: 15, original italics; 1995: 60). He contends that ascent routines allow us to ascribe beliefs and other mental states to ourselves accurately and reliably as well as non-inferentially and without appeal to evidence, e.g. without consulting environmental circumstances or behavior, as the TT would have it, or introspectible properties, as Goldman suggests. For example, one answers a question concerning a mental state, such as “Do you believe that Mickey Mouse has a tail?”, by asking oneself an “outward-looking” question about the world, “Does Mickey Mouse have a tail?” and then “ascending” with the answer from the “object-level” (“Yes, Mickey Mouse has a tail”) to a level which concerns the mental state of belief (“Yes, I believe that Mickey has a tail”) (1996: 15). Gordon calls this procedure an ‘ascent routine’ because it answers a question about belief, i.e. a question about a mental state that is about p (whether I believe M.M. has a tail), by answering a question “at a lower semantic level”, namely a question which is directly about p (whether M.M. has a tail). According to Gordon, the ascent routine allows one to move from “p” to the self-ascription of a mental state without exercising introspection, reasoning or inference. In his opinion, the ascent routine procedure reflects the way adults usually determine whether or not they believe that p; they simply ask themselves the question whether or not p. He claims that even those who possess the concept of belief rely on an ascent routine when answering questions about what they believe (1996: 16). In addition, he hypothesizes that there are ascent routines for propositional attitudes other than belief and extends the ascent routine proposal to the ascription of pain as well (1996). By introducing ascent routines, Gordon’s approach to the self-ascription of mental states avoids any appeal to introspection or inner recognition in contrast to Goldman’s. The ascent routine approach is non-introspective and nonrecognitional; it does not require that one detect an inner state in order to make a mental self-ascription. On the contrary, Gordon claims that in an ascent routine one directs one’s thoughts “outward” to the world (1996, this volume). He also maintains that the ascent routine procedure does not require that one recognize one’s own mental states as one’s own mental states, e.g. as my pain or my belief, in order to self-ascribe them, nor does it require that one possess the relevant mental state concepts (1996: 17). Only the ability to express the mental state is required, e.g. the ability to assert “Mickey Mouse has a tail” in the case of a belief.
Self-Ascription and Simulation Theory
135
Gordon points out that we can train a two-year-old child to prefix his utterances of “p”, e.g. “M.M. has a tail”, with the formula ‘I believe that ...’ or to prefix his expressions of desire with ‘I want ...’. As a consequence, the child will produce accurate and reliable utterances of the form ‘I believe that p’ or ‘I want q’ (1995, 1996). Gordon takes this to show that using an ascent routine to ascend from “p” to the level concerning mental states “I believe that p” does not require possession of the relevant mental state concept. The reliability of such mental “self-ascriptions” via an ascent routines is simply due to training, not evidence. However, he acknowledges that such answers do not qualify as “genuine, comprehending” self-ascriptions, even though the ascent routine allows children to generate the right answers to questions about their mental states, (1995b). A genuine, comprehending self-ascription of a belief, for example, requires more than proficiency in prefixing the linguistic form ‘I believe [that]’ to assertive utterances (1996). As Gordon emphasizes, training in prefixing the linguistic forms of self-ascription, ‘I believe [that]’ or ‘I want’, to verbal expressions of beliefs and desires does not afford a child a mastery of the concept of belief or desire. Comprehending ascriptions of belief require “conceptually prising one’s own present beliefs apart form the facts, so that, like the beliefs of another they can be at variance with the facts” (1995: 63). A child must understand that a belief that p can be false, that someone may fail to believe that p even though p is true, or that what one individual believes may differ from what another believes, i.e. that there are different points of view concerning the facts. Until children have mastered these points, Gordon says, “they will not have learned that they believe it is raining, or even that they have beliefs” (1995: 61), i.e. they will not grasp belief as a mental concept. They also will not understand that the question “Do you believe M.M. has a tail?” is a question about themselves, not about Mickey Mouse. Hence Gordon concedes that the ascent routine procedure by itself will not provide a child with the concept of belief or, more generally, with the concept of a mental state. As he notes, an ascent routine procedure will not even teach a child that ‘I’ refers to the speaker (1995: 59). Although ascent routines by themselves do not provide the components necessary for a mastery of mental concepts, Gordon suggests that the procedure does indicate how one might acquire the concept of a mental state like belief as well as the capacity to make genuine, comprehending ascriptions of mental states (1995). He asserts, [W]hat is needed to elevate such object-level thinking to an understanding of mental states as such is the ability to reconceptualize such ‘objects’ as the
136
Louise Röska-Hardy pain in the foot – or the facts about Mickey Mouse – as having a mental location. When the relevant ascent routine is used within a simulation ... a logical space is opened that enables us to think of pains and facts as located in (or at) individual minds. Thus simulation and ascent routines fit hand-inglove (1996: 19, original italics).
For Gordon the notion of (something’s being at) a mental location is central to the idea of the mental. In order to understand a state as mental, one must, he says, “[think] of pains and facts as located in (or at) individual minds” (1996: 19), realize that “a fact is a fact to some individual” (1996: 16), or that “a pain is mentally at a location; it is someone’s pain and not just a pain in a part of someone’s body” (1996: 18). This is achieved by embedding ascent routines in simulations of different persons or oneself at different times or by memory demotion (1995: 62). In order to acquire the idea of different mental locations or points of view regarding “the facts”, Goldman claims one must be able to simulate others and to employ ascent routines within simulations to make other-ascriptions of mental states. In order to self-ascribe mental states, one must be able to simulate another person, simulating oneself. Consequently, in Gordon’s ascent routine approach genuine, comprehending (present-tense) self-ascriptions of mental states “come last of all” (1995: 62-3). The first-person use of mental predicates is thus construed by Gordon as an “intersubjective achievement”, which “borrow[s] from the intersubjectivity of the public language, the language of ‘public properties of public objects’” through intersubjective training in ascent routines (this volume: pt. III). Gordon holds that ascent routines move directly from expression “p” to self-ascription “I \[p]” without introspection or inference, because they rest solely on the coordination of self-ascription with the verbal expression of belief (1995: 65, ftn.7; this volume: pt. III). In learning to speak a language, he suggests that we learn to link our outward expressions of mental states, e.g. our utterances, to other people’s beliefs about our mental states; this training is what makes ascent possible. For in learning to speak sincerely, we learn to utter “p” only if p obtains. The ensuing coordination between the expression of a mental state “p”, and the self-ascription of that mental state, “I \[p]”, grounds the ascent routine procedure. Ascent routines insure that the coordination involved in mental self-ascription is only with the verbal expression of belief – not with other verbal or non-verbal evidence (TT) or with introspective evidence (Goldman), because they are based on intersubjective training. In the case of belief, for example, training in making assertions about the world teaches us how “to piggyback on the public language” in Gordon’s words (this volume: pt. III).
Self-Ascription and Simulation Theory
137
The notion of the ascent routine is the linchpin of Gordon’s approach to the self-ascription and other-ascription of mental states. It underwrites his non-introspectionist account of first-person mental ascription and his nonanalogical account of third-person mental ascription as well as his claim that simulation does not require prior possession of mental state concepts. Gordon uses ascent routines to explain how we arrive at self-ascriptions of mental states, to provide the basis for an account of mental state concepts in terms of simulation and to indicate how we go about mastering mental state concepts. The notion of the ascent routine is essential to satisfying the three conditions on an adequate account of mental self-ascription. Thus, the hypothesis that we self-ascribe mental states to ourselves and to others by ascent routines merits close examination. If the ascent routine approach is to account for our mastery of mental concepts, it must indicate how first- and third-person mental ascriptions are related to each other. This seems to pose no problem on Gordon’s approach, because ascent routines are equally suited to identifying another’s beliefs as to identifying one’s own (1995: 60). Whether self-ascribing or other-ascribing mental states, one directs one’s attention “outward” to “the world” or “the facts” and then ascribes a mental state by an ascent routine either to oneself or to oneself-as-the-other. Importantly, the mental predicates employed in firstperson and third-person ascriptions are univocal by default, since first-person and third-person ascriptions are both essentially cases of first-person mental ascription of mental states via ascent routines. The crucial assumption in Gordon’s version of simulation and third-person mental ascription is that one can mentally transform oneself into the other. Third-person mental ascription requires a “personal transformation”, in which one recenters one’s egocentric map, considers “the world”, lets one’s cognitive abilities work and, in the case of belief, reports what is there by asserting “p”. The assertion “p” expresses one’s belief-as-the-other within the context of simulation. By an ascent routine one gets “I believe that p”. But why is producing the utterance “I believe that p” within the ascriber’s recentered firstperson perspective a case of ascribing a belief to another person? This question highlights two problematic aspects of Gordon’s account. First, what does the metaphor of mentally transforming oneself into the other amount to? Gordon’s other characterizations, e.g. “recentering one’s cognitive map”, “identifying with” or “becoming the other”, offer additional metaphors, not clarification. How does one “become the other” or “recenter one’s egocentric map”? The answer is important, because in Gordon’s view imaginative transformation or recentering eliminates the need for analogical inference from oneself to the other in third-person ascriptions of mental states. However, to
138
Louise Röska-Hardy
serve this purpose, talk about identifying with the other, recentering one’s cognitive map or imaginatively transforming oneself into the other must be understood literally. This raises the question of how one does it. It cannot simply be a matter of spatial or temporal perspective-taking. For, as hermeneuticists from Schleiermacher to Gadamer have emphasized, our beliefs, desires, and other thoughts determine not only how we view “the world”, but also what we take to be “the facts”.1 In order to recenter or personally transform oneself into the other, an affinity of mind is required. Recentering cannot generally be accomplished directly, e.g. on the basis of perception, because an ascriber will only rarely be in the same physical environment as the person she is simulating. And, she will never have exactly the same personal history as the person she attempts to simulate. Nonetheless, the ascriber must take into account those features of a situation which are salient for the other. This requires substantive assumptions about the other’s beliefs, desires, intentions, etc., which must be specified in contentful terms, i.e. treated as a belief that p, a hope that q, and so forth in order to generate further thoughts or decisions. The fact that all this transpires within the first-person point of view of the ascriber-as-the-other points to a second problem. This problem concerns the endpoint of simulation when one ascribes the result of the simulation to oneself-as-the-other (in the first-person) by an ascent routine. In order for the “self”-ascription within simulation to be understood as a genuine, comprehending other-ascription, a de-transformation must take place. Otherwise, the idea that one is ascribing a mental state to someone else has no purchase. This broaches the question of whether “flagging” mental states as states of the other is necessary in Gordon’s version of simulation (cf. 1992: 66 ftn. 12). Without answers to these questions, Gordon’s ascent routine account of third-person mental ascription is Pickwickean. Until the idea of mental transformation or recentering is more fully elaborated, the difference between first-person and third-person mental ascriptions and our grasp of this difference remains unclear. The ascent routine procedure itself is also problematic. Gordon claims that the ascent routine procedure shows that one can ascribe mental states without prior possession of mental state concepts. However, in order to get the ascent routine going, one must understand the initial question. In the Mickey Mouse example one must understand the predicate ‘believe’ in contrast to ‘Do you
1
Gordon acknowledges as much when he writes, “The most problematic adjustments to projection are adjustments, not in spatial or temporal perspective or in attitude, and thus in the ‘moving’ powers of various facts, but in the facts themselves”, (1992): 23.
Self-Ascription and Simulation Theory
139
hope/fear/desire that p?’ etc. Even yes/no-answers to the initial question presuppose by anaphora that the self-ascriber understands what it means to believe something, e.g. in contrast to hoping or fearing it. Moreover, one must understand the mental predicate ‘believe’ in order to ask the non-mental “object-level” question appropriate to a particular mental state. Consequently, an understanding of mental predicates is required in order to get the ascent routine procedure going. This implies that a grasp of mental state concepts is necessary as well, if mental predicates express concepts, as they surely do on the view that mental ascriptions are cognitively significant. Gordon discusses the ascent routine with respect to belief. Indeed, the ascent routine seems tailored to the case of belief and, arguably, belief constitutes a special case where the ascent routine is concerned. Typically, an indicative utterance of a sentence which does not contain an explicit performative formula will be construed as an assertion and assertion is generally taken to be the expression of belief. Plausibly, in the case of belief, the “ascent” from ‘p’ to ‘I believe that p’ does not require possession of the concept of belief, because the concepts of belief and assertion are internally related. This prompts the question whether there are ascent routine questions for mental states other than belief (cf. Goldman, forthcoming: sect. 3). Therefore, the ascent routine for belief does not show that simulation in general does not require prior possession of mental concepts. As a result, the ascent routine procedure does not supply the foundation for an account of mental state concepts or their mastery, as Gordon avers. Rather, the meanings of mental predicates and a grasp of mental concepts are presupposed in order to get the routine started. Gordon says that ascent routines “piggyback” on the intersubjectivity of public language and identifies public language as the language of “public properties of public objects” (this volume: pt. III). However, mental predicates precisely do not concern “public properties of public objects” in any straightforward sense. The meanings of mental predicates and the content of mental concepts require another explication. In this light Gordon’s claim that the ascent routine procedure can supply a “bootstrap” by which one can acquire mental state concepts and the capacity to make genuine, comprehending ascriptions of mental states must be reassessed.
5. Conclusion Neither Goldman’s introspectionist approach nor Gordon’s ascent routine approach provides an adequate account of the self-ascription of mental states.
140
Louise Röska-Hardy
Both have fatal shortcomings, even though each attempts to provide an account of the content and mastery of mental concepts to complement the advocated version of simulation. Goldman’s attempt to ground mental state concepts in internally detectable non-relational, categorial properties of mental state tokens neglects the connections between mental states and environmental circumstances and behavior. This leads to the “latching” problem and difficulties concerning third-person mental ascription. By contrast, Gordon’s ascent routine approach suggests how self-ascription of mental states could proceed without relying on introspective access. However, his account circumvents central questions about the contents of mental concepts and their mastery by invoking the idea of imaginative transformation. In the end, it presupposes an account of the meanings of mental expressions and the contents of mental concepts, rather than providing one. Thus, despite the fact that Goldman’s and Gordon’s accounts address all three desiderata for an adequate account of mental self-ascription, both fall short of being satisfactory. The shortcomings of Goldman’s and Gordon’s approaches to self-ascription should not, however, be taken to rehabilitate a version of the ‘Theory’Theory account. For the deficits of the ‘Theory’-Theory approach to concepts and concept mastery in first-person mental ascription remain in place. Rather, the failings of the TT and the two simulation accounts of mental self-ascription point to another approach. Taking a cue from Gordon, who follows Wittgenstein and Evans in several important points, we should once again consider the role that the use of language plays in making ascriptions of mental states to ourselves and to others. Gordon’s ascent routine approach highlights the use of language in expressing and characterizing ourselves with a view to the world and accords language acquisition an important role in ascribing mental states. However, in developing the ascent routine hypothesis Gordon fails to recognize and fully acknowledge the social character of mental ascription. In his approach Gordon emphasizes the Wittgensteinian point restated by Evans that in self-ascribing a belief “one’s eyes are, so to speak, ... directed outwards – onto the world” (1982: 225). However, Gordon’s emphasis on “the world” and “the facts” obscures the social dimension of language use and the extent to which social interaction in specific contexts provides the foundation of the ascent routine. Although a full consideration of this dimension requires another paper, some brief remarks should indicate the approach to be investigated. Importantly, the language we employ in making mental ascriptions is both acquired and used in social settings from the very start. In general, we acquire words and learn to use them from other speakers in social contexts. This is as true of expressions which pertain to the mental as it is of expressions which
Self-Ascription and Simulation Theory
141
refer to “public objects”. Although the meanings of some words are conveyed by reference to “the public properties of public objects”, e.g. through ostensive teaching in a social context, this is clearly not true of mental predicates. But this should not lead us to conclude that we acquire mental predicates by linking “verbal labels” to internal representations based on introspectible properties, i.e. through introspection cum private ostension. Rather, we acquire mental predicates and their cognates and learn the conditions for their application by being trained, shown and instructed by mature speakers’ uses of these predicates in specific situations, e.g. by their ascribing mental states to themselves and to us in social settings. As a result of this training during the acquisition of language and our use of these expressions in making ourselves understood to others, we hold mental expressions applicable in certain circumstances, i.e. under those conditions in which we have acquired them and used them in making ourselves understood. These circumstances or conditions insure that our uses of mental predicates are aligned with intersubjectively accessible features of utterance contexts, including spontaneous pre-linguistic behaviors and reactions, linguistic and non-linguistic behavior and environmental circumstances. This suggests a point of departure for an account of the cognitive significance of mental predications, the univocality of mental predicates and the objectivity or intersubjectivity of first-person and third-person mental ascriptions. For in language acquisition, other speakers offer us words to express and characterize our mental conditions in view of features in the physical, social and linguistic context that they find salient to the use of these words. They thereby initiate us into a way of understanding ourselves and others, which crucially depends on the ability to use language. As we master our language with the help of others, we become capable of expressing and characterizing our mental condition in words.1 When the need arises, words come to mind immediately as a result of our socialization into a linguistic community. In the first-person case, everyday characterizations of occurrent mental states are “direct” in the sense that they are not inferred nor are they made on the basis of observation or introspective evidence.2 They are also privileged and authoritative, although not incorrigible, because they are made against the background of our language acquisition and history of language use that endows the our words with meaning. For, we express and 1
Note that there is no distinction between characterizing a mental state and expressing a mental state in uses of psychological verbs in the first-person singular; in first-person singular uses they come to the same thing.
2
There are, of course, exceptions to the ordinary case, e.g. when one “observes” and reports on one’s stream of consciousness or when one reflects on why one has certain beliefs, desires, hopes.
142
Louise Röska-Hardy
characterize our mental states in our own words.1 Moreover, we use these same words to characterize the mental states of others. However, the subject term in self-ascriptions differs from other-ascriptions in that it refers to the person producing the characterization. In basic cases of mental self-ascription the characterization of the mental state \[p] is indexed by the explicit use of ‘I’. This again suggests that the role of language in self-ascriptions should be more carefully considered. These brief remarks suggest a third approach to the self-ascription of mental states, which differs from the ‘Theory’-Theory and from the simulation accounts of Goldman and Gordon, although it has affinities with each. This approach focuses on the role of language and language use in social interaction and seeks to account for the meanings of mental predicates, our everyday concepts of mental states and our mastery of mental concepts in these terms. This approach suggests that we arrive at self-ascriptions of mental states by having been socialized into the use of mental predicates to express and characterize our mental condition, when the occasion arises. In learning a first language from mature speakers we acquire a linguistically communicated framework for explaining the thoughts and actions of persons in the natural and social world. The meanings of mental predicates and the content of mental concepts derive from this framework, which one might call the ‘person framework’. This linguistically communicated framework is characterized by connections among concepts of mental states, environmental factors, behavior and other mental state concepts, which inform the use of mental predicates in everyday mental ascriptions. These connections are in the first instance conceptual, not empirical.2 The person framework is not a theory in any substantive sense of the term. It is not “applied” in making mental ascriptions, nor does it exhibit nomological principles or causal regularities. It also clearly differs from Goldman’s introspection-simulation approach and Gordon’s ascent routine approach to mental ascription. However, it agrees with the two simulation approaches on a central point, namely, that the first-person point of view, the point of view indexed by the explicit use of ‘I’, is essential to mental concepts. Whether such a third approach to self-ascription of mental states will prove adequate remains to be seen. However, in view of the philosophical issues 1
This important point has been consistently emphasized by Donald Davidson, cf. RöskaHardy (1997).
2
This thesis is articulated and defended in: Röska-Hardy (1995). The person framework is not a version of the ‘rationality approach’ as characterized by Goldman (1989), although I cannot argue this point here. For example, it does not presuppose ideal rationality or canons of rationality.
Self-Ascription and Simulation Theory
143
involved in the ‘Theory’-Theory vs. Simulation Theory debate, it merits further consideration.
References Carruthers, Peter 1996 “Simulation and Self-Knowledge” in: Theories of Theories of Mind, P. ——— and Peter K. Smith (eds.), Cambridge: Cambridge University Press. ——— and P. K. Smith (eds.) 1996, Theories of Theories of Mind, Cambridge: Cambridge University Press. Churchland, Paul 1970 “The Logical Character of Action Explanation” in: Philosophical Review 79: 214-236. ——— 1981 “Eliminative Materialism and the Propositional Attitudes”, Journal of Philosophy 78: 67-90. ——— 1988 Matter and Consciousness, Cambridge, MA: MIT Press (1984, revised edition). ——— 1989 A Neurocomputational Perspective: the Nature of Mind and the Structure of Science, Cambridge, MA:MIT Press. Davies, Martin 1994 “The Mental Simulation Debate”, Proceedings of the British Academy 83: 99127. ——— and T. Stone (eds.) 1995a, Folk Psychology, Oxford: Blackwell’s. ——— and T. Stone (eds.) 1995b, Mental Simulation, Oxford: Blackwell’s. Evans, Gareth 1982 The Varieties of Reference, John McDowell (ed.), Oxford: Oxford University Press. Fodor, Jerry 1983 The Modularity of the Mind, Cambridge, MA: MIT 1983. ——— 1987 Psychosemantics: the Problem of Meaning in the Philosophy of Mind, Cambridge, MA: MIT. ——— 1992 “A Theory of the Child’s Theory of Mind” in: Cognition 44: 283-96. Goldman, Alvin 1989 “Interpretation Psychologized“ in: Mind and Language 4: 161-85. ——— 1992 “Empathy, Mind and Morals” in: Proceedings and Addresses of the American Philosophical Association vol. 66, no.3.; and in: Mental Simulation, Davies, Martin and T. Stone (eds.), Oxford: Blackwell’s 1995. ——— 1993a “The Psychology of Folk Psychology” in: Behavioral and Brain Sciences 16 (1993): 15-28. ——— 1993b Philosophical Applications of Cognitive Science, Boulder, CO: Westview Press. ——— (forthcoming) “The Mentalizing Folk” in: Metarepresentation, Dan Sperber (ed.), New York: Oxford University Press. ——— (MS.) “Desire, Intention, and the Simulation Theory”. Gopnik, Alison 1993 „How We Know Our Own Minds: the Illusion of First-Person Knowledge of Intentionality“ in: Behavioral and Brain Sciences 16: 1-14. Gordon, Robert 1986 “Folk Psychology as Simulation” in: Mind and Language 1: 158-71. ——— 1992 “The Simulation Theory: Objections and Misconceptions” in: Mind and Language 7: 11-34. ——— 1995b “Simulation without Introspection or Inference from Me to You” in: Mental Simulation, M. Davies and T. Stone (eds.), Oxford: Blackwell’s. ——— 1996 “’Radical’ Simulationism” in: Theories of Theories of Mind, P. Carruthers and P. K. Smith (eds.), Cambridge: Cambridge University Press. ———(this volume) “Sellars’s Ryleans Revisited” in: Proto-Soziologie vol. 14.
144
Louise Röska-Hardy
Heal, Jane 1986 “Replication and Functionalism” in: Language, Mind and Logic, J. Butterfield (ed.), Cambridge: Cambridge University Press. ——— 1995b “How to Think about Thinking” in: : Mental Simulation, M. Davies and T. Stone (eds.), Oxford: Blackwell’s. ——— 1996a “Simulation, Theory and Content” in: Theories of Theories of Mind, P. Carruthers and P. Smith (eds.), Cambridge, Cambridge University Press. ——— 1996b “Simulation and Cognitive Penetrability” in: Mind and Language 11:44-67. ——— 1998 “Co-cognition and Off-line Simulation: Two Ways of Understanding the Simulation Approach” in: Mind and Language 13: 477-498. ——— (this volume) “Understanding Other Minds from the Inside”, and in: Current Issues in the Philosophy of Mind, A. O’Hear (ed.), Cambridge: Cambridge University Press 1998: 83-99. Leslie, Alan 1987„Pretense and Recognition: the ‘Origins’ of the Theory of Mind” in: Psychological Review 94: 412-26. Lewis, David 1966 “An Argument for the Identity Theory” in: Journal of Philosophy 63: 17-25. ——— 1970 “How to Define Theoretical Terms” in: Journal of Philosophy 67: 427-46. ——— 1972 “Psychophysical and Theoretical Identifications” in: Australasian Journal of Philosophy 50: 249-58. Moran, Richard 1997 “Self-Knowledge: Discovery, Resolution, and Undoing” in: European Journal of Philosophy 5: 141-161. Perner, Josef 1991 Understanding the Representational Mind, Cambridge, MA: MIT Press. Rosenthal, David 1992 “Thinking that One Thinks” in : Consciousness, Martin Davies and G. Humphreys (eds.), Oxford: Blackwell’s: 197-223. ——— 1997 “A Theory of Consciousness” in: The Nature of Consciousness, Ned Block, O. Flanagan and G. Güzeldere (eds.), Cambridge, MA: MIT Press: 729-753. Röska-Hardy, Louise 1995 “Denken, Handeln und Erklärung durch Gründe: Ein Diskussionsbeitrag” in: Deutsche Zeitschrift für Philosophie. Bd. 7 Heft 2 (1995): 259-270. ——— 1997 “Interpretation und Erste-Person-Authorität” in: Donald Davidsons Philosophie des Mentalen, W. R. Köhler (ed.), Paderborn: Schöningh: 139-169. ——— 2000a “‘I’ and the First-Person Perspective” in: Proceedings of the Twentieth World Congress of Philosophy 1998, vol. 6, Bowling Green: Philosophy Documentation Center. ——— 2000b “Idealism and the ‚I’ of Self-Ascription” in: Indexikalität und Bewußtsein/Indexicality and Idealism, Audun Øfsti and Truls Wyller (eds.), Paderborn: Mentis Verlag: 53-68. ——— (forthcoming) “Zuschreibungen und Personenverstehen” in: Dresdener Hefte für Philosophie Bd. II. Shoemaker, Sydney 1993 “Special access lies down with the theory-theory” in: Behavioral and Brain Sciences 16: 78-79. Stich, Stephen and S. Nichols 1992 “Folk Psychology: Simulation or Tacit Theory?” in: Mind and Language 7: 35-71. Strawson, Peter F. 1959 Individuals: an Essay in Descriptive Metaphysics, London: Methuen. Tugendhat, Ernst 1979 Selbstbewußtsein und Selbstbestimmung, Frankfurt am Main: Suhrkamp/ Self-Consciousness and Self-Determination 1989. Wittgenstein, Ludwig 1953 Philosophical Investigations/Philosophische Untersuchungen, 2nd Edition, New York: Macmillan 1958.
Why Intentional Systems Theory Cannot Reconcile Physicalism
145
BRIAN P. MCLAUGHLIN Why Intentional Systems Theory Cannot Reconcile Physicalism With Realism about Belief and Desire Abstract In this paper, I examine Daniel Dennett’s well-known intentional systems theory of belief and desire from the perspective of physicalism. I begin with a general discussion of physicalism. In the course of that discussion, I present familiar ways that one might attempt to reconcile physicalism with belief-desire realism. I then argue that intentional systems theory will not provide a reconciliation of physicalism and belief-desire realism.
The purpose of this paper is consider whether Daniel Dennett’s well-known intentional systems theory can be used to reconcile physicalism with realism about beliefs, desires, intentions, and the like.1 I begin with a discussion of physicalism itself. Physicalism is a thesis about what there actually is. It is the thesis that all there actually is whatever there has to be for the actual world to be exactly as it is physically.2 That is to say, it is the thesis that all there actually is is whatever there has to be for the actual world to be exactly as it is in every physical respect – in respect of what physical objects there are, what physical properties they have, what physical relations they bear to other objects, what physical laws govern their behavior, etc.3 Thus, if physicalism is true and there really are people, cities, and countries, then, given the way the world is physically, there have to be people, cities, and countries. Contrapositively, if there really
1
As will become apparent in due course, I shall be concerned exclusively with the mature “real patterns” version of intentional systems theory found in Dennett 1987, not the earlier instrumentalist version found in 1978.
2
Frank Jackson (1998) formulates physicalism as the thesis that any minimal physical duplicate of the actual world is a duplicate simipliciter of the actual world. By a minimal physical duplicate of the actual world, he means a physical duplicate of the actual world that contains nothing other than what is required to be such a physical duplicate. My formulation of physicalism is logically equivalent to his; and I am indebted to his discussion of physicalism.
3
It is a nontrivial problem to say what‘physical’ means in‘physicalism’. I favor the view that the notion of physicality must be tied somehow to current physics. But I shall not discuss the nature of the tie here. Instead, I shall rely on a rough and ready intuitive notion of physicality.
146
Brian P. McLaughlin
are people, cities, or countries, yet there do not have to be such things for the actual world to be as it is physically, then physicalism is false. Picturesquely, if physicalism is true, then to create an exact duplicate of the actual world, all Mother Nature would have to do is to create an exact physical duplicate of it. Her work would then be done. She would have, thereby, created a world exactly like the actual world in every respect whatsoever – a duplicate simpliciter of the actual world. If, however, after creating an exact physical duplicate of the actual world, Mother Nature would have to go on to put people or anything else into the world, then physicalism is false. By the totality of physical facts, let us mean all the physical facts, particular and general, nomological and counterfactual, and a “that’s all” fact to the effect that the physical facts in question are all the physical facts. Then, the point can be put less, picturesquely, as follows: physicalism (i.e., the thesis that all there actually is is whatever there has to be for the actual world to be exactly as it is physically) is true if and only if the conjunction of physicalism and the totality of physical facts entails all the facts.1 Of course, if physicalism is true, it is only contingently true. Even if all there actually is is whatever that has to be for the actual world to be exactly as it is physically, things might have been otherwise. There might have been things in addition to those required for the world to be exactly as it is physically. There might have been facts not entailed by the totality of physical facts and the assumption of physicalism. When, in opposition to physicalism, it is claimed that there is a certain actual phenomenon that is not required for the actual world to be exactly as it is physically, physicalists have two options. They can deny that there actually is any such phenomenon, and so take an eliminativist or irrealist position visà-vis it. Or they can take a realist view of the phenomenon and argue that, despite appearances to the contrary, it is in fact required for the actual world to be the way it is physically. There are some would-be phenomena that all physicalists are eliminativists about. They are all eliminativists about disembodied spirits, for instance. But in some cases physicalists are divided. Some physicalists are eliminativists of a 1
Three points of clarification are in order. First, the notion of entailment that I am appealing to here is purely modal. Thus, for instance, any necessary fact is trivially entailed by any fact whatsoever. Second, pysicalists do not maintain that the totality of physical facts entail all of the facts. For they allow that a world can be a physical duplicate of the actual world without being a duplicate simpliciter of it. It is only in conjunction with the contingent thesis of physicalism that the totality of physical facts entail all of the facts. Third, unlike Jackson 1998, I hold that the entailment in question need not to be a priori; and, while I am a physicalist, I do not believe that the entailment will be a priori. I shall not, however, pursue these matters here.
Why Intentional Systems Theory Cannot Reconcile Physicalism
147
very radical sort: they are eliminativists about beliefs and desires (and intentions, hopes, wishes, fears, and the like).1 It is fair to say, however, that most physicalists are realists about these matters since they believe they can see, if only through a dark glass dimly, how a plausible case could be made that, given the way the actual world is physically, there must be beliefs and desires. Our main concern here is, as we noted, whether intentional systems theory will enable physicalists to reconcile belief-desire realism with physicalism. But before pursuing our main question, let us first ask the following more general one: How can physicalists argue that a certain phenomenon is required for the actual world to be exactly as it is physically? If the phenomenon in question is straightforwardly physical, then the issue is simply a matter of whether there actually is such a phenomenon; and, so, physicalists that are realists about the phenomenon would simply offer empirical evidence that it exists. If, however, the phenonmenon is not straightforwardly physical, then there are two options available to the physicalist. One is to argue that, despite appearances to the contrary, the phenomenon in question is nevertheless physical. This is the option taken vis-à-vis types of sensations (such as pain) by physicalists that argue that types of sensations are types of neural states.2 Another option is to concede that the phenomenon in question is not itself physical, but to argue that it is nevertheless the case that it is required for the actual world to be as it is physically. The second option is the one that physicalists who countenance nonphysical, functional state types (or properties) take. A functional state is a secondorder state: it is the state of being in some state that plays a certain counterfactual or nomological role. Any state that plays the role in question is a realizer of the functional state; and a functional state may be multiply realizable in that there may be more than one state that can actually realize it. A functional state type is nonphysical if and only if it is logically possible for it to be realized by a nonphysical state.3 Physicalists can countenance nonphysical, functional states provided that the totality of physical facts in conjunction with physicalism entail the laws (if any) and counterfactuals in which they figure, and entail as well their actual world-wide pattern of distribution. The totality of physical facts will imply that there are certain functional state types. And physicalism will imply that (a) they are the only functional state types realizable in the actual world, and, that (b) certain physical states are their only 1
See Churchland 1978; and Stich 1999.
2
See, e.g., Hill 1992.
3
It may be that virtually every functional state is a nonphysical state since virtually every functional state is such that it is logically possible for it to be realized by a nonphysical state. I leave the issue open.
148
Brian P. McLaughlin
realizers in the actual world.1 For any physical state type, the fact that it has a certain world-wide pattern of distirbution will be a physical fact. It thus follows that, if physicalism is true, then the world-wide pattern of distribution of the functional states in question will be entailed by the totality of physical facts and the assumption of physicalism. Moreover, the totality of physical facts (nomological, counterfactual, and otherwise) will, in conjunction with physicalism, entail all the counterfactuals and laws (if any) in which functional states in question figure. Consider the following paradigm dispositions: water-solubility, fragility, and ductability. According to the leading view of dispositions, dispositions are functional states, and their bases are the states that realize them.2 Thus, for instance, water-solubility is (roughly) the state of being in some state that disposes its occupants to dissolve in water. And being composed of sodium chlorine is a basis of the disposition, though of course not the only one. Fragility is (roughly) the state of being in some state that disposes its occupant to shatter when struck; and it is realized by the various crystalline structures that are its bases. Water-solubility and fragility may be nonphysical, functional states since it may be logically possible for them to be realized by structural states of nonphysical substances made of, say, “ectoplasm.” A nonphysical ectoplasmic being might be water-soluble or fragile. Likewise, ductability (roughly, the state of being in a state that disposes its occupants to change shape under slight pressure) may be a nonphysical, functional state; for perhaps it is logically possible for there to be ductable ghosts made up of nonphysical “spooky stuff.” Of course, physicalists cannot countenance the existence of “ectoplasm” or ghosts made of “spooky stuff;” but they can countenance water-solubility, fragility, and ductability, even though it is logically possible for them to be possessed by such nonphysical things. For physicalist can argue that, on the evidence, the actual world-wide pattern of distribution of water-solublity, fragility, and ductibility, and all the laws and counterfactuals in which these dispositions figure, are entailed by the totality of physical facts in conjunction with physicalism. So, water-solubility, fragility, and ductibility are, on the evidence, physicalistically kosher, even though it is logically possible for them to have nonphysical bases. Let us turn, then, to types of beliefs and desires. They are certainly not straightforwardly physical; indeed, they might well be thought paradigms of nonphysicality. And few physicalists would argue that, despite appearances to 1
It does not matter to physicalism whether the functional states have other realizers, even nonphysical realizers, in other possible worlds.
2
See McLaughlin 1994 for further discussion, and for references.
Why Intentional Systems Theory Cannot Reconcile Physicalism
149
the contrary, they are nevertheless physical. Instead, physicalists that are realists about types of beliefs and desires typically argue as follows: beliefs and desires are so related to certain dispositions or functional states that, given that the world-wide pattern of distribution of those dispositions or functional states and the laws and counterfactuals in which they figure are entailed by the conjunction of physicalism and the totality of physical facts, the world-wide pattern of distribution of beliefs and desires and the laws and counterfactuals in which they figure are so entailed.1 Consider logical behaviorism.2 Logical behaviorists maintain that each type of belief or desire is, by definition, a certain disposition to peripheral bodily behavior (i.e., overt or outward bodily behavior, bodily behavior involving movements of the periphery of the body). It is consistent with logical behaviorism that there is a ghost made of spooky stuff that has beliefs and desires, for such a ghost might have the appropriate dispositions to peripheral behavior. Logical behaviorism thus does not imply physicalism. (But, of course, we knew that anyway since logical behaviorism purports to be necessarily true and physicalism is a contingent truth.) Logical behaviorism is, however, compatible with the conjunction of physicalism and realism about beliefs and desires. And, more importantly for present purposes, were logical behaviorism true, one could appeal to it to try to reconcile physicalism with belief-desire realism. For a physicalist could argue that the actual world-wide pattern of distribution of dispositions to peripheral behavior is entailed by the totality of physical facts in conjunction with physicalism. And the physicalist could further argue that since logical behaviorism is true, the actual world-wide pattern of distribution of beliefs and desires is so entailed. Now, to be sure, it is generally acknowledged that logical behaviorism is false: it is not the case that individual beliefs and desires are, by definition, individual dispositions to peripheral behavior. So, logical behaviorism won’t serve to reconcile physicalism with belief-desire realism. But the point to note for now is simply that logical behaviorism is the sort of theory that were it true, one could appeal to it to square physicalism with realism about beliefs and desires. While it doesn’t entail physicalism, it is, so to speak, physicalismfriendly. For it identifies beliefs and desires with phenomena that we are in a position to see are physically kosher; that is to say, it identifies beliefs and 1
Hereafter, to avoid prolixity, I shall speak simply of the world-wide pattern of distribution of some state or property being entailed by the totality of physical facts in conjunction with physicalism, rather than of the world-wide pattern plus the laws and counterfactuals in which the state or property in question figures being so entailed. But it should be understood that the silent clause is required for the points that I will be making.
2
Ryle 1949.
150
Brian P. McLaughlin
desire with phenomena that we are in a position to see must exist, given the way the actual world is physically. Consider, then, a progeny of logical behaviorism, analytical functionalism. It comes in two versions. On one version, beliefs and desires are, by definition, certain functional states specifiable in physical and topic neutral terms, on the other version they are, by definition, states that realize certain functional states of that sort. Of course, neither version entails that physicalism is true. However, both versions are compatible with the conjunction of physicalism and realism about beliefs and desires. And, more importantly for present purposes, physicalists who embrace either version can appeal to it to vindicate beliefdesire realism. For they can argue that given that the actual world-wide pattern of distribution of the functional states in question is entailed by the totality of physical facts in conjunction with physicalism, the actual world-wide pattern of distribution of beliefs and desires is so entailed. Like the first version of analytical functionalism, scientific functionalism implies that beliefs are desires are certain functional states specifiable in physical and/or topic neutral terms.1 However, unlike analytical functionalism, scientific functionalism does not claim that such identity claims are true by definition. According to scientific functionalism, the identities are not knowable a priori (as, for instance, it is not knowable a priori that water = H2O). They are, rather, a posteriori justifiable on the basis of evidence and supraempirical considerations of overall coherence and theoretical simplicity. But while a posteriori, the identities will, nevertheless, be necessary.2 Thus, a physicalists could appeal to scientific functionalism to vindicate belief-desire realism by arguing that since the actual world-wide pattern of distribution of such functional states is entailed by the conjunction of physicalism and the totality of physical facts, the actual world-wide pattern of distribution of beliefs and desires is so entailed. Thus, if scientific functionalism is true, then physicalists could appeal to it to reconcile physicalism with realism about beliefs and desires. Turn, at last, to another progeny of logical behaviorism, one that more closely resembles its parent than does analytical functionalism, namely, Dennett’s intentional systems theory. Intentional systems theory is a kind of holistic logical behaviorism according to which the belief-desire profile of a system characterizes the system in terms of its global pattern of dispositions to peripheral behavior. The bases of the dispositions are treated as “black boxes” 1
Fodor’s (1975) language of thought theory of beliefs and desires is an example of a scientific functionalist theory. According to that theory, beliefs and desires are dispositions to compute in certain ways with mentalese sentences.
2
See Kripke 1971, and Kripke 1980, on the necessity of identity.
Why Intentional Systems Theory Cannot Reconcile Physicalism
151
by theory. For, according to the theory, it is irrelevant to a belief-desire profile what is in the black boxes, what the bases for the dispositions in the pattern are like; all that matters is the global pattern of dispositions itself; it is irrelevant what grounds the pattern. And the theory eschews any attempt to pair individual beliefs and desires with individual dispositions to peripheral behavior. According to the theory, the entire pattern of beliefs and desires of a system is its global pattern of dispositions to peripheral behavior.1 If this presentation of intentional systems theory is correct, then it is open to a physicalist to endorse the theory and to try to vindicate belief-desire realism by appeal to it. For it would be open to a physicalist to argue that since in fact the world-wide distribution of global patterns of dispositions to peripheral behavior of systems is entailed by the totality of physical facts in conjunction with physicalism, the world-wide distribution of beliefs and desires is entailed by the totality of physical facts in conjunction with physicalism. On the evidence, the world-wide pattern of distribution of global patterns of dispositions to peripheral behavior of systems is indeed entailed by the totality of physical facts and the assumption of physicalism. Thus, given that there is such an entailment, whether this appeal to intentional systems theory would vindicate belief-desire realism turns on whether intentional systems theory is true. I have argued elsewhere (with John O’Leary Hawthorne)2 that, while a definite improvement on logical behaviorism, intentional systems theory nevertheless fails to escape all of the pitfalls of that mistaken doctrine. For, like logical behaviorism, it asserts too intimate a tie between beliefs and desires and 1
See, for example, Dennett 1987, p.58. I should note that this presentation is incomplete in two ways. First, Dennett is an externalist about certain aspects of mental content. He distinguishes a system’s organismic contribution to a belief-desire profile from the environmental contribution to such a profile (1987, 134). A system’s global pattern of dispositions to peripheral behavior is only its organismic contribution to its belief-desire profile. Thus, intentional systems theory is the theory that the belief-desire profile of a system is a characterization of the system in terms of its global pattern of peripheral behavioral dispositions within an environmental context. Second, Dennett (1987, p. 40) holds that a system’s relationship to its environment is such that there is an indeterminacy of mental content, and so that the beliefdesire profile of the system is an interpretative characterization of it in terms of its global pattern of peripheral behavioral dispositions within an environmental context. (For detailed textual defense of this characterization of intentional systems theory, see McLaughlin and O’Leary-Hawthorne 1994.) I shall ignore the issue of belief-desire profile indeterminacy here; if anything, it would appear to make trouble for any would-be reconciliation of physicalism with belief-desire realism that appeals to intentional systems theory. But I shall not push such a line since I think that belief-desire realism can tolerate systematic indeterminacy. I shall also ignore mental content externalism. Suffice it to note that Dennett holds that the relevant environments are physical environments, or at least supervene on physical environments.
2
See McLaughlin and O’Leary Hawthorne 1994.
152
Brian P. McLaughlin
dispositions to peripheral behavior. As Dennett readily acknowledges, intentional systems theory implies the following supervenience thesis: necessarily, if two organisms are exactly alike in respect of their dispositions to peripheral behavior, then they are exactly alike in respect of their beliefs and desires.1 And that supervenience thesis is false. I presented two sorts of counterexamples to the supervenience thesis, one due to Ned Block,2 the other my own. The Block example is of a gigantic look-up table. For any believer-desirer, there will be some possible look-up table, “a Blockhead,” that has exactly the same dispositions to peripheral behavior as that believer-desirer.3 The global pattern of dispositions to peripheral behavior of the Blockhead is exactly the same as that of the believerdesirer, but the internal architecture of the Blockhead is inadequate to support beliefs and desires. Since the Blockhead is devoid of beliefs and desires, the supervenience thesis is false. The other example is of a creature named Manthra that is, in outward appearance, exactly like an extraordinarily large (much larger than they come) preying mantis. In addition to looking like a gigantic preying mantis, Manthra is, moreover, exactly as a gigantic preying mantis would be in its peripheral behavioral dispositions. Moreover, the part of Manthra’s brain that controls its peripheral behavioral dispositions is preying mantis-like. However, a large part of Manthra’s brain is very much like a normal human brain. It is just that the human-like part exerts no control over peripheral behavior. Manthra has a rich 1
He has been very explicit about this. He says: “I unhesitatingly endorse the claim that necessarily, if two organisms are behaviorally exactly alike, they are psychologically exactly alike” (Dennett 1993, p.923). In the context in which he makes this supervenience claim, it is clear that by‘behaviorally exactly alike’ he means exactly alike in respect of dispositions to peripheral behavior, and by‘psychologically exactly alike’ he means exactly alike in beliefdesire profiles. This supervenience thesis is, of course, not true to Dennett’s mental content externalism. But it could be restated to include a reference to environment in the subvenient base, or it could be recast to say that a system’s organismic contribution to a belief-desire profile supervenes on its dispositions to peripheral behavior. I shall, however, follow Dennett in omitting these qualifications in the present discussion. For nothing I shall say below would be essentially effected by them. My criticisms of the supervenience thesis below could be recast to accommodate any such recastings of the supervenience thesis.
2
Block 1981.
3
“Blockhead” is Frank Jackson’s (1993) apt term. Block himself (1981, p.20) describes a lookup table that will pass the Turing Test (introduced in Turing 1950). But Block and Jackson are well aware that we can extrapolate to a look-up table that mediates the transition from all sorts of simulus inputs to all sorts of motor ouputs in such a way as to match the overt or peripheral responses of any human being. It is logically possible for any pattern of dispositions to peripheral behavior (that is, in principle, nonintentionally characterizable) to be possessed by a Blockhead. (Some years ago, D.M. Mackay (1952) pointed out that any peripheral behavioral test for any mental attribution could in principle be passed by an automaton.)
Why Intentional Systems Theory Cannot Reconcile Physicalism
153
mental life, full of belief and desire. Realizing that its peripheral behavior is completely controlled by automatic subpersonal mechanisms, Manthra accepts its fate stoically. There is, however, another possible creature that is outwardly exactly like Manthra, not only in appearances, but also in its peripheral behavioral dispositions, but that has a preying mantis-like brain through and through. That possible creature and Manthra differ in respect of beliefs and desires, despite being exactly alike in respect of peripheral behavioral dispositions. Thus, the supervenience thesis is false. In the case of Manthra, we find behavioral dispositions that are exactly similar to those of a creature devoid of belief and desires (or with only a very impoverished belief-desire profile), yet Mantha’s internal architecture supports a rich pattern of beliefs and desires. The tie between belief-desire profiles and global patterns of dispositions to peripheral behavior is, at best, only criteriological. Global patterns of dispositions to peripheral behavior (relative to an environment) are, at very best, only defeasible sufficient conditions for belief-desire profiles; they perhaps provide us with prima facie reasons for certain belief-desire profiles, but those reasons are defeasible. Blockheads show that no global pattern of dispositions to peripheral behavior is a sufficient condition for a belief-desire profile. And Manthra shows that no such pattern is a necessary condition for any beliefdesire profile. Like logical behavorism, intentional systems theory goes wrong in its peripheralism. Now in response, Dennett1 has acknowledged that Blockheads and Manthra are counterexamples to the supervenience thesis when ’necessarily’ is taken to mean’ it is logically necessary that’. But he maintains that intentional systems theory does not imply the supervenience thesis so understood. Intentional systems theory implies the supervenience thesis, but only when ’necessarily’ is understood in a weaker sense than’it is logically necessary that’. He tells us that in using ’necessarily’ in stating the supervenience thesis, he meant to be expressing “real necessity,” not truth in literally every possible world. In places, his response seems to suggest that the kind of necessity he has in mind is nomological necessity. But he does not explicitly commit himself to that. And, unfortunately, he tells does tell us what, exactly, he means by “real necessity.” Presumably, Blockheads and Manthra are not counterexamples to the supervenience thesis if the necessity at issue is merely nomological necessity, for presumably they are nomologically impossible (as is, perhaps, a ninetyfoot tall human being or a signal that travels faster than the speed of light). So, let us suppose that intentional systems theory is intended to imply only that it 1
See Dennett 1994, pp.517-22.
154
Brian P. McLaughlin
is nomologically necessary that if two systems have exactly the same global pattern of dispositions to peripheral behavior, then they have exactly the same beliefs and desires. And let us call this the nomological supervenience thesis, and call the stronger supervenience thesis, with “necessarily” meaning “it is logically necessary that,” the logical supervenience thesis. While Blockheads and Mantra yield counterexamples to the logical supervenience thesis, they do not yield counterexamples to the nomological one since they are nomologically impossible. If, however, intentional systems theory implies only the nomological supervenience thesis, then the theory would not be the theory that the beliefdesire profile of a system is a characterization of the system in terms of its global pattern of peripheral behavioral dispositions. For the theory would not entail that the global pattern of beliefs and desires of a system is its global pattern of peripheral behavioral dispositions. The system will have two “real patterns,” a real pattern of beliefs and desires, and a real pattern of dispositions to peripheral behavior. And, according to the theory, the link between the patterns would be merely nomological. The belief-desire profile of a system would be based on the system’s global pattern of peripheral behavioral dispositions just in that the global pattern would nomologically necessitate the belief-desire profile. But if this were how intentional systems theory should be understood, then even if the theory is true, it would not reconcile physicalism with realism about belief and desire. The reason is this. Since a system’s global pattern of peripheral behavioral dispositions merely nomologically suffices for its beliefdesire profile, from the fact that the totality of physical facts in conjunction with physicalism entails the world-wide distribution of global patterns of peripheral behavioral dispositions of systems, it would not follow that the totality of physical facts in conjunction with physicalism entails the worldwide distribution of beliefs and desires. For there will be some possible world (albeit a nomologically impossible one) with exactly the same distribution of patterns of global dispositions to peripheral behavior as the actual world, but with a different distribution of beliefs and desires or no beliefs and desires at all. Thus, even if intentional systems theory so understood were true, it would not enable physicalists to vindicate belief-desire realism.1 It should be apparent that if the necessity in the supervenience thesis is understood as anything weaker than truth in every possible world, then the
1
There is a serious question of whether intentional systems theory so understood is true. Paralytics may provide a counterexample. But for a way that Dennett might respond to the case of paralytics, see McLaughlin and O’Leary-Hawthorne 1994.
Why Intentional Systems Theory Cannot Reconcile Physicalism
155
same point will hold. For if it is weaker than truth in every possible world, then from the fact that the totality of physical facts in conjunction with physicalism entails the world-wide pattern of distribution of global patterns of dispositions to peripheral behavior of systems, it will not follow that the totality of physical facts in conjunction with physicalism entails the worldwide distribution of beliefs and desires. So, intentional systems theory, understood as failing to imply the logical supervenience thesis, will not serve the purpose of allowing physicalists to vindicate belief-desire realism. Moreover, if intentional systems theory implies only the nomological supervenience thesis, then it is compatible with physicalism only if the supervenience laws linking subvening global patterns of peripheral behavioral dispositions to supervening belief-desire profiles are themselves entailed by the totality of physical facts and the assumption of physicalism. For if the supervenience laws were not so entailed, then the actual world would not be a minimal physical duplicate of itself: it would contain something other than what is required for it to be a physical duplicate of itself, namely the supervenience laws in question. And, so, physicalism would be false. To illustrate this point using our creation myth, were the supervenience laws fundamental, irreducible laws, then Mother Nature would not have succeeded in creating a duplicate simpliciter of the actual world once She had created a physical duplicate of it. For, having created the physical duplicate, She would then have to go on to dictate the laws linking global patterns of peripheral behavioral dispositions with belief-desire profiles. We see, then, that if the supervenience laws are fundamental, irreducible laws – emergent laws –, then intentional systems theory is incompatible with physicalism. Intentional systems theory would be a form of emergent behaviorism, and so inconsistent with physicalism.1 Moreover, if intentional systems theory implies only the nomological supervenience thesis, then whether it is a form of emergent behaviorism will turn on matters about which the theory is silent. It will turn on whether the supervenience laws linking global patterns of peripheral behavioral dispositions to belief-desire profiles are entailed by the conjunction of physicalism and the totality of physical facts. An intentional systems theorist may, of course, assert that there is such an entailment. But it would remain to explain why there is such an entailment, for intentional system theory would itself be silent on the matter. Whether the supervenience laws are so entailed will depend on whether the laws in which beliefs and desires figure are entailed by the totality of physical facts in conjunction with physicalism. If they are so 1
For discussions of emergence, see Broad 1925, McLaughlin 1992, and McLaughlin 1997.
156
Brian P. McLaughlin
entailed, intentional systems theory (understood to imply only the nomological supervenience thesis) would not explain why they are. An independent physicalism-friendly theory of belief and desire would be required to explain why. It is to that physicalism-friendly theory of belief and desire that physicalists would have to look to vindicate belief-desire realism, not to intentional systems theory. In conclusion, then, intentional systems theory faces a dilemma. Either it implies the logical supervenience thesis or it doesn’t. If it implies the logical supervenience thesis, then it is false. And, so, physicalists cannot successfully appeal to it to vindicate realism about beliefs and desires. On the other hand, if it doesn’t imply the logical supervenience thesis, but implies only a modally weaker supervenience thesis, then it is not the sort of theory that a physicalist could appeal to in order to vindicate realism about beliefs and desires. Thus, either way, intentional systems theory will not enable physicalists to reconcile physicalism with belief-desire realism. Physicalists who seek to do so must look elsewhere.
References Block, N: ’Psychologism and behaviorism’ Philosophical Review 90 (1981): 5-43. Broad, CD: The Mind and Its Place in Nature (London: Kegan Paul, 1925). Churchland, PM:’Eliminativism and the prepositional attitudes’ The Journal of Philosophy 78 (1978): 67-90. Dennett, D: Brainstorms: Philosophical Essays in Mind and Psychology (Montgomery, VT: Bradford Books, 1978). Dennett, D: The Intentional Stance (Cambridge, Mass.: MIT/Bradford Book, 1987). Dennett, D: ’The message is: there is no medium’ Philosophy and Phenomenological Research 53 (1993): 922. Dennett, D: ’Get real’ Philosophical Topics 22 (1994): 517-522. Fodor, J: The Language of Thought (Hassocks, Sussex: Harvester Press; Scranton, PA: Crowell, 1975). Hill, C: Sensations: A Defense of Type Materialism (Cambridge: Cambridge University Press, 1992). Jackson, F: ’Appendix A (for philosophers)’ Philosophy and Phenomenological Research 53 (1993): 901. Jackson, F: From Metaphysics to Ethics: A Defense of Conceptual Analysis (Oxford: Clarendon Press, 1998). Kripke, S: ’Indentity and necessity’ in: MK Munitz (ed.) Identity and Individuation (New York: New York University Press, 1971). Kripke, S: Naming and Necessity (Oxford: Basil Blackwell, 1980). Mackay, DM: ’Mentality and machines’ Aristotelian Society Supplement 26 (1952): 61-81. McLaughlin, BP: ’The rise and fall of British emergentism’ in: A. Beckermann, H. Flohr, and J. Kim (eds.) Emergence or Reduction? (Berlin: Walter de Gruyter, 1992), 49-93. McLaughlin, BP: ’Emergence and supervenience’ Intellectica 25 (1997): 25-43.
Why Intentional Systems Theory Cannot Reconcile Physicalism
157
McLaughlin, BP: ’Dispositions’ in E. Sosa and J. Kim (eds.) A Companion to Metaphysics (Oxford: Basil Blackwell, 1994), 70-73. McLaughlin, BP and O’Leary Hawthorne, J: ’Dennett’s logical behaviorism’ Philosophical Topics 22 (1994): 189-258. Peacocke, C: Sense and Content: Experience, Thought, and Their Relations (Oxford: Clarendon Press, 1983). Ryle, G: The Concept of Mind (London: Hutchinson, 1949). Stich, S: ’Eliminative materialism’ in: RA Wilson and FC Keil (eds.) The MIT Encyclopedia of the Cognitive Sciences (Cambridge, Mass.: MIT/Bradford, 1999), 265-267. Turing, AM: ’Computing machinery and intelligence’ Mind 59 (1950): 433-66.
158
Gerhard Preyer
GERHARD PREYER Primary Reasons: From Radical Interpretation to a Pure Anomalism of the Mental Abstract The paper gives a reconstruction of Donald Davidson’s theory of primary reasons in the context of the unified theory of meaning and action and its ontology of individual events. This is a necessary task to understand this philosophy of language and action because since his article “Actions, Reasons, and Causes” (1963) he has developed and modified his proposal on describing and explaining actions. He has expanded the “unified theory” to a composite theory of beliefs and desires as a total theory of behavior. At first, the anomalous monism was a byproduct. Yet, the step to this philosophy of the mental is not contingent. It will be shown that the anomalous monism as the primary feature of the mental does not lead us to a materialic monism in our ontology. If the evidence of radical interpretation has no echo in natural science, we have to give up materialism. Following this, I will argue for a pure anomalism. Or, in other words, for the ascription of propositional attitudes given by the semantical analysis of sentences of such attitudes does not commit us to a materialistc framework in our ontological thinking. Yet, to understand Davidson’s unified theory, we have to grasp the role of individual events in this framework since the ontological reduction of individual events explains par example the relation between the mind and the body, and the possibility of autonomous action in the world of causality. I do not argue for a dualism of properties between the physical of the mental in the manner of other philosophers, since from the radical interpreters point the empirical restrictions of the speaker’s attitudes and behavior cannot be fixed for all cases.
The unified theory of meaning and action is a total theory of behavior. Consequently, it is a composite theory of beliefs and desires. From the analysis of the logical form of action sentences in “The Logical Form of Action Sentences” follows a synkategorematic account of action and event sentences (Davidson (1967a): 1980). Therefore, the identity thesis of actions is reasoned by a logical argument. In the relation between the both acts A-A’, there is only one action which we can describe in different ways: AA’. The logical form of action sentences gives us a particular analysis of adverbial modifi-
Primary Reasons
159
cation. An adverbially modified action sentence such as “Peter went intentionally into the bathroom at midnight” has therefore the logical form: (}e) (went (Peter, e) & (into the bathroom) & (intentionally, e)), e is the event which verifies the sentence. Following this theory, most adverbial modifications like “slowly”, “intentionally”, “deliberately” etc. cannot change the reference to the expressions which they modify. They are of no importance to the semantical analysis of action sentences. An agent doing something intentionally cannot be considered as a feature of his acting, but as the description of his respective action. Adverbial and attributive adverbial modifications must therefore be analysed at the same level. Davidson’s critique of the logical connection argument by R.S. Peters, H.L.A. Hart, A.M. Honore, A.I. Melden, N. Malcolm, G.H. Wright and others is a consequence of the analysis of logical form of action sentences and singular causal statements. The result is the theory of primary reasons. Hence, it is necessary to give an answer to the following question: What is the logical relation between the description of actions and the propositional content of explained attitudes, provided that these attitudes stand in a logical relation to such descriptions, and, how can we explain actions with these attitudes? The answer to the question is given with the primary reasons and the syncategorematic account of intentions. We have found the logical relation between the description of an action and the propositional attitudes, provided that such attitudes explain the performance of an action. The primary reasons will give us therefore an answer to the “mysterious connection” (S. Hampshire) between reasons and actions. It is the claim of the ontology of individual events to understand actions, their explanation, causality, and also the relationship between the mental and the physical. The unified theory of meaning and action as a total theory of behavior has argued for a materialistic ontology which should be compatible with an anomalous monism: “The mental and the physical share ontology, but not ... classificatory concepts.” (Davidsons: 1990b, S. 18) The presupposed monism – the “shared ontology” – leads us to a substantial problem with this proposal in the philosophy of the mental. To make this clear, I will proceed with the following steps. Firstly, it will be shown how the explanation of action by primary reasons is connected with the syncategorematic account of intentions, and how we analyse the causal relations between the components of actions: pro attitudes, beliefs, and primitive actions. Primary reasons define intentional actions by their causes.
160
Gerhard Preyer
Do we have – with the causal theory of actions – a law of behavior based on analytical considerations in our hand from which we normally would assume that it is an empirical generalization? Yet, in this context, we are faced with the problem of the correct causation of causal chains. Secondly, the thesis of the anomalous monism of the mental is: there are no strict laws of mental events or psychophysical laws. This version of a monism detects the autonomy (freedom, self-rule) and anomaly (failure to fall under a law) of the mental by arguing that there is no systematic way to express a description of mental states in a physicalistic vocabulary. This argument is derived from Quine’s thesis of indeterminacy of translation. But it is assumed as well: every single mental event is a physiological event (tokenidentity). Yet, from my point of view, the evidence of RI advises us a pure anomalism of the mental – on an outline of Davidson’s philosophy: G. Preyer, M. Roth: 1998.
Pro Attitudes, Beliefs, Primitive Actions and their Causal Relations The distinction between intention, intentional actions and intending goes back to G.E.M. Anscombe. In “Actions, Reasons, and Causes” (1963) Davidson emphasizes that intentional actions mean acting with an intention. Therefore, the concept of intention is the basic concept. Action is performed with an intention if such doings are caused by primary reasons. The concept of intention as the basic concept is no longer used in the further development of this account. The question for the correct way causing the upshot of acting also emerges. What does mean a syncategorematic account of intentions and the explanation of actions by primary reasons? “The expression ‘the intention with which James went to church’ has the outward form of a description, but, in fact, it is syncategorematic and cannot be taken as a reference to an entity, state, disposition or event. Its function in this context is to generate new descriptions of actions in terms of their reasons; ...” (op. cit. 1980, p. 8). If this assertion was true, we could not intend any states, events, etc. “James” is the agent of an event, if, and only if, there is a description of what he has done which makes a sentence true, and that this sentence states that “James” did something intentionally. In accordance with this theory the statement “James did A intentionally” is equivalent to the assertion that an action is caused by a sort of primary reason. These reasons are given by analysis of the word “because” in sentences with which we explain actions: “If we interpret this ‘because’ as implying (among other things) a causal relation – and I believe we must – then in describing an action
Primary Reasons
161
as performed with a certain intention, we have described it as an action with a certain causal history. So in identifying the action with a physical history of the physical event includes events or states identical with the desires and cognitive states that yield a psychological explanation of the action.” (op. cit. 1980, p. 254, my italics) This citation is good evidence for the reason why the unification of the theory of meaning and action is based on an ontology of individual events. The claim that a participant does something intentionally implies an interpretation in the following way: (*) The act of doing something intentionally is caused by primary reasons, therefore, by a complex mental state. Such states are firstly composed of pro attitudes toward an action a with a particular property p, and, secondly, the belief that this action a has, under a given description, the desirable property p. The role of such beliefs can be explained by the distinction between primafacie-judgements – pro attitudes – and all-out judgements. The mental states themselves are no actions or events. “The only hope for the causal analysis is to find states or events which are causal conditions of intentional actions, but which are not themselves actions or events about which the question whether the agent can perform them can intelligibly be raised. The most eligible such states or events are the beliefs and desires of an agent that rationalize an action, in the sense that their propositional expression put the action in a favorable light, provide an account of the reasons the agent had in acting, and allow us to reconstruct the intention with which he acted.” (op. cit. 1980, p. 72.) Therefore it is valid: “To know a primary reason why someone acted as he did is to know an intention with which the action was done.” (op. cit. 1980, p. 7). Yet, knowing the intention does not mean that we know all primary reasons of a certain agent. Examples for pro attitudes are desires, economic preferences, aesthetic evaluations, social conventions, private and public goal orientations, etc. which reveal attitudes of an agent toward his actions. Therefore, these are original mental states or dispositions in the sense that the statement “James goes intentionally to church” is true just in the case, when his behavior is caused by the complex of pro attitudes and beliefs. Yet, the pro attitudes are not events or episodes or anything else of what ”James” intends. The statement “James did a intentionally” is therefore equivalent to the statement “James did a as a result of his primary reasons R”. James is an agent who caused an event e by the antecedent of a complex state. This state consists of a pro attitude toward the event e with a certain property and the belief that the event e, under a certain description d, has that property e. The pro attitude toward the event e and the belief cause directly such a complex state of the
162
Gerhard Preyer
agent: this state – which is not an event or an action – is a primary reason. Actions are physical upshots of a mental state and therefore merely body movements of an agent. If an agent moves his hand, he brings about an event e by his doing: he is an agent of this event. It is the aim to construct a class of events which are actions. “If we can say, as I am urging, that a person does, as agent whatever he does intentionally under some description, then, although the criterion of agency is, in the semantic sense, intentional, the expression of agency is itself purely extensional. The relation held between a person and an event, when the event is an action performed by the person, holds regardless of how the terms are described. Therefore, we can speak without confusion of the class of events that are actions which we cannot do with intentional actions.” (op. cit. 1980; p. 47) The analysis of the logical form of action sentences and the explanation of action by primary reasons apply to the principle of extensionality. This is a further reasoning for the ontology of events: actions are intentional body movements. These are events we distinguish from other events because they are caused by primary reasons. Actions are to be ascribed directly to agents. The theory of primary reasons is an objection to the intentionalists: 1. Knowing the intention does not mean to know the cause, and therefore the primary reasons in detail, such as if someone knew my intention about visiting the girl friend of my mother in hospital; one does not know whether I only do my duty, or connect it to buy a new suit in the shop near the hospital etc. 2. Descriptions could give us an better explanation for motives and intentions. The assumption of a class of intentional actions is wrong. The ascription of actions shows us a semantic opacity: “Hamlet intentionally kills the man behind the arras, but he does not intentionally kill Polonius. Yet, Polonius is the man behind the arras, and so Hamlet’s killing of the man behind the arras is identical with his killing of Polonius. It is a mistake to suppose that there is a class of intentional actions: if we took this tack, we should be compelled to say that one and the same action was both intentional and not intentional.” (op. cit. 1980, p. 46) This “opacity” makes the solution of ascribing intentions possible by analysis of “Someone is the agent of the event” by primary reasons. The rejection of a class of intentional action is concluded from the syncategorematic account of intention and the logical form analysis of action sentences. The agency is therefore more basic than the intention of the agent. 3. The logical connection argument gives us no analysis of the word “because” in sentences which explain action with given details about desires and beliefs. “If, as Melden claims, causal explanations are ‘wholly irrelevant to understand we seek’ of human action then we are without an analysis of the
Primary Reasons
163
‘because’ in ‘He did it because ...’, where we go on to name a reason.” (op. cit.1980, p. 11) The logical-connection argument states that the reasons which are given for an explanation of action cannot cause an action, because there is no logical or conceptual tie between causes and effect, and this is just to assume for explanations of action with reasons. Davidson’s objection is that there is no logical and conceptual relation between reasons and action as well as between causes and effects: from my preferences and beliefs the performance of any action can neither logically nor conceptually be concluded. This argument is relevant to the interpretation of practical syllogism, because such syllogisms are no helpful model for evaluative reasoning – see Davison op. cit. 1980, p. 15. The idea of a causal explanation of actions by primary reasons is this: we talk about such explanations without referring to statements of laws. We name laws only when we justify our statements. “In brief, laws are involved essentially in ordinary causal explanations, but not in rationalization.” (Davidson op. cit., p. 15). Yet, the explanation of primary reasons follows the logical-connection argument in two points: 1. we explain action starting from intelligible redescription, and 2. primary reasons are themselves no events or actions. In “Agency” (1971) it is distinguished between intention and intentional actions. However, it should be emphasized: not all intentional actions can be reduced to intentions. The agent causality (I. Thalberg) must be distinguished from the ordinary causality of events in the following way: ”What distinguishes agent causation from ordinary causation is that no expansion into a tale of two events is possible, and no law lurks. By the same token, nothing is explained. Therefore, it seems unreasonable using such expressions as ‘cause’, ‘bring about’, ‘make the case’ to illuminate the relation between an agent and his act.” (op. cit. 1980, p. 53). Yet: “Causality is central to the concept of agency, but it is ordinary causality between events that is relevant, and it concerns the effects and not the causes of actions (discounting, as before, the possibility of analyzing intention in terms of causality).” (Ibid.) It is the body movement as primitive actions whereby we ascribe causality. Actions are a physical upshot of a mental state and consequently mere body movements of the agent: “We must conclude, perhaps with a shock of surprise, that our primitive actions, the ones we do not do by doing something else, mere movements of the body – these are all the actions there are. We never do more than move our bodies: the rest is up to nature.” (op. cit. 1980, p. 59) In this sense we can conclude that we are as agents identical with our body. Therefore, the theory of primary reasons is also an objection to a causal interpretation of the relation between A-A’ (R. Chisholm, D. Bennett, A. Kenny, R. Taylor and others), because the intentional action and the body
164
Gerhard Preyer
movement are not causal distinct items. The ordinary event-causality can be useful to explain how the primitive action transfers itself to other described actions, but it does not explain the agent causation. In this sense, we have to classify some action as primitive, because they are not standing in a causal relation to any other actions of the agent. In this respect, it is not the case “... that if I do A by doing B my doing A and my doing B must be numerically distinct.”(op. cit. 1980, p. 57, note 16) If an event is an action, we classify it on the level of its description as primitive or intentionally. But we cannot “stretch” these primitive actions – with the help of the “accordion effect” (J.L. Austin, J. Feinberg) – and transfer actions to other actions because there are no further actions but only other descriptions of these doing. Following this account, intentions are no basic concept to understand agency. The logical form of action sentences has shown that the concept of agency is purely extensional. In a further step this logical argument is completed by analysis of the action causation: ”Doing something that causes a death is identical with causing a death. But there is no distinction to be made between causing the death of a person and killing him.” (op. cit. 1980, p. 58) The thesis of identity between A-A’ is reasoned with a logical argument. There is only one action A in the relation ARA’ which we can describe in different ways. That is the reason why the distinction between basic and non-basic action (A. Danto) is rendered invalid. ”But what is the relation between my pointing the gun and pulling the trigger, and my shooting the victim? The natural and, I think, correct answer is that the relation is that of identity.” ”Excuses provide endless examples of cases where we seem compelled to talk about ‘alternative descriptions of the same action’ seriously, i.e., literally. But there are plenty of other contexts in which the same need presses. Explaining an action by giving an intention with which it is done provides new descriptions of the action: I am writing my name on a piece of paper with the intention of writing a cheque, with the intention of paying my gambling debt. List all the different descriptions of my action. Here are a few for a start: I am writing my name. I am writing my name on a piece of paper. I am writing my name on a piece of paper with the intention of writing a cheque. I am writing the cheque. I am paying my gambling debt. It is hard to imagine how we can have a coherent theory of action unless we are allowed to say that each of these sentences is verified by the same action. Redescription may supply ( support ? ) the motive (‘I was getting my revenge’), place the action in the context of a rule (‘I am castling’), give the outcome (‘I killed him’), or provide evaluation (‘I did the right thing’).” (op. cit. 1980, p. 110) But the argument shows that each further redescription introduces another context-reference. There is also an alternative analysis of the act-pairs A-A’ to the identity thesis such as of A.I. Goldman
Primary Reasons
165
(1970) who has found a transitive relation between the pairs of acts and gives a reconstruction of the identity thesis with a level-generation of act-tokens. Davidson makes the assumption that we individuate events with the principle: ”... if events are identical, they consume identical stretches of time.” (op. cit. 1980, p. 177) Whereby the caused effect of an action is its end. All primitive acts are body movements and in connection with such acts a production of body movements. Actions described in causal items must be analyzed with such elementary activities and events. I think that this is a problematic point of Davidson’s proposal. How do we explain the paradox that the effect of an action will happen at a later time, such as the poisoning of water in the water pipe which would kill the inhabitants, even though their death would occur three month later? For Davidson this problem is not a dramatic one. If it took a long sequence of time for an effect such as the one between the poisoning of the water in the water pipe and the death of the inhabitants, the action of poisoning could also be unsuccessful, and other events could cause their death. Finding the cause of the end of an action is an empirical question; and the empirical knowledge of an agent makes his actions more or less paradox and fix the understanding of a statement like “You have killed the inhabitants” – see op. cit. 1980, p. 177-78. In “Intending” (1978) “pure intending” is, as opposed to intentions and intentional actions, the foundation for our explaining of action. “To intend performing an action is, on my account, holding that it is desirable to perform an action of a certain sort in the light of what one believes is and will be the case. But, if one believes no such action is possible, there can be no judgement that such an action consistent with one’s belief is desirable. There can be no such intention.” (op. cit. 1980, p. 100-01). But forming an intention is not saying anything. Therefore, a performative theory of intending is not argued for. “Forming an intention may be an action but it is not a performance and having an intention is not generally the aftermath of one.” (op. cit. 1980, p. 90). What – then – is the relation between intending and an intentional action? To answer this question, Davidson introduces – in contrast to prima-faciejudgements – so-called all-out or unconditional judgements like “this action is desirable”. Prima-facie-judgement are “judgements that actions are desirable in so far as they have a certain attribute.” (op. cit. 1980, p. 98). Such judgements can not be connected directly to actions, because: “We can hardly expect to learn whether an action ought to be performed simply from the fact that it is both prima facie right and prima facie wrong.” (op. cit. 1980, p. 37) Therefore, we have to introduce an all-out judgement by which we decide on the desirable property of an action as a sufficient reason to do something or not. “For a judgement that something I think I can do, and that I think I see
166
Gerhard Preyer
my way clear to doing, a judgement that such an action is desirable not only for one or another reason but in the light of all my reasons; a judgement like this is not a mere wish. It is an intention. (This is not to deny that there are borderline cases.).” (op. cit. 1980, p. 101). Acting with an intention means nothing more than: an intention is an all-out judgement; and we distinguish intentional actions by such judgements. Wishes, desires, principles, felt duties etc. are reasons and also intentions for actions we express by prima-faciejudgement. The relation between an intention and an intentional action can not be connected with prima-facie-judgements, it can only be connected with all-out judgements. With this proposal Davidson has modified his concept of primary reasons in “Action, Reasons, and Causes” (1963). These refer to the following issues: 1. The redescription of actions by primary reasons which states that the action was done intentionally does not allow the conclusion that the action is simply desirable to perform. We can only deduce that the action has such a property. Therefore, every explanation of action cannot elaborate all primafacie-reason from which it is concluded that a certain action may be desirable. 2. We may equate an action with a conclusion of practical inferences, and this means, that an action is born by an all-out judgement. This causes us to do the specific action. In the first proposal of primary reasons in 1963, it was not clear how the intending and intentional actions can be connected – see on this modification op. cit. 1980, p. 98-99. 3. If we express the way how beliefs and desires cause an action in a practical syllogism, we must never exclude that this action could be caused by another way than assumed. This leads us to the problem of the right causal chains and the lunatic internal causal chains. I will discuss this later. Therefore, it is not necessary to elaborate in particular how attitudes cause actions which we redescribe by primary reasons (op. cit. 1980, p. 79). For this account the difference between intended or not-intended effects of action is only concerned with the description of the upshots, like, in the case of torpedoing the Bismarck instead of the Tirpitz, it refers to a description of the result of an action as not being intentional. The analysis of intending is coherent with acting with an intention – the unconditional beliefs are equal to an intention – but the prior intention is not the basic concept. It has to be emphasized that for the theory of belief and meaning, the concept of reason is of prime importance. If we explained actions by beliefs – but also desires –, we would make the assumption that the agent we ascribe such beliefs to has also a reason to fulfil his beliefs. Such beliefs are not restricted to an individual agent. Primary reasons which consist of pro attitudes
Primary Reasons
167
and unconditional beliefs are causes of actions. But a causal explanation of reasons must also satisfy the demand for rationality because the description of action has to clarify the rationality of the content of beliefs. This is the difference of explanations given by natural science. Intentional actions are performed by reasons and the agent has an answer to the question: Why have you done that? Three different sorts of reasons must be distinguished: cognitive reasons such as beliefs, conative reasons such as values, goals and orientations, and artificial reasons like competencies and professional experiences. Causation of an action, in connection with our intentions and pro or con attitudes, is the decisive factor of beliefs. On the role of beliefs see Davidson: 1992, p. 1. Yet, it has to be emphasized that there is no absolute principle of evaluating our reasons: “Principles, or reasons for acting, are irreducibly multiple.” (op. cit. 1980, p. 34) – on the plurality of good reasons see J. NidaRümelin: 1994. Therefore, there is no absolute reasoning. The redescription of actions and their explanations by primary reasons leads us back to RI and the principle of charity – on the connection of the theory of interpretation, propositional attitudes and the causal explanation by primary reasons see J. Heil: 1996. The result of Davidson’s theory of action is: in order to understand an action with an intention, we require pro attitudes, beliefs and the primitive action itself. The analysis of the causal relations of these components has to be added to it. Davidson has argued that if we have an ontological reduction at our disposal – the ontic decision of individual events – it “is enough to answer many puzzles about the relation between the mind and the body, and to explain the possibility of autonomous action in the world of causality.” (op. cit.1980, p. 88). From my point of view, the problem of this account is that the concept of primitive actions is reasoned by the ontological commitments of individual event. By analytical considerations the causal theory of intentional actions should give us a law of behavior, a “law” – as we would normally assume it – is a result of an empirical generalization – op. cit. 1980, p. 76; on the exposition p. 76-81. “These things, then, stand together: a law stating conditions under which agents perform intentional actions; an analysis of freedom to act that makes it a causal power; and a causal analysis of intentional action. They stand together, or they fall together. In my opinion they coincide if we want explicit, non-question begging analyses, or laws without generous caveats and ceteris paribus clauses.” (op. cit. 1980, p. 76) Yet, this account is a problematic one also for Davidson, because with it we cannot give a solution to the problem of ascribing attitudes. This is really a matter of explanatory role of such
168
Gerhard Preyer
laws of behavior in their relation to the explanation by primary reasons. In my view this account is falsified by the evidence of RI: 1. The central problem of this account is: How does giving the causal condition of intentional action provide us with an empirical law of prognosticate behavior? This law of causality would not mention any concepts of causality. The use of causal concepts is a “cloak for ignorance”. In ordinary life we normally use rules of thumb for the prediction of events and the explanation of action because we do not know a sufficient and exact law. If we had such laws, we would often miss the knowledge of the particular events from whose descriptions we conclude such laws. If we grasped the antecedent’s condition of the attribution of propositional attitudes with physical and behavioral concepts, the empirical character of this law would be clear. Yet, this would not be successful by the anomalous monism of the mental, a radical externalism of the individuation of propositional attitudes and the evidence of RI (radical interpretation) holding a sentence true. 2. The ascription of propositional attitudes by oblique or direct considerations of empirical laws, the way we take laws into account by explained reasons, has to its explanandum only individual agents. Such laws do not improve our explanations by primary reasons. The explanation of actions also requires a general knowledge of the agent, because a single action by itself is not evidence enough whether it is right to ascribe any attitude. Therefore, a general knowledge is informative to our explanations of actions. It tells us something about permanent preferences and beliefs, but we do not apply it in explaining primary reasons (op. cit.1980, p. 273-74). We may argue that such general knowledge are assumptions to make our behavior comprehensible, but we cannot conclude anything about the primary reasons, in particular why, par example, “James” gives a promise to his colleague; may be he wants to curry favor with him. 3. There is, however, an open question: What gives an explanation by primary reasons its empirical thrust? For Davidson, the explanatory power of primary reasons is due to the ascription of attitudes and preferences and is not given from the axioms of decision theory or from rationality as a human trait like Hempel has argued – see op. cit. 1980, pp. 266-67, 273-74; on rational pattern of preferences op. cit. 1980, p. 268-69. Yet, if we connected general laws of behavior with the explanation by primary reasons, Davidson’s answer would lead us to the problem of making behavior intelligible in principle. There is a conflict between the two points of view: grasping the action from the inside and its description from the outside.
Primary Reasons
169
Every description of action from the outside as a happening seems to ignore the performance of actions – T. Nagel: 1986, on Nagel see J. Hornsby: 1993. This problem leads us to the role of decision from which our doings are born and also to our responsibility for actions. Yet, the perspective from the inside is taken into account by RI because every ascription of attitudes has to ensue from the belief system of the agent under study. If this is right, all mental events (states), or the network of propositional attitudes are to be explained in a frame of reference which has no echo in physical laws. “Such accounts of intentional behavior (the explanation by primary reasons) operate in a conceptual framework removed from the direct reach of a physical law by describing both cause and effect, reason and action, as aspects of a portrait of a human agent.” (op. cit. 1980, p. 225) Natural science can not explain such states as a class. It may be that we have such explanations for single mental events insofar as we know the particular identities (ibid.). Just this is clarified by the anomalous monism of the mental. Yet, how do we answer the question of the right way of causation, insofar as we explain actions by primary reasons? For the explanation by primary reasons the “because” in “... he did it because ...” leads us to the concept of causality. In principle, it can not be excluded that the causal chains show an anomaly. An agent may have pro or con attitudes and beliefs which rationalize and cause his action without us stating the described action as intentional. Such cases are anomalous extern causal chains like in the example of the man who wants to kill someone by shooting him, misses the victim, but the shot stampedes a herd of wild pigs which trample him to death – the example goes back to D. Bennett: 1965. Another, with this example connected, case are lunatic internal causal chains like the well known example of the climber (op. cit. 1980, p. 78-79). The example is: “A climber might want to rid himself off the weight and danger of holding another man on a rope, and he might know that by loosening his hold on the rope, he could rid himself of the weight and danger. This way, this belief and wish might unnerve him to cause him to loosen his hold, and yet, it might be the case that he never chose to loosen his hold, nor did he do it intentionally. I think, it will not help to add that the belief and the wish must be combined to cause him wanting to loosen his hold, for there will remain the two questions of how the belief and the wish caused the second wish, and how wanting to loosen his hold caused him to loosen his hold.” (op. cit. 1980, p. 79) Lunatic internal causal chains such as the example of the climber show us that the desired effects are not only caused by the intended effect by primary reasons. In the example of the herd of wild pigs which trample the victim to death, we may ask ourselves whether marksman has killed the victim inten-
170
Gerhard Preyer
tionally. Therefore, we are asking for the right causal chain. The lunatic internal causal chains lead us to another problem. It shows that if the primary reasons have caused the actions in the right way, such actions could also be caused in a different way. Therefore, Davidson concludes that we could not extract in particular cases how attitudes factually have caused action in order to redescribe actions as rationalized by primary reasons. This leads us to the following question: how do we grasp the relationship between intentional and not-intentional effects of action? The identity thesis states that there is only one action in the relation between A-A’ which could have different effects and consequences. These are no parts of action but only one we redescribe in different ways. “The analysis of the concept of an action comes, then, if I am right, in two stages, the first bringing in intention directly, and the second extending the concept to action in the first sense, redescribed in terms of unintended consequences or other unintended characteristics.” (op. cit. 1980, p. 71) Yet, how far can the expression “x makes happen y” in general be interpreted as “x is the agent of the event” in the sense of Davidson’s syncategorematic account of intentions? Davidson’s theory of action argues: a person who moves his hand brings about a movement: he is the agent of the event. Hence, someone is the agent of an event, if there is a description of his doing which makes the sentence “true” saying that it was an intentional action. The statement “James brings about something” has two different readings: 1. “James” does something a which causes c or 2. “James” is the agent in the sense of Davidson. “x brings about c” usually means: x does something a which brings about c such as he pours gas onto the ground, lights it with the result: (that) there is a fire. The second interpretation refers to Davidson’s theory, because, in the case if someone did something which caused the fire – in correspondence with the accordion-effect (J.L. Austin, J. Feinberg: 1968) – we would describe the doing as an act of making a fire. But the normal use of the expression does not mean: x is the agent in the sense of Davidson, because with the expression we ascribe intentions. For Davidson this is not the case. Intentions are by no means sufficient conditions of any ascription of action, because g, the action x par example, the spilling of the coffee, may be the case if non-p par example “James” does not have the intention of doing it. It is certainly not absurd to explicate the expression “x brings about c” in the sense of Davidson. In this case explication means finding a technical equiv-
Primary Reasons
171
alence for a vernacular talk. If we interpreted the expression this way, we could ask the question to this account: how do we ascribe propositional attitudes with these expressions? The ascription of attitudes has to be handled on this basis: normally we use the expression “x brings about c”, in the sense of doing c, to ascribe intentions. If a speaker uttered in a given situation s “James did x with the intention to do a”, we interpreted the expression in such a way that the speaker s refers by using his words to the same state which conventionally corresponds to “I will (want) do a”. But “doing something wilful” is only then the case, when I do it intentionally. We interpret the utterance “I will (want to) do something, now” the word “will (want to)” as an expression of an attitude. If we make the assumption that the series of acts overlap (Anscombe) –I think it is a misleading assumption - the answer to the question of the identity of actions presupposes that all parts of actions refer to the performed action, and all tells about the same entity. The question: (1) why does James give a piano concerto? must be distinguished from the question: (2) why does James play the piano? The expressions (1) and (2) may refer to the same event but (1) gives us no answer to the intention of playing the piano, the answer could be, par example, ”James was invited to do it”. Only then, provided that we know the intention, we could answer (2), par example, James plays piano because he wants to be a little better. We do not explain the real happening of events, but why a certain event happens or not. This is a further reasoning for the notion that the semantical analysis of action verbs can not provide for instances of designated expressions. In the expression “I will/want (shall) now do something” we interpret the word “will/want (shall)” as an expression of an attitude. Anscombe: 1959 has written: “The primitive sign of wanting is trying to get.” For Davidson Hampshire’s: 1965 it becomes clearer when he argues “‘A wants to do X’ is equivalent to ‘other things being equal, he would do X, if he could’. Here I take (possibly contrary to Hampshire’s intent) ‘other things being equal’ to mean, or anyway to allow, the interpretation ‘provided that there is not anything he wants more’.” (op. cit. 1980, p. 23-24) This interpretation of “I will/want (shall) now do something” connects intentions with a desire, resp., a motivation with an unconditional judgement: pure intending is such a judgement. This construction goes back to a syncategorematic account of event and action sentences and the truth-centered theory of interpretation to ascribe attitudes. Davidson concedes that “to do something is wanting to do it” but intending is
172
Gerhard Preyer
not wanting. Wishes and desires “are best viewed as corresponding to, or constituting, prima facie judgements.” (op. cit. 1980, p. 102). We make the given behavior intelligible by explaining it with the assumption that the agent realizes his strongest beliefs, such as, why an agent wants to do x more then he wants to do y, and this causes his intentional actions. This is a judgement that ”consists with one’s beliefs is desirable.” (Ibid.) Such judgements are no wishes but intentions. The syncategorematic account of intentions does not argue from the point of view that mental concepts do not refer to events, states or dispositions. In the case of the ascription of action on the basis of events which verifies a description of an action like “James spills the coffee intentionally”, we quantify over linguistic entities. The use of sentences about propositional attitudes and descriptions of action requires in this respect a semantical analysis of such sentences. We cannot exclude cases such as a person having an intention or a disposition, but not acting in this way, e.g. he intends to sink the Tirpitz, but torpedoes the Bismarck, or he has a disposition to drink, which he ignores. The singular term also can denote an intention as a ‘good’ one which becomes true, thus we find a way to analyze intentions with a non-Davidson account. The word “intending” stands in a relation with “choice” and “decide” (Austin). Both refer to a mental activity which implies a deliberation. The analysis of the word “because” of action explaining sentences requires us then to give a further answer to the relation between an intention and an intentional action – on a first version G. Preyer: 1984. The analysis of intending by all-out judgements has to be developed in that direction that the realization of an intention will happen only by our decision to do something. In this context the prior intention is significant, because intending is forming a decision by our strongest beliefs as exclusionary reasons. Yet, this is relevant to the conclusion about the problem of lunatic internal causal chains, that we cannot rationalize in particular how attitudes factually have caused any action. For Davidson, however, we could also come to have an all-out judgement without such modes as deciding, choosing, and deliberating to arrive at it (op.cit. 1980, p. 99).
A Pure Anomalism of the Mental For Davidson actions are intentional body movements that we describe as intentionally produced physical events, and we explain such events which we call actions by primary reasons.
Primary Reasons
173
How far is a unified theory of meaning and action of anomalous monism reasoned by the truth-centered theory of RI, and to what extent does this lead us to a pure anomalism of the mental against an ontological monism of the physical and the mental? RI has a mental counterpart which is the network of propositional attitudes. For RI it is required that an interpreter makes the assumption that this network is extensively coherent and consistent. Therefore, the speaker has a logic at his disposal. This is a necessary condition for every interpretation and ascription of attitudes. Irrationality is an inner inconsistency as a deviation from a belief from a set of otherwise consistent beliefs of the speaker. The application of RI is not only the ascription of beliefs and actions, but also of intentions, desires and passions. Also passions have a propositional content such as pride is pride to something which causes the passion. Causality and rationality have to go hand in hand in interpreting behavior. But the attribution of the mental and the ascriptions of action lead us to the other hand back of RI: ”It is a feature of the mental that the attribution of mental phenomena must be responsible to the background of reasons, beliefs, and intentions of the individual. There cannot be tight connections between the realms if each is to retain allegiance to its proper source of evidence.” (op. cit. 1980, p. 222, my italics) For an anomalous monism and the assumption of supervenience of the mental and the physical we should be committed to physicalistic ontology, but not to a conceptual reduction of the mental to the physical. There are no psycho-physical laws which could motivate us by reasons of ontological economizing to such a reduction. Since the mid-eighties Davidson’s account has initiated an extensive debate. J. Kim has argued, par example, that a nonreductive materialism is a myth – J. Kim: 1993, S. 265-284, on Davidson S. 267-271, 275-277; on the debate between Kim and Davidson, on the anomalous monism in general J. Heil, A. Mele (eds.): 1993, on a defence of the anomalism N. Melchert: 1986, on Kim’s concept of supervenience causality, the problem of mental causality and the concept of causality in particular an analysis with contrafactual conditionals see E. Lepore and B. Loewer (1987), (1989). In a first step I will recall what an anomalous monism means. It is helpful to recognize why we will go to a pure anomalism: 1. Between mental events and physical events there are causal relations. For this assumption an ontology of particular, unrepeatable and dated events is presupposed. Therefore, there is an identity of tokens between the mental and the physical, but not an identity of type. For all events it is valid that their descriptions cannot change the causal power of such events. Causality and
174
Gerhard Preyer
identity are relations between individual events. But “The principle of the anomalism of the mental concerns events described as mental, for events are mental only as described.” (op. cit. 1980, p. 215). 2. By finding laws, the explanation, not the stating, of singular causal relations can be given. There is a difference between singular causal statements, between events und causal explanations which refer to statements. Covering by law means: a law is expressed in a language which entails a statement (sentence) about the corresponding (described) event and a statement (sentence) about the happening of the corresponding effect. A generalization is not only lawlike if it is true, it also has to be deterministic, and its validity has to state without ceteris paribus restrictions. Therefore, the universe of a closed system has to be presupposed. Such explanations are complete physical explanations of a “finshed physics.” It is an open question whether this model of explanation is adequate for natural science in general – see L.R. Baker: 1993. Yet, this is not my subject in this context. Generalizations which do not fulfil this condition are so-called non-strict laws. In accordance with his account in 1970, 1967b, for Davidson: 1993 the distinction between statements on simple causal relations between events and their explanation is the essential point. The confusion between the both emerges from our everyday understanding of event. Often we do not distinguish between causal relations between particulars such as the short-circuit in the electric cables which sets a house on fire and types of events which are lawlike (Davidson: 1993, S. 15-16). The latter have to be sufficient. Therefore, the proposal given in “Mental Events” (1970) is valid: not every true singular causal statement is covered by law (ibid. p. 213). 3. The mental counterpart of RI is the network of propositional attitudes. The holism of this network is first evidence for the autonomy and the anomaly of the mental. It follows from the truth-centered and holistic theoretical framework of RI, that the concepts with which we describe mental events must in principle be distinguished from such concepts we us to describe physical events. For Davidson: 1993, p. 11 this argument is not reasoned by the objection of psycho-physical laws, because: if the mental and the physical are connected by causal relation, a shared ontology between both is given. The extent to which we make a reduction is the degree of anomaly. For Davidson supervenience means that a change of mental properties is caused along the correspondent change in physical properties, but this does not imply that the same physical properties change with the same mental properties: ”Supervenience of any form implies monism, but it does not imply either definitional or nomological reduction.” (Ibid., p. 5). Therefore, both following accounts must be rejected: firstly, a reduction of the mental to the physical
Primary Reasons
175
with an explicit definition and, secondly, a reduction by bridge laws, whereby we connect both properties. “But if mental concepts are not reducible to physical concepts, there is no reason to suppose we would lose interest in explanations in mental terms just because we had a complete physical explanation. What is true, of course, is that psychological explanations are never full and sufficient; like most explanations, they are interest-sensitive, and simply assume that a vast number of (unspecified and unspecifiable) factors that might have intervened between cause and effect did not.” (Davidson: 1993, p. 16, my italics) The ontological monism should be coherent with a conceptual dualism. 4. The anomalous monism of the mental is not only reasoned by RI but also by a radical externalism. Without the distal individuation of the content of propositional attitudes in triangulation between speaker, interpreter and the external world itself it has no echo in neuronal processes. For such individuation we presuppose a shared picture of the world. Therefore, the authority of the first person, the social character of language and the distal individuation of thoughts and meaning belong together. To a large degree we have accepted that the essential feature of the mental – refraining from unconscious motives – is its intentionality (Brentano). As mental states we normally distiunguish sensations like pain, visual awareness, thoughts and propositional attitudes. Such attitudes have a propositional content which we grasp semantically. The feature of all mental states is the authority of the first person. From my point of view, the problem of anomalous monism is in principle the acceptance of an identity of the token and claims to analyse the relation between the mental and the physical with the paradigm of strict or non-strict laws. So long as we do not give up an identity of the token, the philosophy of the mental poses us problems we cannot solve, and it is also impossible to form an understanding of our self-knowledge. This is not a matter of the consistency of anomalous monism but a substantial issue. Therefore, I will give up the idea of the identity of the token between the mental and the physical and I will argue for a pure anomolism. This follows from the principle of charity and that the ascription of attitudes is dependent on our theory of interpretation. Such anomalism is reasoned by the evidence of RI. To summarize, problems with the philosophy of the mental such as the problem of epiphenomalism, the different versions of supervenience or the problem causal-explanatory exclusion are a result of handling the metaphysical framework of materialism. A pure anomalism is reasoned by the following four arguments: 1. The mental is holistic in nature because propositional attitudes are not single attitudes. The relationship between attitudes is of a logical matter. The
176
Gerhard Preyer
holism of the mental has no echo in the network of brain processes. Yet, I do not argue for an ontological dualism of properties between the mental and the physical, because from RI is concluded that the restrictions of the ascriptions of attitudes are not fixed in the end. 2. We can only ascribe propositional attitudes to creatures who have rationality at their disposal. This follows by the application of the principle of charity or, respectively, by the principle of natural epistemic justice, how I call it. This means for RI: only creatures who are able to communicate meet the concept of an objective world. It is a dimension of the triangle. Beliefs and other attitudes can only be ascribed to by the procedure of RI. Such ascriptions have to cope with the difficulties of the circle between belief and meaning. This is only then possible, when a radical interpreter makes the assumption that most beliefs of the speaker are true, and we presuppose the triangulation model of RI similarly to reactions caused from the external world. This is presupposed on the interpreter’s side. Therefore, we conclude that attitudes cannot be ascribed to by neuroscience whatever the development of such a science might be. The principle of charity has not only no echo in natural science, but its knowledge is also not instructive for the ascription of attitudes. This is reasoned by the evidence of RI such as the background of reasons, beliefs and intentions. We conclude from that: if we are required to make the assumption of rationality to make the mental intelligible, we assume the existence of both domains of different synthetic constitutive principles: “Hence, if each domain is to retain its own integrity, there cannot be laws connecting them. ... I have always thought that the power of the Davidson argument for mental anomalism is seen in fact that, if it works at all, it should work against laws of all kind ...” (J. Kim: 1993, S. 25, see also on this problem J. Hornsby: 1993). This is a correct conclusion from Davidson’s philosophy of the mental, because we are not forced by RI to make the anomalism coherent with an ontological monism between the mental and the physical: ”The heteronomic character of general statements linking the mental and the physical traces back to this central role of translation in the description of all propositional attitudes, and to the indeterminacy of translation. There are no strict psychophysical laws because of the disparate commitments of the mental and physical schemes.” (op. cit.1980, p. 222). These “commitments” characterize the primary feature of the mental. 3. The ascription of action is a result of the causal explanations of primary reasons, and mental concepts have a causal origin. The mental does not “constitute a closed system” (op. cit. 1980, p. 239) It is connected with the external world and not excluded from the outside. Mental phenomena are not an
Primary Reasons
177
instance of explanations with strict and deterministic laws – on RI and the model of triangulation G. Preyer: 1998, R. Wilburn: 1998, on content externalism A. Bruecker: 1998 4. Davidsons monism is a weak monism. Therefore, it seems reasonable to go the step to a pure anomalism which is reasoned by the evidence of RI and the triangulation. The anomalous monism is no eliminativism, because the ascription of mental states is of an attributive matter. Such ascriptions are not rendered invalid by any scientific progress. The ontology of individual events is supported by a monism which means: there is no psychological phenomenon without the physical. But anomalism of the mental is also no dualism of properties of the mental and the physical, because from the radical interpreter’s point the empirical restrictions of the speaker’s attitudes and behavior cannot be fixed for all cases. Because what an addressee understands about the words uttered by any speaker is not only determined by the meaning of the words, but also by the speaker’s beliefs about the world. In this respect, all understanding is underdeterminated by the network of propositional attitudes. Having this in mind, we give up materialism and argue in favor for a pure anomalism. How we do unify the theory of truth with the theory of meaning with respect to the ascriptions of actions? The answer to this question gives us a further clarification of the pure anomalous: because the post-empirical theory of meaning has applied understanding meaning to utterances, the intelligible redescription and explanation of actions is connected. The ascription of meaning to an utterance stands always in relation to empirical theories on the respective linguistic behavior. This is a rejection of all analytical proposals of meaning. Meaning systematically depends on sentence meaning, and such meaning is inseparable from the speaker’s propositional attitudes. The theory of meaning has to be expanded to intelligible redescription and explanation of actions. Therefore, with propositional attitudes we redescribe reasons for actions and explain such acting by unconditional judgements. The principle of charity and its unrestricted application is the presupposition of all interpretation, therefore, it is a guarantee for an independence of evidence for an ascription of attitudes. The truth-centered theory of interpretation is continued in the theory of action. The unification of the theory of meaning and the theory of action is carried out in the framework of the theory of truth. We explain a given action by conclusions of a presupposed theory of interpretation. The ascription of attitudes is successful, if the agent satisfies the ascribed attitudes concluded from our the truth-centered theory. Therefore, the theory of interpretation has to be completed by a total theory of explanation of behavior which enables us
178
Gerhard Preyer
to ascribe beliefs and desires. The theory of interpretation, decision and communication does not stand alone. The connection between RI and the explanation of actions by primary reasons is often misjudged – on the framework in general, Davidson: 1984, p. 160-162. In contrast to that, what we should have in mind: we explain actions by finding the attitudes of the speaker and order them into a rational pattern of behavior. The redescription of linguistic behavior does not explain why a speaker performs a certain utterance or action. RI states the meaning of any utterance, and it is not clear what we mean in explaining it also by RI. But if we understand the utterance as an action, the interpreter should be able to attribute beliefs and desires of the agent We have emphasized that the concept of reason (belief) plays a crucial part in RI. This concept of reason has to be applied to our attitudes. Attitudes which explain our actions are rational when we make the assumption that our actions are caused by beliefs which are relevant to them and realize the pro attitudes. This exactly leads us to the concept of unconditioned beliefs as sufficient reasons for actions. From RI it follows that a total theory of behavior qualifies us not only to interpret uttered sentences but also to ascribe beliefs and desires. This composite theory can explain behavior in general. Therefore, the primary reasons lead us back to RI and to coping with difficulties of the tension between belief and meaning. Considering holding a sentence true is only then an independent evidence for RI, when we are successful in ascribing and explaining actions which are connected with our intentions by a composite theory. Therefore, the theory of interpretation has its place “within a more comprehensive theory of action and thought” (op. cit. 1984, p. 162) with its own constitutive principles. If this was not the case, “the notion of a true belief” could not be connected with the “notion of a true utterance” and “this in turn there cannot be without shared language.” (op. cit. 1984, p. 170)
References Anscombe, G.E.M., “Thought and Action in Aristoteles“, in New Essays on Plato and Aristotle, ed. R. Bambrough (London: Routhledge and P. Kegan 1995). Brueckner, A., „Content Externalism and A Priori Knowledge“, Protosociology Vol. 11 1998: Cognitive Semantics II - Externalism in Debate. Davidson, D., “Thinking Causes”, in Mental Causation, eds. J. Heil, A. Mele (Oxford: Clarendon Press, 1993). ——— “Spinoza’s Causal Theory of the Affects”, Berkley, Manuskript 1992. ——— “Representation and Interpretation”, in Modelling the Mind, K.A. M. Said, W.H. NewtonSmith, R. Viale, K.V. Wiles eds. ( Oxford 1990).
Primary Reasons
179
——— “Intending” (1978), Actions and Events (Oxford: Clarendon Press, 1980). ——— “Hempel on Explaining Action” (1976), op. cit. 1980. ——— “Thought and Talk” (1975), Truth and Interpretation (Oxford: Clarendon Press, 1894). ——— “Freedom to Act” (1973), op. cit. 1980. ———“Agency” (1971), op. cit. 1980. ———“How is Weakness of the Will Possible” (1970), op. cit. 1980. ——— “The Individuation of Events” (1969), op. cit. 1980. ——— “Action, Reasons, and Causes” (1963), op. cit. 1980. ——— “The Logical Form of Action Sentences” (1967a), op. cit. 1980 ——— “Causal Relations” (1967b), op. cit. 1980. Feinberg, J., “Action and Responsibility” (1965), in The Philosophy of Action, ed. A.R. White, (Oxford: Oxford University Press 1968). Goldman, A.I., Theory of Human Action, (Englewood Cliffs, NJ.: Printice Hall, 1970). Hampshire, S., Freedom of the Individuals, (Princeton NJ: Princton University Press 1965). Heil, J. (1996): “The Propositional Attitudes”, in: Protosociology, Vol. 8/9 1996: Rationality II/III. Hornsby, J., “Agency and Causal Explanation” in eds. J. Heil, A. Mele op.cit. 1993. Kim, J., “Can Supervenience and ‚Non-Strict Law’ Save Anomalous Monism?”, in: Mental Causation, eds. J. Heil, A. Mele (Oxford: Clarendon Press, 1993). Lepore, E., Loewer, B., “More on Mind Matter”, in: Philosophical Topics Vol. XVII, No. 1 1989. ———, Loewer, B., “Mind Matters”, in: Journal of Philosophy Inc. 1987. Melchert, N. (1986): “What’s wrong with Anomalous Monism?”, in Journal of Philosophy Inc. Nagel, T., Die Grenzen der Objektivität. Philosophische Vorlesungen, (Stuttgart: Reclam 1991). Nida-Rümelin, J., “Die Vielheit guter Gründe und die Theorie praktischer Rationalität”, in: Protosociology vol. 6: Rationality I, 1994. Preyer, G., M. Roth, “On Donald Davidson’s Philosophy: An Outline”, Protosociology, Vol. 11 1998: Cognitive Semantics II - Externalism in Debate, Frankfurt am Main. Preyer, G., “Interpretation and Rationality: Steps from Radical Interpretation to the Externalism of Triangulation”, Protosociology, Vol. 11 1998: Cognitive Semantics II - Externalism in Debate, Frankfurt am Main. ———, “Rationalität: Absichten - Primärgründe - praktisches Denken”, in Language, Mind and Epistemology On Donald Davidson’s Philosophy, G. Preyer et.al., (Dordrecht: Kluwer Synthese Library 1994). Wilburn, R., “Knowledge, Content, and the Wellspring of Objectivity”, in: Protosociology, vol. 11 1998: Cognitive Semantics II - Externalism in Debate.
180
Rebecca Kukla
REBECCA KUKLA How to Get an Interpretivist Committed1 Abstract I argue that interpretivists ought to broaden and enrich the constitutive standards of interpretability and epistemic agency that they have inherited from classic Davidsonian theory. Drawing heavily upon John Haugeland’s recent account of objective truth- telling, I claim that in order to be an interpretable epistemic agent at all, a being must have various kinds of practical (yet genuinely epistemic) commitments that cannot be reduced to combinations of beliefs and desires. On the basis of this claim, I argue that radical interpreters must appeal to many commitments held by their interpretees other than assents to observation sentences and commitments to sincerity; hence the interpretive tools available in the Davidsonian toolbox are insufficient. I suggest that we ought to take the behaviors manifesting the various commitments that constitute epistemic agency as straightforwardly available from a third-personal observational perspective, and thus as no threat to the basic spirit of interpretivism. At the same time, I claim that these behaviors cannot be individuated in non-normative, physicalist terms, so my account should indeed pose a threat to naturalists of a certain stripe. I end by revisiting and moderately revising Davidson's notorious deflation of the problem of radical skepticism.
For several decades, Davidson has promoted his influential brand of interpretivism, which consists of at least two primary, interdependent theses: The first is that the meaning of an utterance, thought, or action is holistically determined by the best possible interpretation of the meaning-maker, where the best interpretation is the one that optimizes the attribution of true, rationally connected beliefs and truth-claims to the interpretee. The second is that the constitutive criterion for being an epistemic agent – a believer and meaningful claim-maker – is being interpretable as believing and saying mostly true, rationally connected things of the right sort. From this second thesis, 1
This paper has benefited from conversations with many people, but overwhelming debts must be acknowledged in particualr to John Haugeland, for serving as the ongoing source of philosophical inspiration for it, to Ernest LePore, for creating the possibility for me to write it and to Richard Manning, for giving me detailed, enormously helpful comments on my penultime draft, and for productivly fighting with me about everything I ever say about Davidson, every step of the way.
How to Get an Interpretivist Committed
181
Davidson has argued, a third central thesis follows fairly directly: If being an epistemic agent essentially involves having mostly true beliefs and making mostly true claims, then we cannot be radically wrong about the way that the world is, and the threat of radical skepticism is no more than a philosopher’s fantasy.1 Many have worried that Davidsonian interpretivism leads to a problematic coherentism. Although any agent’s set of beliefs and utterances must, on Davidson’s account, not only fit together but be mostly true, Davidson’s insistence that ‘nothing can count as a reason for a belief other than another belief’ seems to imply that this truth does not devolve from our confrontation with and response to the empirical world that our beliefs purport to be about. If the beliefs of a subject must be mostly true, while at the same time the content of these beliefs is fixed through holistic reference to other beliefs in interpretation, then truth itself seems to become a product of rather than a constraint upon coherence – a conclusion with unsettling idealist and relativist overtones. Davidson does insist that at least some beliefs must be caused by the objects or occasions they are about, in order for believing to get off the ground at all, but he explicitly denies the role of these causal events in justifying the normative claims of a belief. Hence, while objects may provide an enabling condition for belief, it does not appear that within a Davidsonian picture, our beliefs are formed and transformed in response to the normative claims of objects. Davidson himself argued that his interpretivism implies a type of coherentism, but he has denied the idealist and relativist implications of his account, claiming that despite his coherentist criteria for fixing the content of beliefs, these beliefs are directly and immediately about the real, external objects that cause them.2 But can Davidson’s causal, non-normative relationship between beliefs and objects give to those objects, as he claims, an objective role in fixing aboutness, or do the objects remain no more than a nonepistemic enabling condition for belief? Several recent philosophers have doubted that Davidson’s coherentism can save itself from a pernicious loss of the objectivity of our epistemic states. They have denied Davidson’s causal theory of belief, and argued that we must 1
This characterization of Davidson’s position is intentionally broadly drawn, and raises obvious questions, such as which beliefs are those ‘of the right sort’. In several papers, Davidson argues that a necessary condition for being an epistemic agent is having certain sorts of beliefs about one’s own and others’ beliefs (see, for instance, “Thought and Talk” (1984b) and “Subjective, Intersubjective, Objective” (1996b), among others). For my purposes at this point, only the broadest and most inclusive characterization of Davidsonian interpretivism is necessary or helpful.
2
See, for instance, “A Coherence Theory of Truth and Knowledge” (1986), and “Seeing Through Language” (1997).
182
Rebecca Kukla
understand the connection between mind and world as tighter than the one he gives us. John McDowell’s recent and highly influential Mind and World (1994), and John Haugeland’s more recent – and in my view enormously important – “Truth and Rule-Following” (1997), offer kindred critiques of Davidson along these lines. We can describe the general shape of such critiques as follows. Being an epistemic agent necessarily involves having objective beliefs, and that objective believing must be responsive to normative standards imposed by the objects themselves; these objects, and not just good fit with other beliefs and claims, must serve to normatively govern agents’ epistemic practices. If confrontations with objects play a merely causal role in our epistemic processes, then Davidson’s arguments concerning the necessary byand-large truth of our beliefs will not constitute a sufficient response to the qualms coherentism should raise for us. Objective believing seems to require that in each particular case, such beliefs are responsive and answerable to the objects that they are about. While Davidson may have a story about why beliefs about the world cannot generally fail to correspond to how the world actually is, it does not seem that objective beliefs are merely those likely to correspond to the world. The proper relationship between object and objective beliefs must be not just one of correspondence but of authoritative criteriality for the holding of the belief, from the point of view of the believing agent. Thus, according to this genre of critic, Davidson’s account dooms us to a set of beliefs which ultimately spins free of the world in the sense of failing to be accountable to it. In this essay, I do not intent to pursue the much-discussed question as to whether Davidsonian interpretivism does indeed lead to malignant coherentism. Instead, I will examine Haugeland’s recently developed account of the nature of objectivity and epistemic agency, which is designed to escape the traps of coherentism. His account makes central the need for subjects to be responsive to the claims of objects in order for them to be epistemic agents or wielders of meaning at all. While I do believe that this account succeeds in escaping Davidsonian coherentism, my interest here is in how Haugeland’s theory, in responding to the threat of coherentism, implies the need to work backwards and revise Davidson’s interpretivism, by broadening and enriching the constitutive standards of interpretability and (hence) epistemic agency. Haugeland himself does not consider these implications of his theory, but it is not surprising that a theory of objectivity and objective believing and claimmaking designed to avoid Davidsonian coherentism would end up requiring a revision of his interpretivism as well. After all, the interpretivist holds that the constitutive condition of epistemic agency is being interpretable as making true claims. If we revise our understanding of objectivity, then presumably we
How to Get an Interpretivist Committed
183
may well have to revise our understanding of the criteria and conditions of possibility for claiming something to be true. I will argue that Haugeland’s account can be used to demonstrate that there are transcendental preconditions on what enables us to make true claims or have true beliefs at all, which cannot themselves be characterized in terms of the possession of more such beliefs. These constitutive conditions on epistemic agency will enrich our understanding of the standards by which we can justifiably interpret someone as a being with a genuine stance on the truth, capable of making objective claims and having beliefs about the world.
Davidson on the Constitutive Conditions of Epistemic Agency Davidson claims that “If we cannot find a way to interpret the utterances and other behaviour of a creature as revealing a set of beliefs that is largely consistent and true by our own standards, we have no reason to count that creature as rational, as having beliefs, or as saying anything at all”. 1 Thus interpretability is the condition for the legitimacy of taking someone as an epistemic agent and meaningful claim-maker. In turn, the standards and methodology of interpretation provide a procedure for determining (at the same time) the existence of epistemic agency and the content of the intentional states of the agent. In a breath, Davidson’s claim is that “[a] person comes to understand the speaker of a different language when he can systematically match up his own sentences with those of the alien speaker on the basis of same or similar causes”.2 The primary method of interpretation, for Davidson, is to correlate sentences held true by the interpreter, with sentences held true by the speaker, with events (apparently) causing the speaker’s utterances of these sentences and thereby determining sentence-content. Our attempt to do this is based on our initial hypothesis that the speaker is a claim-maker whose utterances are meaningful sentences. To avoid the circularity involved in suggesting that we must understand meaningful sentences as such in order to be justified in beginning the process of interpreting them, Davidson makes the crucial claim that prior to understanding the meanings of a speaker, we can provisionally interpret some of her behaviour as assent behaviour, whereby we notice that something she is doing counts as committing herself to the truth of a sentence, independent from our ability to interpret the sentence.3 Such assent 1
“Radical Interpretation” (1984a), 137.
2
“Pursuit of the Concept of Truth” (1995), 18.
3
“Subjective, Intersubjective, Objective” (1996b), 164.
184
Rebecca Kukla
relations are empirically observable, as are the relations between utterances and their occasions, and out of these observed relations, holistically interpreted, the interpreter develops and refines a ‘theory of truth’ for the interpretee, which attaches meanings to her sentences. Crucially, Davidson wants these same observed correlations to provide transparent evidence for the beliefs of the interpretee; the interpretation of a speaker’s language and of his beliefs is one and the same process for Davidson. “The result [of the development of this theory of truth] is a theory of both belief and meaning for the speaker, for it must yield an interpretation of the speaker’s utterances, and if one knows both that the speaker assents to the utterance, and what it means in his mouth, one also knows what he believes.”1 For Davidson, there is no trick to determining what a speaker holds true, beyond the trick of correlating his utterances with the world in the right way (where ‘in the right way’ means, among other things, in a way sensitive to holistic constraints of rationality). These correlations provide the criteria for the content and existence of holdings-true. For Davidson, in a situation of radical interpretation (wherein we raise the question of an agent’s epistemic status ex nihilo), the evidence provided by a purported agent’s assents to utterances, and by the occasions of those utterances, is all the evidence we have or need for determining the meanings of the speaker’s sentences, her beliefs, and in turn her status as an epistemic agent who makes truth-claims. Furthermore, he takes this evidence to be self-standing and itself available without the benefit of having already engaged in a process of interpretation. “What is directly available to the learner of a first language or to the jungle linguist is the high correlation between easily observed behaviour on the part of the teacher or informant and easily observed events, situations or occasions in the shared world”.2 Thus, because such correlations are both necessary and sufficient for interpretability, and because interpretability is the constitutive condition of epistemic agency, such correlations are a necessary and sufficient condition for epistemic agency. In this paper, I support taking interpretability as the constitutive condition of agency, for reasons that will become clear later. My difficulties will be with what Davidson takes as the evidence and methodology for interpretation, and with the status of that evidence. I will claim that in order to take someone as a truth-claimer, it is not enough that he utters sentences that can be correlated with truths. In other words, a being who did nothing but tell the truth is an impossibility, on my account – we need to do things other than tell truths in order to tell the truth at all. I will argue that our ability to observe that some1
Ibid.
2
“Pursuit of the Concept of Truth” (1995), 19.
How to Get an Interpretivist Committed
185
one is making a truth claim is not self-standing in the way Davidson thinks, but depends on observation and interpretation of all sorts of other epistemic behaviour on her part. Intertwined with this denial will be my argument that doxastic commitments to propositionally structured facts are not the only type of epistemic commitments that we need to have, in order to count as an epistemic agent.1 Thus I claim that Davidson’s allowed evidential base is too meager to justify assignment of epistemic agency. But in order to actually get to the point where I can give all these arguments, I must take a detour through Haugeland’s account of objectivity and epistemic agency.
Haugeland on the Constitutive Conditions of Epistemic Agency Both Davidson and Haugeland agree that epistemic agents are essentially norm-followers, and that objective claims – whether beliefs or utterances – involve a normative commitment to the truth of their content, or a holding true on the part of the agent. But while Davidson is committed to this normative account of meaning, emphasizing that the possibility of error is the mark of the mental,2 he is not concerned with giving an analysis of the nature or varieties of normative commitment. Rather, his analysis of epistemic agency and interpretability relies, as we saw, on only one variety of committed action, namely straightforward assent to the truth of a sentence, or its psychological correlate, belief. Furthermore, for Davidson, as we also saw, even this one type of committed action receives no analysis, for he treats it as more or less primitive, simple, observable and self-standing. Hence his account, and his methodology for interpretation, focus not on the nature and determination of normative commitments, but on giving content to those commitments by correlating them with propositions. For Haugeland, on the other hand, being able to hold 1
Davidson notoriously rejects the existence of facts, which makes my use of such language potentially problematic in this context. (See, for instance, “The Structure and Content of Truth” (1990)). I find this rejection implausible, and find it extremely difficult to even talk coherently about what makes a truth-claim true without making references to facts. Even Davidson seems to find his own eschewing of facts difficult enough to sustain linguistically that he sometimes slips up and speaks of the ‘fact of the matter’ (for instance, in (1996b), 169). Thus I will refer to facts from time to time in this paper, although I believe that doing so is safe in virtue of the fact that none of the debates concerning the ontology of facts are directly or importantly relevant to the issues discussed here. The diligent reader could, I think, translate all my sentences that refer to facts into sentences that refer to whatever it is that makes truth-claims true, remaining neutral between possible ways of cashing out this reference. For a recent, detailed argument against the idea that Davidson can do without facts, see R. Manning, “All Facts Great and Small” (1998).
2
For instance, see “Seeing Through Language” (1997), 25.
186
Rebecca Kukla
something true requires our ability to be responsive and committed to not just one but several sorts of norms, and the philosophically interesting project is analyzing and giving criteria for these various sorts of norms, as well as figuring out how they have to fit together for a single subject if epistemic agency is to be possible. In Haugeland’s account, the determination of the content of beliefs and claims is intertwined with the determination of what sort of norms are at play and how. This implies that we cannot unproblematically determine the set of commitments an agent has prior to fixing their content, nor is the process of fixing content merely a matter of matching contents with commitments. In “Truth and Rule Following” (1997, hereafter ‘TRF’), Haugeland begins from the observation that there exists at least one fundamental difference between the norms of truth-telling and more traditional social norms such as norms of etiquette: the norms of truth-telling, unlike the others, need to be made up of at least two different sorts, namely norms of propriety, which in this case are the socially accepted norms of epistemic discovery, justification and expression, and norms of correctness, which hold us to getting the world right. In the domain of etiquette, if we are living up to norms of propriety then there is no way that we can be getting anything wrong – the norms of propriety and correctness coincide. But it is essential to the possibility of objective claim-making that it be possible for us to be following the epistemic norms of propriety just fine and still get the facts wrong; hence these two sorts of norms need to be able to conflict. Objective claim-making requires not just the possibility of error, but the possibility of error given social propriety, for otherwise truth becomes solely determined by social agreement. Accounting for this possible gap between the two sorts of norms is, in Haugeland’s view, a minimum requirement for avoiding radical coherentism and for making our claims responsive to the world they are about. We need both types of norms and the possible conflict between them in order to allow objects to ‘resist’ or ‘stand up’ to our claim-making, so that they can be criterial for the correctness of those claims.1 Haugeland’s project is to explain how these norms of correctness are possible: how do we succeed in letting objects govern and set the standards of correctness for the claims we make about them? A Davidsonian picture in which objects cause beliefs but do not justify them will not be in a position to provide an answer, since a merely causal relation between objects and claims is precisely not a normative relation at all. For Haugeland, the key lies in understanding that our epistemic practices must actually constitute objects with the right kind of independence, capable 1
TRF 338.
How to Get an Interpretivist Committed
187
of playing the right criterial role in these practices. He argues that for any domain of objective phenomena, the ability to identify objects in that domain and make claims about their character and relations necessarily involves the ability to apply ‘constitutive standards’ to these objects – the standards for something’s getting to count as an existent case of the sort of thing that it is. These standards are the rules that “govern phenomena that make sense as what they are only in term of these very rules”.1 Haugeland uses game phenomena to make his point vivid; to recite his favorite example, to be a chess piece at all is to be something that is governed by the legal rules of chess in the right way. Only something which moves properly along diagonals in the course of a chess game can be a bishop. While Haugeland makes it clear that we cannot assimilate the case of everyday objects that are not arbitrarily constructed to the game case, he does argue that all objects must be held to and meet such constitutive standards in order to exist at all. Midsize physical objects, for instance, must be subsumable under the laws of nature, must not pop in and out of existence, must be bearers of multiple consistent properties, and so forth.2 In order for us to be able to hold objects to constitutive standards, and to judge authoritatively when and how they have met them, we must possess epistemic skills. We must have skills that enable us to recognize when objects are meeting these standards and behaving ‘legally’, and we must be able to manipulate, recognize and generally cope appropriately with objects that we determine to meet the constitutive standards. Haugeland calls these ‘constitutive’ and ‘mundane’ skills respectively. It is only insofar as we have and exercise such skills that objects can show up for us as determinate and existent at all. But crucially, the reverse also holds. That is, there can only be objects that meet constitutive standards in the first place insofar as there are standardenforcers, who engage in the normatively inflected interactions with objects that require the objects to meet the standards. As Haugeland puts it, constituted phenomena and the skills for recognizing and coping with them are ‘made for each other’. Thus for Haugeland, we do not come into contact with objects via a merely causal relation which produces a correlative belief or utterance. Instead, even the possibility of interacting with an object in such a way that we can come to make a claim about it involves the confluence of a large number of normgoverned and object-involving events of various sorts. If the basic shape of our confrontation with objects is through holding them to standards, then the 1
From Haugeland, “Originary Intentionality” (1998).
2
TRF 348-51.
188
Rebecca Kukla
basic relationship to these objects is not one of mere causality but of authority. In order for an object to be one about which we have beliefs and make claims, it needs more than mere causal powers – it needs to make a specific kind of normatively defined sense. It needs to behave not only efficaciously but legally, and in such a way that we can cope with it in the course of our epistemic practices. The normatively governed interactions between epistemic agents and objects, wherein agent exercise the skills required to make discoveries about objects, collectively comprise the constitution of objects, in Haugeland’s terminology. As he puts it, “Constituting objects is letting them be by finding and showing that they make sense in some determinate way, and can consistently be told and coped with as such”.1 Lest all this talk of constituting objects through holding them to constitutive standards sound arcane, we should remember that the point that the existence of phenomena can be enabled by the imposition of constraints is not a new one. We know that without the rules of discourse there could be no language, without the rules of hockey there could be no hockey, without the rules of voting there could be no voting, etc. In each of these cases a domain of phenomena is enabled by our insistence upon the imposition of bodies of norms, on each other and on the world. Haugeland’s twist is to argue that not only social practices but objects themselves are constituted in this way. But this should not seem mysterious: all sorts of rich social practices are essentially co-constituted along with the objects of those practices. Nothing counts as a chess-piece, a word or a penalty box except insofar as there exists norm-governed chess-play, language and hockey. Haugeland’s goal is to extend this story of the co-constitution of practices and objects to epistemic practices and objectivity in general. However, if this account of the constitution of objects is to become an account of objectivity and objective epistemic practices, then we need understand how the relation between agents and objects can be governed by norms of correctness, as opposed to the more philosophically familiar norms of propriety. In order for Haugeland to escape coherentism and allow objects to be independent tribunals of correctness, it is, crucially, not enough for him to explain how we hold them to various standards, in the sense I have just discussed. It is just as important that the objects be able to hold us to the standards of truth that they impose. This is just the point of his original propriety/correctness distinction, and it is what prevents his constitutive story from turning ‘objective’ truths into mere social agreements produced by conventional norms of propriety. Even if we stick to games and other highly con1
TRF 329.
How to Get an Interpretivist Committed
189
vention-bound domains of social practice, we can see that the objects of these practices are not, in general, ‘mere’ social constructions that we can make however we want. A baseball or a penalty box is a material object that must serve specific material purposes, and the world may well resist our efforts to constitute things as we wish. Merely ‘deciding’ that something is a baseball, or that baseballs have this or that property or behave in this or that way, doesn’t make it so – a baseball, though it is constituted within social practices, is an objective thing about which we can be wrong. Haugeland’s claim that everyday objects are constituted within social practices goes hand in hand with his insistence upon this objectivity which transcends mere social construction. His mark of the objective is supposed to be that objects themselves serve as the authoritative standard with respect to the truth of our claims; they are what our claims must live up to in order to be true, and they must be able to surprise us, rather than being simply subject to our authoritative imposition of norms. It is only thus that our claims have the possibility of turning out to be false despite our best efforts. But it is not obvious, if we believe that objects are constituted by our standards, how such falsification is possible. In Haugeland’s view, to be an object of some sort is to behave according to the standards for being such an object – for instance, an object whose properties change radically and randomly is not just an unruly object, but not an object at all. How, then, can we recognize a violation of our expectations or a challenge to our claims? How do we recognize aberrant entities, as we must if truth is to avoid being fully derivative upon our beliefs, expectations and practices? Haugeland argues that what we need, for any given domain, is a distinction between possible phenomena – the set of phenomena our constitutive standards allow – and conceivable but impossible phenomena. These conceivable impossibilities form what Haugeland calls an ‘excluded zone’ of aberrant events that lie outside the boundaries of how objects are supposed to be able to behave, but which are still sufficiently ‘ruly’ to be intelligible, or recognizable as the behaviours of specific objects. For instance, if we are engaged in a regular chess game, and in the late mid-game our opponent moves her knight over two columns and up two rows, we know that she has done something illegal, which doesn’t actually count as a move in the game, but we do not throw up our hands at this point, and become unable to recognize that the chess game is a chess game and that the knight is a knight. This example move belongs neatly within what Haugeland calls the excluded zone. It is for this reason that we can intelligibly say that she did something wrong which needs correction. If, on the other hand, our opponent regularly moved pieces of each
190
Rebecca Kukla
colour, several to a turn, occasionally flinging them off the board, then we would not count her as playing chess at all, and her actions would not count as illegal moves but as non-moves. Somewhere between these two examples lies the fuzzy boundary between the conceivable but impossible, and the simply nonsensical. It is the excluded zone, Haugeland argues, that makes aberrant entities intelligible. When events occur in the excluded zone, we must explain them away, for they serve as an unacceptable challenge to our constitutive standards. Haugeland explicitly intends these excluded events to function much like Kuhn’s anomalies. For any domain, the excluded zone must be ‘intelligible but virtually empty’.1 If there are more than occasional events in the excluded zone that cannot be explained away, then we know that we have gotten the nature of things fundamentally wrong – we must substantially revise our constitutive standards for how things in the domain are supposed to behave, in which case the phenomena are not of the sort we thought they were. It is this excluded zone which enables the possibility of a relationship to objects whereby we discover rather than merely dictate their nature. It is the possibility of our constitutive standards being refuted by excluded events which makes those constitutive standards responsive to and dictated by objective phenomena themselves, rather than our socially accepted whims. Because our constitutive standards face the tribunal of the world, the objects and events which we discover through the application of these constitutive standards are themselves objective rather than merely socially stipulated. Because our standards can be refuted by the world, they are justified through living up to the claims of the world. Haugeland writes: “The crucial requirement that [excluded events] don’t in fact occur is subject to empirical test. The availability of such tests is what lets disclosure be hard – lets it be an achievement – because it is vulnerable, liable to fail.”2 The excluded zone opens up the possible gap between meeting the epistemic norms of propriety and the norms of correctness. When we make a truth claim, our epistemic authority comes from our having lived up to legitimate norms of epistemic propriety in arriving at that claim. But this legitimacy, in turn, comes from the fact that these norms of propriety are responsive to the way the world actually reveals itself to be. Unless there is a concrete way in which they could fail to live up to the world, they cannot meaningfully be said to be true to or legitimized by it. Thus, without this gap, our truth claims in fact have no authority at all.
1
TRF 333
2
Ibid.
How to Get an Interpretivist Committed
191
But for the excluded zone to produce this gap and enable objectivity, it is not enough that there simply be conceivable events which violate our constitutive standards. For this story to work, the epistemic agent must not only believe that excluded events are impossible in some passive, contemplative sense; she must be committed to taking responsibility for their impossibility. The apparent existence of excluded events must be more than just a curiosity – it must be cause for concern, and it must demand of the agent that she take on the project of explaining the anomaly. This leads us to the final (and for my purposes, the most important) element of Haugeland’s story, namely the need for specific types of practical commitments on the part of agents in order to get objective claim-making off the ground. Commitments bind our actions in various ways – they hold us to projects involving responsibilities. If we are committed, then we care that things go a certain way, which does not just mean that we have some inner affective feeling with respect to them, but that we are normatively bound to make sure that they go that way and to take responsibility (in this case, epistemic responsibility) for their doing so or failing to do so. Davidson’s story, of course, requires commitments from agents, who must hold various things to be true, and hence count as having failed if they are in error. However, this is the only type of commitment he discusses, and he does not analyze the nature of this commitment in any detail, mentioning only that it is manifested in easily identifiable assent behaviour. Haugeland, on the other hand, draws upon many epistemic commitments of different sorts in order to give his account of objective claim-making. To hold objects to constitutive standards is to be committed to their living up to these standards, and this is already an epistemic commitment distinct from that of holding something propositional to be true. To hold something to a standard, we must do more than understand the standard and notice that the object meets this standard. We must be invested in the project of keeping it to that standard, and in turn, this requires an authoritative relationship between us and the object which is more active than that of just having contemplative beliefs about it. Imagine the impossibility of playing chess with someone who had a passive understanding of the rules – that is, she believed that bishops are supposed to move diagonally, and so forth – but who didn’t care at all whether she or her opponent won or lost, or even played legally. If, when you point out to me that I just moved my rook diagonally, hence illegally, I acknowledge the point but show complete indifference, you would be forced to conclude that I am simply not playing chess; the chess norms are not in fact gripping or governing me at this time, even though I have correct beliefs about what those norms are. A range of enthusiasm and investment in the game is
192
Rebecca Kukla
allowable, but I must be committed to playing correctly and well in order to play at all. Understanding the rules is thus not enough to constitute being bound by them – this binding depends on our commitment to avoiding breaking the rules, and to caring when they are being broken by others. Likewise, I do not count as being bound by norms for determining the truth and justifying truth-claims unless I am committed to upholding these norms (the norms of epistemic propriety), and doing so involves committing to the enforcement of constitutive standards. For Haugeland, this is condition for the existence of objective belief itself. The excluded zone and its role in objectivity force us to deepen our account of the nature and varieties of epistemic commitments. We saw that events in the excluded zone must be intelligible but (for the most part) not actual. But this is not the ‘must’ of mere descriptive prediction; it is the ‘must’ of an active, committed epistemic project on the part of agents. An integral part of the commitment to holding objects to constitutive standards is commitment to the non-occurrence of excluded events. It is not enough that excluded events indeed don’t happen; the excluded zone essentially functions to get objectivity off the ground through its mattering to agents that excluded events not occur. Agents must not only believe that excluded events are impossible; they must greet an excluded event with care and concern rather than indifference.1 That is, in the face of a seemingly impossible observation, she must pursue the project of finding an explanation which shows that an excluded event did not occur after all. For instance, when a child sees, for the first time, a puddle on the road on a sunny day that disappears as she gets close to it, she makes inquiries and finds out about puddle mirages, just as a scientist does not ignore an unexpected result in the laboratory. Furthermore, if excluded events persist and begin to intrude upon an agent’s attempts to cope with the world, then at a certain point she must admit that her standards for when the world is working ‘properly’ – her constitutive standards which disclose objects to her – must be revised, If this happens, then, importantly, she has failed, because holding objects to constitutive standards was a commitment of hers, and the correctness of those constitutive standards was her responsibility and necessarily a matter for her concern. Haugeland’s favorite examples of this type of failure occur within the domain of natural science, where they amount to the catalyst for a Kuhnian scientific revolution, in which the standards constituting the being of phenomena must be replaced. But if Haugeland’s account of objectivity is to be a general one, and if we are to assume that everyday observations and judgments are objective if anything 1
Haugeland puts the point this way in “Original Intentionality” (1998).
How to Get an Interpretivist Committed
193
is, then we have to assume that the constitutive standards for the being of everyday, midsize objects themselves produce an excluded zone, consisting of intelligible events which could (but won’t!) challenge our fundamental common-sense ontology.1 The possibility of the illegitimacy of our epistemic norms – and not just the possibility of needing to revise our particular beliefs about empirical objects – is an enabling condition of our being committed to these norms at all. It is only once the norms we follow can turn out to be the wrong norms that our relation to those norms can be one of substantive commitment. We can only count as committed to a norm to the extent that we at least in principle recognize it as justified or legitimate. If we do not take ourselves to be responsible for knowing why we should follow a norm, then we cannot really say that it binds us.2 If we act in accordance with it under these circumstances, then it is binding us through something like blind, automated habit or physical force, but not through its legitimate grip. But this is tantamount to saying that our relationship to it is not really normative, as it cannot have any authority over us. To be committed to and bound by the claims of a norm is to understand the norm not just as something that happens to describe our actions, but as something it is right to follow. We can do this without having easy access to a story about why it has this legitimate authority, but at a minimum, to be bound by a norm, we must be capable of recognizing that it is our responsibility that there be such a story and that it be right. In the case of epistemic norms, what makes them legitimate must be that they are true to the world and enable justified, true belief. But this is exactly what the excluded zone makes possible: our ongoing attempt to exclude events in the excluded zone, and our success or failure at this in the form of their nonoccurrence or occurrence, provide the objective, empirical test against which the correctness of our epistemic norms can be measured, and which they are not allowed to fail. It is because these norms have been put to this test that we can tell a story about why they are justified and authoritative and hence worthy of our commitment, which we must be able to do in order to count as committed to them at all. A commitment to the discovery of objective truth cannot be reduced to a commitment to follow epistemic norms; it must involve a commitment to the legitimacy of these norms in virtue of their ability to let us get things right. It is because of this necessary possibility of a radical epistemic failure belonging to the agent – a failure for one’s epistemic practices to live up to the 1
See the final section of this paper for a detailed discussion of this point and its implications.
2
See Hegel’s brilliant development of this point through his reading of the myth of Antigone, in the Phenomenology of Spirit.
194
Rebecca Kukla
tribunal of the world – that Haugeland’s account of the constitution of objects and the conditions of possibility for objective claims purports to supersede the threat of idealist coherentism. The threat of failure is constitutive of the possibility of getting things wrong, which is in turn the condition for the possibility of getting things right (rather than just making them up as we go along). We cannot merely invent some norm-bound game that results in our uttering sentences and count the playing of this game as the making of truth-claims; we need a justification for thinking that these norms are not only norms of practice but norms which yield truth. Luckily, because objects themselves can resist our constitutive standards, they can be genuinely independent of us even though they are constituted and disclosed by our practices. This general independence is what enables specific truth-claims to be made true or false by the nature of the objects they are about. It is through our commitment to holding them to standards that they can come to be authoritative tribunals of our claims, so that they in turn hold us to the standards of truth that they set. Because of epistemic commitments that we make, we enable the objective world to have a character that is not up to us. Objects participate in the achievement of committing to an adequate set of constitutive standards that is justified by its fealty to the world, and as Haugeland puts the point, “objects can bring this achievement down, even though they would destroy themselves in the process”.1
Beyond Coherentism and Behind Interpretivism In this section, I will argue that on the basis of Haugeland’s account of objectivity and objective claim-making, we can discover and articulate substantial limitations of Davidson’s version of interpretivism. Haugeland sums up his account of objectivity by giving three criterial features of objective phenomena: (1) They are accessible through the exercise of our mundane skills; (2) “They have normative status as criterial for the correct exercise of objective skills (they are authoritative)“; (3) They are independent of consensus and of any specific instance of the exercise of skills.2 Criterion (1) denies radical skepticism by asserting that we are not epistemically cut off from the world, and (3) establishes the realist credentials of the world. The trick is to find a way to give an account of objective phenomena which satisfies both (1) and (3), which seem to be both necessary to objectivity, and in tension with one 1
TRF 353-4.
2
TRF 325.
How to Get an Interpretivist Committed
195
another, given the classical picture of mind/world relations within which things must belong on one or the other side of this division and not both. For Haugeland, it the satisfaction of (2) which is the bridge enabling objects to satisfy both (1) and (3). As we know, it is precisely (2) which represents a radical departure from a Davidsonian picture. Davidson denies (2), and asserts that there is a merely causal, non-epistemic relation between objects and agents. More generally, for Davidson, epistemic space is defined by normative relations of authority between intentional states, and the external world of objects is defined by merely causal properties and relations. But this division forces us to choose between two possibilities – the objects of beliefs and truthclaims must be either normatively or merely causally defined. If they are normatively defined, then they are no longer external to the knower in the right way as demanded by (3), and if they are not (which is what Davidson claims), then we will not have access to objects via our epistemic practices as demanded by (1), since these epistemic practices will not be object-involving. Haugeland encapsulates the relevant charge against Davidson as follows: All is ... lost if, from the outset, object and response are understood in terms of fundamentally disjoint conceptual frameworks – the causal/law-governed and the normative/rational, say. For then the relation between them could not intelligibly be either causal or normative; whereas any external constraint that is to transcend mere coherence must be both.1
Haugeland, we saw, solves this problem by making objects constituted by our normative practices, and yet properly independent of these practices in virtue of the resistance enabled by the excluded zone, thus ‘exorcising the specter of mere coherence’.2 Thus objects enter directly into to structures of epistemic authority and practices, thereby satisfying (1), while at the same time having the independence they need to meet (3), by being wielders of authority in addition to subjects of authority. The condition for the possibility of the existence of an object is that it can be coped with through the exercise of agents’ epistemic skills, in virtue of its living up to imposed constitutive standards, and thus the normative structure and epistemic accessibility of objects is built into their ontology. On the other hand, these accessible objects are independent, because the norms governing our epistemic interactions with them are not up to us but may fail to be legitimate in the face of the behaviour of the objects. This independence gives objects the authority over what is true of them. Hence they meet criterion (2). Despite their constituted status, they are ‘out of our control’, for “the constitution of the domain determines what 1
TRF 346.
2
Ibid.
196
Rebecca Kukla
it comes to for them to be or to behave in this or that way, but whether or not they do is ‘up to them’ – and skillful practitioners can tell”.1 Criterion (2) combats the threat of coherentism, because Ordinary or ‘mere’ coherence ... lacks this element of independent constraint. It leaves no intelligible space for the objects that the propositions are supposed to be about to resist those propositions ... Constituted objects themselves, unlike any beliefs or statements allegedly about them, can ‘talk back’ ... skillful mundane performances can be normatively beholden to [objects] for the correctness of their results.2
Thus Haugeland’s avoidance of coherentism turns on his account of how we can come to be ‘beholden’ to objects in virtue of our own normative commitments. It is the role of commitment and beholdenness in epistemic agency which will force us to enrich the Davidsonian account of the enabling conditions and methodology of interpretation. We know that if we accept Haugeland’s story, then the causal relationship between objects and agents that Davidson allows will not be sufficient to enable objective truth claims, no matter how tight of a systematic correspondence between utterances and events we manage to detect in an agent. Our claims can only be about objects if those objects function within our epistemic practices as the tribunals for the correctness of our claims, and for Haugeland, this means we need to stand in a specific, complex normative relationship to objects. Furthermore, we can now see that evidence that an agent assents to a set of properly connected claims cannot provide a self-standing foundation for the ascription of epistemic agency. An agent who can make objective claims needs to be committedly engaged in specific sorts of projects involving objects, where the failure of these projects is always a governing threat, in that the failure would count as a failure precisely because it matters to us that we not fail. This mattering cannot be reduced to a systematic pattern of asserting certain truth-claims and not others. Rather, the possibility of failure must affect our committed activities, and our commitment is manifested in our acting so as to avoid or explain away a certain result, and as one whose goals have been frustrated when these results occur. The possibility of our developing a theoretical belief that a former belief of ours turned out to be false is not sufficient for grounding epistemic agency; we need to practically care about such results. Davidson believes that epistemic agents need the ‘concept’ of objective truth, and, equiprimordially, the concept of the possible incorrectness of our 1
TRF 347.
2
TRF 347-8.
How to Get an Interpretivist Committed
197
claims. He writes, “someone cannot have a belief unless he understands the possibility of being mistaken, and this requires grasping the contrast between truth and error – true belief and false belief”.1 But Davidson characterizes the need for these concepts as though understanding the relevant possibilities is just a matter of the possession of a set of beliefs with the right content, manifested in our ability to make the right truth-claims about these matters. However, from within Haugeland’s account, the constitutive condition for understanding the possibility of false belief, in the way needed to make us into epistemic agents, is practically behaving so as to exhibit our commitment to various epistemic projects, which demand more from us than correct doxastic commitments. We must actively pursue the truth and avoid false belief, and doing so involves a commitment to a set of norms of justification, observation and worldly negotiation. This commitment itself is grounded in our commitment to putting these norms to the test and making them survive the threat of the excluded zone. The excluded zone must be more than a passively conceivable realm of phenomena that we believe won’t occur; it must, as Haugeland puts it, function in our lives and activities as a ‘practical burden’, a threat for which we take practical epistemic responsibility. All of these commitments must be manifested in concrete practices that go far beyond the mere uttering of what are consistently interpretable as true sentences about these issues. For Davidson, only beliefs justify beliefs, and hence the norms of justification are norms for relating beliefs to other beliefs. But for Haugeland, the norms of justification govern the practices whereby we confront and interact with objects, where these practices take the form of holding objects to constitutive standards through the exercise of constitutive and mundane skills, and being held by objects to standards of truth. The objects themselves, on this account, participate essentially in both setting and meeting the norms of justification. These norms cannot be cashed out without reference to the whole structure of our epistemic commitments, and, unlike in Davidson’s account, these commitments will not be a uniform set of doxastic commitments, but must be of various sorts. This structure of commitments, together with the objects themselves, determine which beliefs we must have and which we must give up, where these ‘musts’ have practical repercussions beyond impacting the future content of our beliefs. The question that opens “Radical Interpretation” (1984a) – Davidson’s original, seminal text on interpretivism – is, what kind of evidence could we have that would enable us to know that someone means some particular thing by a sentence? Such knowledge, of course, involves knowledge of what in 1
“Thought and Talk” (1984b), 170.
198
Rebecca Kukla
particular she means, but also knowledge that she is an epistemic agent who means things and that her utterance is a meaningful expression in the first place. For Davidson, it is the process of interpretation which fixes the particular meaning, and this same process is what confirms and justifies the hypothesis that the interpretee is an epistemic agent, though the initial evidence for this is provided by our pretheoretical, fallible ability to observe that she is the kind of being who assents to claims and hence has doxastic commitments. Given that we can observe this type of commitment, and that it is the only one relevant to getting interpretation off the ground, the justification of the assignment of epistemic agency through interpretation takes the form of correlating assents with sentences and causally efficacious events, pairing up contents with assents, and making sure that it is possible to consistently do this in such a way that the interpretee turns out to have mostly true, rational doxastic commitments. But based on what we have already discussed, we can see that this is not a sufficient account of the methodology of interpretation and the justification of epistemic agency if we believe Haugeland’s account. Within a picture like Haugeland’s, an interpreter will not be able to get away with merely holistically fiddling with the contents of commitments, while taking the structure of the commitments themselves as an independently available given. Instead, the interpreter will also have to engage in the intertwined process of holistically interpreting the nature of the commitments by understanding them in the context of other commitments. That is, in order to take something, quite independently of its content, as an assent indicating a doxastic holding-true, the interpreter will have to understand the agent as the kind of being who has a sufficiently rich structure of commitments of the right sort to be a doxastic truth-holder at all. This will involve understanding the agent not only as one who holds things true, which we can no longer understand as an independent ability, but as one who holds various excluded events to be impossible, who holds objects to constitutive standards, who holds herself to her responsibility for the rightness of those standards and so forth. It is only when an agent has such a host of different commitments, manifested in various practices, that fit together in the right way – i.e. when her apparently epistemic practices such as her mundane skills are responsive to correction and challenge from apparently excluded results – that she can be said to have doxastic commitments at all. Hence the process of interpretation must interrogate and holistically fix not only the content but the nature of epistemic commitments, and it cannot treat doxastic commitments as self-standing evidence for epistemic agency. Of course, if we know that an agent is making a truth-claim, then, if we can figure out its content, we also have sufficient evidence to justifiably attribute
How to Get an Interpretivist Committed
199
epistemic agency to her. For if she does make interpretable truth claims then she is an epistemic agent. The point is not that such evidence would be insufficient, but that such evidence is itself only available within the context of all sorts of other, interdependent evidence concerning an agent’s commitments and normative practices. I can think of two ways in which a Davidsonian could try to defend classic interpretivism, while accepting Haugeland’s account. Firstly, she could argue that all these various sorts of commitments can be understood just as special types of beliefs or other doxastic holdings-true. If so, then the Davidsonian can stick to her claim that the interpretive process is just a matter of fixing the content of these holdings-true, and that being a truth-holder is an independent, self-standing property sufficient for constituting epistemic agency. Secondly, she could claim that these other types of commitments, such as a practical commitment to the impossibility of excluded events, are not provisionally observable from a third-person perspective, but are at best inferable from other evidence, and hence cannot provide part of the initial data which is first recognized and later legitimized during the interpretive process. (As we will see, Haugeland himself holds a position similar to this). I will discuss these two possible defenses in this section and the next, respectively. Can the commitments an agent must have in order to be an objective truthclaimer be understood as elaborate doxastic commitments? For instance, we have seen Haugeland’s arguments for why, in order to be an epistemic agent, we must hold events in the excluded zone to be not only non-actual, but impossible. But can’t we just understand this as the requirement that agents have some doxastic commitments of the form, “It is true that this excluded event is impossible”? This will not work, because the relevant commitment needed for objectivity is not merely a belief in the impossibility of excluded events, but a practical commitment to taking responsibility for their impossibility. An epistemic agent must do some work to keep the excluded zone empty – she must search for explanations for anomalies, caring and taking it as a threat when apparent excluded events occur and taking it as a failure when they cannot be explained away. In turn, this caring, working and the like cannot be reduced to the holding of a set of truth-claims. The ‘practical burden’ of commitment could not be exhausted by the uttering of the right claims; it must be manifested in expressions of frustration, effort, disappointment and the like. The norms for coping with an event for whose impossibility we have a practical responsibility are different, in content and in form, from those for coping with an event which we merely didn’t believe would happen. Mere belief in the impossibility of an event – if there could be any such thing, which Haugeland’s account makes doubtful – is consistent with indifference to the
200
Rebecca Kukla
event. In other words, in the face of the ‘impossible’ event a mere believer in this impossibility could revise her beliefs but not any of her other behaviours. But this practical indifference is just what Haugeland’s account rules out – agents must be committed, not just to an understanding of the boundary between the possible and the impossible, but to maintaining that boundary. We can put the general point here in Kantian terms: the constitutive requirements for epistemic agency include commitments of both theoretical and practical reason, and it is simply a category mistake to try to translate or reduce practical commitments to theoretical commitments. The way that Haugeland-style epistemic commitments to constitutive standards impose practical constraints upon our actions, coming along with governing concerns and responsibilities, is simply not the same as the way that our commitments to truth-claims impose constraints upon our other commitments to truthclaims. An interesting and important result here (though not one that was lost on Kant) is that the commitments of practical reason can themselves be epistemic commitments. Our practical commitment to constitutive standards, manifested not just in a theoretical belief in those standards but in the exercise of our constitutive and mundane skills, is an epistemic commitment, in the sense that it plays a direct role in determining how we believe that the objective world actually is. A Davidsonian might respond, at this point, that this discussion of practical reason just reminds us that interpretable agents must have not only beliefs but desires, otherwise their beliefs would not be manifested in any behaviour at all, including even the behaviour of uttering a sentence expressing that belief. Desires are, after all, practical commitments. Thus, this response goes, we can just understand agents as having a necessary desire that the excluded zone be empty rather than a belief in its emptiness, and so on. Or perhaps we could understand the commitment to constitutive standards as something like the belief that these standards hold plus the desire that this belief be right.1 In any case, according to this optimistic critic, we can understand the type of commitments I have been discussing as some combination of traditional beliefs and desires. I think that there are at least two problems with this line of approach. First of all, I have just argued that the commitments we are considering are essentially both epistemic and practical, and it seems dubious that we can understand them as just a concatenation of epistemic commitments and practical commitments. Regular desires cannot be understood as epistemic commitments; an agent’s desire that something be true does not, if she is rational, of itself help to make it true, unless that truth is created by the actions of the 1
Many thanks to Richard Manning for giving me this clear formulation of the objection.
How to Get an Interpretivist Committed
201
agent (as in when we make it true that there be a turkey sandwich in front of us by going and making one). But our commitment to constitutive standards is a commitment to their turning out to be right, which involves our practically holding objects to them, and without relapsing into degenerate idealism we cannot understand this in terms of our simply making it be the case that they are right. A belief in the constitutive standards plus a desire that they be right does not add up to our taking responsibility for their objective rightness, nor does it seem like any addition of regular beliefs and desires to our analysis will help to capture the missing element of this type of responsibility. If the practical commitment to constitutive standards can be understood as a species of desire, it is a special epistemic desire, of a sort that does not figure in regular Davidsonian accounts of the relation between belief and desire. I can also make a more straightforward, methodological point here. Davidson’s story about the evidence which initiates interpretation draws only upon assents and their holistically satisfying correlation with events and sentences held true by the interpreter. It is true that in order for this story to make sense at all, we must presuppose some desires on the part of the interpretee, namely, desires to manifest or communicate her beliefs in assents. However, these desires explicitly do not play a separate evidential role in initiating interpretation, for Davidson. This is an important, motivating correlate of his claim that we can presuppose sincere assent as the default interpretation of agent’s utterances. In provisionally observing something as an assent, we implicitly assume that the agent desires to sincerely communicate her beliefs with this assent, and thus we do not need to attend separately to her apparent desires and motivations in the initial stages of interpretation; we get an interpretation of the relevant desires automatically through the process of building a truth theory for her sentences. Davidson builds in the methodological initial assumption that the agent’s relevant desires correlate transparently with her assertions; furthermore, he claims that we need not know or provisionally observe anything about the rest of an agent’s desires and motivations in order to provisionally recognize assents as assents, and hence to begin the process of interpretation which will hopefully justify and flesh this provisional recognition. Hence, even if it turned out that we could interpret the epistemic commitments that I have been discussing as special types of desires, my position would differ significantly from Davidson’s, in that I am arguing that we must give significant attention to many of the practical, non-doxastic commitments of an agent, and situate their commitments to truths amidst these practical commitments, in order for these commitments to truths to show up as commitments to truths at all.
202
Rebecca Kukla
I have claimed that in order for an agent to be able to make or assent to a truth, she must do so in the context of her other commitments, not all of which can themselves be doxastic commitments. It is necessary to separate two levels at which the interpretation of something as a truth-claim may be holistically dependent upon the attribution of other commitments to the claimer. One is that after the interpretive process is off the ground, we may need to interpret and ensure the existence of the agent’s other epistemic commitments, including her practical commitments such her commitment to the emptiness of the excluded zone, in order to justify and flesh out the content of our attribution to her of beliefs and truth claims. If this were all I were claiming, then my discussion would not constitute a substantive criticism of Davidsonian interpretivism, but at most an intriguing elaboration of certain stages of the interpretive process. But the other level is that of the initial stages of interpretation; here my claim is that when we notice something as an assent in the first place, even provisionally, we do so in the context of noticing that the being we are considering generally behaves like an epistemic agent, and that this means noticing that they manifest various other types of committed behaviour as well. A machine which simply blurted out ‘yep’ in a mechanical voice every five seconds would never occur to us as a candidate for epistemic agency, and its sounds would never be taken for assents. In fact, out of context, there is no self-standing, isolatable behaviour that marks itself as assent behaviour. We take behaviours as assents when they come, under appropriate circumstances, from beings who generally act like agents. I am claiming that ‘acting like an agent’ in the relevant way involves manifesting all sorts of epistemically significant behaviour other than assenting, in the context of which assents count as assents, and that our recognition of this other behaviour must be equiprimordial with our recognition of assent behaviour in order to initiate interpretation. If this is right, then my account departs from Davidson’s, by revising and enriching his story about what evidence is sufficient to begin interpretation, and by challenging the sufficiency of his described method of truth-theory construction for grounding the attribution of agency in a situation calling for radical interpretation.
Observing Agency This last move takes us to my second line of defense for a Davidsonian, namely the argument that the types of commitments I am discussing are not observable from a third-person perspective and hence are not appropriate data for use by the interpretivist. At this point in my argument, this line might go
How to Get an Interpretivist Committed
203
something like the following. The practical epistemic commitments of which I speak are not observable, and hence have no place in a theory of interpretation or interpretability. It may well be that we must have these commitments in order to be an epistemic agent who makes truth claims, but this does not mean that an interpreter must observe them in order to attribute agency and interpret the content of truth claims. Rather, we infer the presence of these commitments from the interpretable utterances of the agent taken together with her other non-epistemic behaviours. Furthermore, it is these nonepistemic behaviours that give the context that allow us to provisionally recognize assent behaviour as such – we need not take other committed behaviours as part of our initial evidence for attributing epistemic agency. Hence while Haugeland’s account may enrich our understanding of epistemic agency itself, and while it may even enrich our more specific understanding of the conditions for the possibility of agency, it does not affect our understanding of the proper methodology of interpretation. I think that a version of this line is, as we will see in more detail, Haugeland’s own position. My response to this argument will take me to the very heart of the version of interpretivism and the account of epistemic agency that I am developing in this paper. I will argue that normative commitments other than those to the truth of various sentences are indeed observable from a third-person perspective, and that once we see this, we remove the intuitive plausibility of the idea that the contextual behaviours which make the recognition of assent possible are themselves non-epistemic behaviours. What I want to urge is that interpretation is first and foremost a process of interpreting the commitments of the other, where these commitments are the initial data for the interpretation, as well as what the process of interpretation justifies attributing. They are, as it were, both the initial resource and the payoff for the interpreter. Davidson agrees, but restricts the relevant commitments to doxastic commitments, whereas I want to say that, at the beginning and the end of interpretation, doxastic commitments show up only in the context of other types of epistemic commitments. First, we must see clearly why no amount of uttering can be justifiably interpreted as truth-claiming, in the absence of some contextual behaviours (a claim that Davidson would not dispute, as such). For any set of utterances whatsoever, even once their content is holistically interpreted (perhaps through Davidson’s correlative method), we can always draw a distinction between a being who holds their content to be true, in which case the utterances are truth-claims, and a being who merely utters the words. Even if an agent utters phrases such as “No, really, I’m not just saying this, I truly believe it”, or “I care if it turns out to be false”, there is no contradiction in the idea
204
Rebecca Kukla
that these phrases are themselves just utterances not backed up by committed belief. Thus the investment we have in our claims always exceeds the fact of their expression. Now this will have no impact upon the nature and methodology of radical interpretation if every utterance can be assumed to indicate sincere assent. But even if this assumption can be applied to epistemic agents, this move just reinstates the issue here, because we need a reason to believe that an utterer is an epistemic agent and hence a sincere assenter, and the point is that utterances alone cannot justify such a belief.1 However, the fact is that typical agents do all sorts of things all the time that indicate their investments and commitments other than just making direct statements of and assents to claims, and hence we in fact have plenty of contextual reasons to interpret their utterances as the sincere assents of epistemic agents, rather than as robotlike emissions not backed up by commitment. Thus an appeal to all these other ‘sorts of things’ must be made in our initial interpretive judgment that we have an epistemic agent on our hands, as well as during the process of justifying (or revising) this initial judgment through working to interpret her truth-claims. The question, at this point, is whether these other manifestations of investments and commitments are themselves observable as such, or whether we observe some other type of contextual behaviours, and infer that they indicate normative commitments of the needed sort. But what motivation could their be for denying the commonsense fact that we often see people trying to do things, getting frustrated, being concerned, and exhibiting other essentially normative, committed behaviours? It seems that the motivation can only be one springing from a certain sort of reductionist impulse. We may wish to hold that it is only naturalized behaviour describable in physicalist, non-normative language that can be observable, in virtue of a more general philosophical commitment to naturalized events being the only observables, and hence we may hold that an essentially normative action such as trying to do something can only be inferred from the observational evidence rather than observed itself. I think that there are several reasons to resist this reductionist impulse. Firstly, we can argue that there is no possible set of ‘natural’2 observations which could possibly justify the attribution of the normative behaviours we need for epistemic agency, given a Haugeland-style account. In order to be a truth-claimer, on such an account, we must be differentially responsive not 1
This point implies a rejection of the Turing test in its classical form.
2
I am using the physicalist reductionists’ sense of the terms ‘natural’ and ‘naturalized’ here, since they are my target, though this usage is contentious, and in my view ultimately untenable. See McDowell (1994) and Hornsby (1997) for active attempts to use these terms in ways that are not reductionist and which do not drain nature of normativity.
How to Get an Interpretivist Committed
205
just to objects such as rabbits and shoes, but to such normative features of the world as authority and claims. Davidson ought to agree with this point – after all, being an interpretable agent requires being an interpreter of one’s self and others, for him, and hence an epistemic agent must not only respond to natural events in the world but to its own beliefs and to the claims of others. Now authority and claims cannot themselves be reduced to natural features of the world, in such a way that, in turn, we can give a reductive behavioural account of what counts as responding to them. We can see this by noticing that for any set of natural behaviours at all, described in absence of any normative commitments, it will always make sense to ask if these are in fact the actions that they appear to be if we do understand them as the actions of a committed agent. In other words, if when we describe some behaviour, we artificially strip away any references to terms indicating commitments, such as ‘believing’, ‘trying’, ‘preventing’, etc., then the question as to whether our described behaviours are in fact committed actions will always be left unanswered by the initial description. For instance, we can always ask whether a being who is making what we take to be legal chess moves is in fact playing chess, and the answer necessarily requires looking at more than whether appropriate movements are being made. As Haugeland says, Obviously... there is more to playing chess than just moving the ostensible pieces in accord with the rules ... For one thing, the players should be trying to win ... Any player of chess ... has an investment in the legality of all the moves in the game ... you must insist that both your own and the opponent’s moves be legal.1
But – as proven by the little ‘chess-playing’ programmes available for free all over the internet, which none but the philosophical extremists would count as genuine, committed chess players, but which can ‘hold their own’ against most amateur players – these investments, insistings and the like, are not reducible to the moves made. Each of these rests for their its possibility upon the ability to recognize authority, or binding normative force, and it seems that we are precisely missing the point of needing this recognition in our account of epistemic agency if we mistakenly think that this recognition is reducible to giving differential responses to natural objects. Recognition of authority is in fact the transcendental condition which makes possible the recognizing of objects at all (as opposed to just differentially responding to them). While I think that the transcendental status of the recognition of authority does preclude this recognition from being either reducible to naturalized behaviour or inferable from naturalistic observations, this doesn’t yet prove 1
TRF 340.
206
Rebecca Kukla
that activities involving such recognition are themselves observable in some other way. But the fact is that we all know perfectly well, for the most part, what it looks like when a being is responsive to claims and the holder of commitments. We know what it looks like for a being to try to do something, to care when it succeeds, and to consider itself held to standards whose authority it recognizes, whether this is epistemic authority or some other sort. This knowledge cannot come from inferences from or translations of naturalistic observations, I have argued, and I think that there is nothing other than a hard-core general commitment to naturalism of this sort which could make us deny that we learn about such things from observation. Therefore, it seems we should conclude that we simply observe them directly. We just need to grant that the processes of initiating and interpretation will always make reference to irreducibly normative actions. We must interpret others as trying to carry out various projects, insisting on various results, and the like. And though we cannot capture these features in a pure, naturalized behavioural language stripped of any normative vocabulary, we also do not need to be looking for some supernatural features over and above material behaviour – it is simply the regular old things we do that are being observed here, expressed in the irreducibly normative terms appropriate to that behaviour.1 Trying, caring and insisting just are the kinds of material activities in which agents engage. Haugeland points out that activities like these are no more to be associated with phenomenological mental states than they are with physicalist movements. For instance, in order for an agent to insist on a result, it is neither necessary nor sufficient for her to have an inner ‘insistence-impression’ or other phenomenal state, any more than it is necessary or sufficient for her to move her body in some way specifiable in naturalistic terms. I would go farther and say that there is nothing above and beyond material behaviour that qualifies an agent as being insistent – insisting is an example of the kind of material behaviour we engage in. We will find this problematic only if we have already committed ourselves to the idea that every material event must have a naturalistic description, which is the most fundamental or primary description of what is real and what is observed. But no interpretivist ought to be committed to such a thing, including a traditional Davidsonian interpretivist. Davidson, after all, bases his interpretive method on the initial noticing of assents, which are of course normative, committed actions. Davidsonian evidence never was naturalistic – the Davidsonian interpretivist took someone as making a contentful claim by holistically interpreting her assents, while taking it as a (provisional) governing 1
For a sustained defense of a similar position, see Hornsby (1997).
How to Get an Interpretivist Committed
207
assumption that the agent’s utterances expressed meaningful, rationally connected doxastic commitments. Davidson himself is notorious for arguing against the reducibility of the space of reasons or mental events to the space of naturalized physical events.1 Therefore, an interpreter who tried to proceed by using merely behavioural evidence stripped of its meaningful content could not get her interpretation off the ground. Hence from a Davidsonian point of view, my position ought not to be read as a new insistence that interpretivism should part ways with naturalism. The difference is that while Davidson acknowledged the observability of commitments, he considered only one kind of commitment (the doxastic), and did not concern himself with the conditions of its observability. Instead, he spoke as though the observation of such commitments was a trivial and self-standing matter not in need of further analysis. Combined with his methodological assumption of sincere assertion, the observation of commitments in Davidson’s account ends up sounding so easy, and so little in need of hermeneutic interrogation, that we do not notice the way that it requires us to detect features of the world that go beyond the naturalistic. But once we notice that to be a Davidsonian, we must already be able to observe essentially normative, committed behaviour from a third-person point of view, then there are no good physicalist motivations for denying the interpretivist the ability to observe other such behaviours. And, once we attend to Haugeland’s arguments, and hence notice that the existence of doxastic commitments is holistically dependent upon other types of epistemic commitments, there is in turn no reason not to include observations of these commitments as part of the fundamental methodology of interpretation, equiprimordial with, and a condition for the possibility for, the observation of doxastic commitments. For Davidson, interpretation negotiates only the contents rather than the nature and structure of commitments. What I am doing is not adding the observation of commitments to Davidson’s story, but shifting the emphasis so as to reveal the need for a substantial focus on the interpretation of the structure of an agent’s broad normative standing. In an effort to reaffirm physicalist inclinations, an objector might worry that normatively governed behaviour cannot be the kind of thing we directly observe, since these observations are blatantly fallible. It may turn out, turning the process of interpretation, that what seemed to be an assertion or a trying on the part of the interpretee was not what it seemed to be, either because the interpretive process failed, or because it revealed the agent as a very different kind of being than the interpreter originally thought it to be. This may make it seem that knowledge of an agent’s normative situation can only come, at 1
For instance, see his “Mental Events” (1980)
208
Rebecca Kukla
best, as the result of inferences at the end of interpretation, rather than as part of the original data for interpretation. Such an objection would be as appropriately leveled at Davidson as at me, since he too takes observed committed behaviour as the empirical starting point for interpretation. But to confuse the inferential nature of the justification of such knowledge with its nonobservability is to misunderstand the heart of the interpretivist’s method. A central point for Davidson, as we have seen, is that the very same meaningful, essentially normative behaviours (in his case, only assent behaviours) which provide the initial, observed data for an interpretive process are the behaviours that get interpreted holistically, and hence whose attribution gets justified, during this process. It is a key element of the interpretivist’s method that one’s initial observations are fallible, since the behaviours observed are both the starting point and the eventually justified pay-back of interpretation. If we can accept that behaviour manifesting doxastic commitments is observable in this special way, then I see no reason to deny observability to other kinds of committed behaviour. And, if we can observe such behaviour, then, to return to an earlier point, there seems to be no special plausibility to the claim that the observed behaviours in the context of which assents show up as assents are themselves naturalized behaviours stripped of their normative significance. If we, like Haugeland, understand assents as possible only in the context of other epistemic commitments, then it makes much more sense to let these very commitments form the interpretive context in which assents are disclosed to us in the first place. After all, the same process of justifying initial fallible observations occurs in the everyday observation of inanimate material objects as well. Our provisional observations of the world are always fallible (as those old philosophers’ favorites, puddle mirages and barn facades, are always there to remind us).1 But, if we find ourselves in a situation where we need to revisit and be more precise about what it is that we are observing, we certainly can and do use those initial observations as part of our data, out of which we construct a better account of what we are seeing, and perhaps reinterpret that original data along the way. The reason that the case of interpreting agents may seem different is that the whole story about interpretation is designed to show the workings of cases where it is not obvious whether to trust or how to understand our initial observations; if the case were clear, there would be no need to do any hermeneutic work. On the other hand, barn facades and puddle mirages 1
Such examples may not be convincing to a ‘mad Quinean’, who is willing to insist that since puddles and barns aren’t reducible physicalist notions, we don’t observe them either, and who allows only observations of patterns of extroceptor stimulation. I will not concern myself with this madwoman here.
How to Get an Interpretivist Committed
209
aside, we rarely have to call our everyday observations of merely material objects into question. However, this is not so much a disanalogy between the types of observation, as an effect of the explanatory purposes for which the philosophical interpretivist’s account was developed. The interpretivist explains how to attribute meaningful states and epistemic agency in situations where little or no shared means of expression can be assumed, and hence substantive hermeneutic work is involved – not because we usually need to do this work, but in order to reveal the constitutive conditions of interpretability and the structure of interpretation. In everyday situations, we are perfectly capable of simply observing, without interrogation, that someone is trying to do something, claiming something, insisting upon something, etc. While such observations are always fallible, such fallibility becomes an issue only when we are dealing with an agent or a situation which tests the boundaries of our established constitutive and mundane skills for observing and coping with agents – an agent whose behaviour is to some extent anomalous or alien and who thus demands hermeneutic work out of us. But in this way, the situation is no different from one of physical observation. Normally, the fallibility of our physical observations is not important and does not shake our basic faith in what we see. We call these observations into question and begin a process of interpretation when our observations disclose phenomena that play in or near the excluded zone, and hence demand our entry into a hermeneutics of observation. This return to Haugeland’s terminology allows me to segue into a critique of Haugeland’s own discussion of how his account fits with interpretivism. Haugeland, of course, believes that a particular structure of commitments of various sorts is necessary for epistemic agency. However, he is explicitly not interested in using this result to rewrite interpretivism. On the contrary, he believes that his account distances him from any form of interpretivism (such as either Davidson’s or my own) that wants to make interpretability into a constitutive condition for agency, and he declares himself to have no special opinions or interest in methodological questions concerning the interpretation and attribution of agency from a third-person perspective.1 He claims that in order to make commitment show up, we need to take up a first-person perspective, because only then can we notice how ‘first-person involvement’ – rather than any third-personally available behaviour of an agent – is necessary to the disclosure of commitment.2 He does not deny that we may be able to tell from a third-person perspective, using indirect evidence, when an agent is 1
TRF 340.
2
TRF 339.
210
Rebecca Kukla
involved in the right way, but he argues that ultimately, committed involvement “gets lost when one restricts oneself to what is outwardly observable”.1 Now I am not sure what he means by saying that such features are first-personally (but not necessarily psychologically) available, but I suspect that he is under the spell of the reductionist’s picture, in which only physicalist facts count as third-personally or ‘outwardly’ accessible – a picture which I think is a particularly bad fit with the rest of his account. I have argued that the facts about someone’s involvements and commitments are perfectly good (though not naturalistically reducible) behavioural facts2 observable from an external perspective, and this argument is only reinforced by Haugeland’s rejection of psychologism. I have also argued that the observation of these facts is not particularly indirect – though the facts are holistically determined, what we are determining, in observation, is such behaviours themselves, rather than something else, such as some ‘internal caring’ inferable from outward behaviour. If Haugeland simply means that in order to be committed, involved, etc., one must have a first-person perspective on things, then this seems clearly right but not particularly remarkable. The Davidsonian interpretivist would certainly agree that being an epistemic agent requires abilities which in turn require a first-person perspective and could not be had by a perspectiveless physical object, such as having beliefs, understanding their possible falseness, wanting to communicate, and so forth. Haugelandesque involvement and commitment cannot be reduced to states that make no intrinsic appeal to a first-person perspective, but the same is true of Davidsonian belief and assent, and in neither case does this prove that these things are not available to be observed from a third-person point of view. Now Haugeland “make[s] no claim as to whether or how the presence and significance of such a feature in some activity could be established by radical interpretation”.3 On the other hand, he brings up the point about first-person involvement in the context of a discussion of radical interpretation, where he frames himself as departing from Davidsonian dogma in a way that I do not understand. His conclusion is that whether or not radical interpretation can detect such features is ‘beside the point’, because his argument concerning the irreducible role of the first person shows than the constitutive mark of epistemic agency is the right kind of first-person involvement, as opposed to 1
Ibid. He rightly reminds us, however, that this does not mean that such commitments are necessarily or only accessible from an introspective, phenomenal point of view.
2
Here, clearly, I specifically do not want only naturalized bodily movements to count as behaviour, though this is sometimes how the term ‘behaviour’ is used. Behaviour, for my purposes, is simply material, public action.
3
TRF 340.
How to Get an Interpretivist Committed
211
interpretability.1 This move seems wrong to me for at least two reasons, given the rest of Haugeland’s account. First, as I have already pointed out, his comments about first-person involvement do not contradict Davidson’s account, and this means that Haugeland has not given us any reason to reject Davidson’s many famous and compelling arguments in favour of the constitutive role of interpretability in epistemic agency. But the second and more important point is that there are features of Haugeland’s own account which not only fail to exclude the constitutive role of interpretability, but actually compel us to accept it. Furthermore, because of this compulsion, an upholder of Haugeland’s project has every reason to worry about the type of thirdpersonal evidence that could ground such interpretability, so that this issue is not ‘beside the point’ at all. If I am right about this, then Haugeland ought to be much more concerned with revisiting the conditions of interpretability in light of his account than he claims to be. The point is simply this. Presumably, epistemic agents are themselves not exempt from Haugeland’s story about the constitution of objects. Hence objective knowledge about beings like us must, like any objective knowledge, be governed by constitutive standards.2 But this means that beings like us are constituted through the enforcement of constitutive standards, as we are coped with by means of constitutive and mundane skills. Therefore, we can ask what the constitutive standards for being an epistemic agent are, which in turn, on my account (or an enriched Davidsonian account) just is a matter of asking about the standards of interpretability; Haugeland’s own explanation of constitution as ‘making sense of’ certainly suggests this translation into the language of interpretation. Likewise, since we are constituted by these constitutive standards, and fail to be constituted by them if we fail to live up to them, we can conclude that our interpretability is constitutive of our epistemic agency. In fact, when Haugeland describes how agents enforce constitutive standards through the exercise of constitutive and mundane skills, he is describing the constitutive conditions for the existence of the objects they constitute, but he is also, whether he considers this explicitly or not, giving us an account of the constitutive standards for being an epistemic agent in the first 1
Ibid.
2
In fact, our understanding of what Haugeland means when he says that we must take responsibility for our constitutive standards and for explaining away excluded phenomena seems derivative upon our understanding of how this works in the case of understanding agents; it is clear to us why our standards for personhood must be right, why we have an ethical responsibility to make sure they are, and why we need to care when an apparent agent acts in ways that appear not to make sense. Even our understanding of what it means for an object to have authority and make a claim on us seems derivative upon our understanding of how other agents can do this.
212
Rebecca Kukla
place. That is, insofar as some being (including myself) is disclosed to me as an epistemic agent, this happens through my successfully interpreting her (through the use of my constitutive and mundane skills) as being a being who herself has such skills, enforces constitutive standards, takes responsibility for the emptiness of the excluded zone, is responsive to the authority of objects, and the like. My commitment to understanding other agents in this way manifests itself not just as a passive belief that I can do so, but as an active project of holding others to these standards, excluding the possibility that they fail them, and demanding that they do not demonstrate an unacceptably radical irrationality or insouciance about their own standards. Understanding someone as an epistemic agent involves at least as much commitment to making sense of her as does understanding any phenomena in a determinate way. Furthermore, whether I am holding myself to constitutive standards and hence interpreting myself, or whether I am interpreting another, as the interpreter I must in effect take a ‘third-personal’ stance towards the interpretee. This is because in order to count as applying constitutive standards to objective phenomena, the phenomena being interpreted must be independent of the interpreter, in Haugeland’s sense of being separate from and capable of resisting the interpreter. Thus the ‘first-person involvement’ that Haugeland wants to make essential to epistemic agency must in fact show up from a third-personal point of view, in order for it to impose itself authoritatively upon an observer. The constitutive first-personality of agency thus comes to nothing more than the fact that some of the claims about an agent that must be made and supported from a third-person point of view must be irreducibly about the first-personal stance of that agent, as they surely are in any roughly Davidsonian interpretation. Thus I would say that Haugeland needs to be at least as committed to the constitutive role of interpretation in epistemic agency as is Davidson. Likewise, whether or not there exists evidence sufficient for interpreting agents as agents, and what form this evidence should or could take, ought to be questions that concern Haugeland as much as Davidson. For unless there exist enforceable constitutive standards for making sense of epistemic agents as such, agency would not get to count as an objective phenomenon at all, given Haugeland’s theory of objectivity. In fact, one might even say that, far from being ‘beside the point’, interpretability counts as a constitutive standard of existence for all objective phenomena, on his account. However, in order to be convinced by this, we need to understand ‘interpretation’ in the broadened sense I have been developing, namely the as process of making sense of normatively governed behaviour in general, rather than just making sense of the content of truth-claims. If Haugeland understands inter-
How to Get an Interpretivist Committed
213
pretation merely as the process of matching contents with sentences, then he will not consider it constitutive of epistemic agency. But we can reclaim the notion of constitutive interpretation once we insist that even for Davidson, interpretations took commitments as their starting point and subject matter, and once we properly enrich our understanding of the commitments at stake. This section has been devoted to an analysis of how we observe agency. But of course, Davidson’s own analysis of this does not focus upon everyday observational situations, but upon the highly specialized, idealized case of radical interpretation. I will end this section by revisiting the conditions, possibilities and methodology for radical interpretation, given my arguments so far. Remember that while the Davidsonian radical interpreter cannot presuppose the meaning of any of an agent’s sentences, she can and must presuppose a provisional, fallible ability to recognize assent behaviour, since she needs some initial reason to hypothesize that the being in front of her is an agent who is making truth claims. But I have argued that there is no reason to take assent behaviour as more provisionally accessible than other types of normatively governed behaviour, and that in fact it is only accessible in the context of this other behaviour. Thus I would claim that the radical interpreter can and must presuppose the ability to provisionally recognize a whole variety of committed behaviour, rather than giving some sort of foundational privilege to assent behaviour. These recognitions, including the recognitions of assent behaviour, are just what will be justified, perhaps falsified, and given content during the process of interpretation; the justification and interpretation of assent behaviour will be necessarily intertwined with the justification and interpretation of behaviour manifesting other types of epistemic commitments. One reason for why interpretivists may not have focused on the plethora of commitments needed to be a truth-claimer is connected to a fact which one might try to turn into an objection to my argument – namely, the fact that most of the time, someone’s status as a committed being of the right sort to be capable of epistemic agency is completely obvious to us. Just as I earlier pointed out the quotidian obviousness of committed behaviour in order to forestall an objection that such behaviour was not naturalistically observable, I also think that this obviousness has gotten in the way of our ability to see the central role of such behaviour in epistemic agency. Generally, we can just tell that someone is the kind of being who can make claims, in part because we can just tell that they care when they are wrong, that they try to explain anomalies, and so on. Thus someone might argue that we don’t have to explicitly interpret any of the committed behaviour I am discussing in order to notice assent, and hence that the interpretivist can turn her attention specifically to the determination of the intentional contents of those assents. I am certainly not
214
Rebecca Kukla
proposing that, during the everyday processes of understanding others and taking them as agents, we must actively check whether they have the right structure of commitments to count as claim makers in the first place. Most of the time, such questions don’t even come under scrutiny. However, most of the time, questions about speaker’s meanings don’t come under scrutiny either. In everyday life, where situations requiring radical interpretation are rarely (or perhaps never) encountered, we take for granted huge swaths of facts about what another’s commitments are and how she manifests them. These include facts about her beliefs and ways of expressing them, as well as her practical epistemic commitments (as well as her non-epistemic commitments to avoiding pain, eating tasty food, finding true love, and all the rest of the obvious stuff). Davidson’s interpretivist story was never designed to describe the hermeneutics of everyday life – we are not supposed to design a truth-theory from ground zero each time we interact with someone. The account of radical interpretation is designed to describe an idealized situation in which we have no knowledge of the meaning of another being’s behaviour or utterances, and the purpose of the exploration of this idealized situation is to reveal the constitutive conditions of the assignment of agency and the fixation of meaning. We only need to make the interpretation of commitments explicit when there is some problem or breakdown in transparent communication. The theory of radical interpretation is designed to make explicit what goes on implicitly and unproblematically during normal interaction, and to fill in the legitimizing story about what would have to happen if our alreadypresent expectations about what people mean and believe, which are almost always justified, became problematized. To say that what we interpret are commitments in general, not just holdings-true, is not to say that we need to engage in some special new interpretive work. It is to point out, rather, that we have already been doing much more than Davidson’s story reveals when we observe others and recognize them as epistemic agents and wielders of meaning. Davidson tells a story in which we begin with the observation of assents, and engage in a process of interpretation which involves fixing the contents of those assents from scratch, but not attending to any other committed behaviours. I claim that this story is insufficient either as a story about radical interpretation, or as a story about what we actually do during our everyday hermeneutic encounters. It is an insufficient story about what we actually do – a story which it was never intended to be – because generally, we don’t have to start assigning meanings to beliefs and actions from scratch. It is an insufficient story about radical interpretation, because, firstly, our initial, provisional observations of assents are dependent upon our initial, provisional observa-
How to Get an Interpretivist Committed
215
tions of other commitments, which provide a context allowing assents to be disclosed as assents in the first place, and, secondly, the justification of our initial attribution of assents cannot occur merely by matching them with true, appropriate contents, but must also proceed by holistically interpreting all of an agent’s committed behaviour in order to make good on our initial hypothesis that we are dealing with an epistemic agent. Davidson has written that his “idea is to take truth as basic and to extract an account of translation or interpretation”1 His arguments against the need for a theory of truth are well-known,2 and I will not rehearse or evaluate them here. We can assume here that we have an unproblematic understanding of what it means for a sentence to be true. But Davidson takes not only truth but holding-true to be methodologically basic.3 He needs to do this in order to support his methodological story, which places assent at the foundation of interpretation. But the basicness of truth, even if we grant it, does not automatically yield the basicness of holding-true, which is a committed attitude that requires the ability to have the right sort of normative relationship to independent objects. My claim is that holdings-true are very special sorts of commitments, which are neither basic in the sense that they form the bedrock of other commitments, nor in the sense that we can recognize them independently of or prior to our recognition that an agent has other commitments as well, in the context of which she holds things true. It is unclear that Davidson has any non-circular reasons to assign these holdings-true the privilege he needs them to have. They might have some special privilege if, as Davidson sometimes suggests, we could justify our ascription of them merely by correlating an agent’s utterances with appropriate true contents, thereby flattening any gap between interpretive evidence for fixing content and interpretive evidence for ascribing belief. But not any being that consistently utters what are interpretable as truths counts as an epistemic agent; the being must also be committed to those truths in just the right way. Davidson’s problem, as I have portrayed it, is not so much with accounting for how sentences that express claims manage to be true, but about how we can be justified in taking things as expressions of objective claims in the first place. It is in determining this that I have urged that we must attend to the entire structure of commitment which binds an agent, rather than focusing merely on fixing the contents of her sentences and taking that as a self-standing interpretive project. Haugeland, we 1
“Radical Interpretation” (1984a), 134.
2
See “The Folly of Trying to Define Truth” (1996a), “The Structure and Content of Truth” (1995).
3
For instance, see “Radical Interpretation” (1984a), 135.
216
Rebecca Kukla
saw, concludes that Davidson is wrong to take interpretability as a constitutive standard of belief. I think, instead, that we should enrich our account of interpretability so that it can serve as such a standard.
Living With the Threat of Skepticism: Kantian Ruminations I want to end this paper with a return to one of the cornerstones of Davidson’s philosophy, namely the rejection of the threat of skepticism. We know that Davidson’s account rules out radical skepticism – the possibility that any individual or language community is substantially wrong about the nature of the world. For Davidson, this result is built into interpretivism. Since beliefs are fixed by causal contact with the world, together with interpretations which optimize truth and rationality, there is no way for them to be radically out of step with this world. Now, Haugeland would have to agree with Davidson that an epistemic agent whose beliefs are radically wrong is an impossibility. For Haugeland, being an agent requires having normative relationships with objects wherein we let objects disclose themselves; but this means that such genuine disclosure is a constitutive condition of epistemic agency, and hence that if a being has contentful beliefs at all, they cannot be radically wrong. So far, the accounts are in accord with one another. However, important differences between them show up when we examine their implications more carefully. Remember that for Haugeland, epistemic agency requires treating excluded events as impossible, and taking responsibility for their non-occurrence. These excluded events are precisely the ones that, if they occur, could make us radically wrong about the world. Luckily, they are by definition impossible, and hence they will not occur. Yet, while impossible, we know that they are not epistemically irrelevant for Haugeland. They are the events whose exclusion we must take responsibility for maintaining, and the threat – though not the fact – that they may occur is what makes our basic constitutive standards threatenable. The fact that the correctness of our constitutive practices is threatened by the conceivability of excluded events is what makes these constitutive practices count as a set of practices for objectively disclosing phenomena, rather than a set of mere non-epistemic conventional behaviours. But these constitutive standards determine what counts as an object and how objects behave – i.e. they determine the very things that are beyond the reach of skeptical interrogation for a Davidsonian. Therefore, Haugeland would have to grant the following result: At one and the same time, we must acknowledge that our most basic constitutive standards cannot in fact be wrong, and likewise that we cannot be radically wrong about the nature of the objec-
How to Get an Interpretivist Committed
217
tive world, and we must acknowledge that the threat that they are wrong, and the corresponding responsibility for making sure that they are right, must genuinely govern our practices. Epistemic agency requires taking responsibility not only for our particular claims, but for the correctness of our standards, while facing the conceivability of these standards falling apart. This higherlevel responsibility is “the ultimate origin or source of power of the excluded zone, hence of objects. Without this power – the power to bring the whole thing down – the bite of the excluded zone would be ultimately toothless”.1 Thus, while both Davidson and Haugeland will agree we cannot be radically wrong about the world, the status and functioning of the ‘cannot’ is very different within the two accounts. Whereas for Davidson, this is a straightforward, theoretical fact about the world, which need not itself function or have weight in our epistemic practices, Haugeland would have to say that this has the weight of an imperative of practical reason. In his case, this has to be the ‘cannot’ a responsibility, indicating a practical burden. The impossibility of being radically wrong is not just a truth – it has to be a normative issue for us in order for us to be epistemic agents at all. In other words, the threat of being radically wrong must actually govern our epistemic practices, and we must take responsibility for this threat at the level of practical reason. This is so even though we already theoretically know that we are not in fact radically wrong – that it is impossible for this threat to be realized, since the general correctness of our standards is the precondition for our encountering the world at all. There is no empirical possibility of the standards being wrong, since the standards constitute the existence of the empirical world. Haugeland does not worry about this necessary threat to our mundane objective contact with the world, nor about the status of radical skepticism more generally. I believe that this is because all of his discussions of constitutive standards and excluded zones concern specialized, highly limited domains of objects, such as that of some special science. The constitutive standards of these limited domains can and often genuinely do need to be radically revised or scrapped in the face of persistent anomalies. However, Haugeland purports to be giving an analysis of objectivity in general, and hence, presumably, his story must apply to the domain of everyday, midsize empirical objects,2 and other domains so fundamental and pervasive that if we turned out to be wrong about them, we’d have to do more than switch research programmes – we’d have to admit radical wrongness about the nature of the world. Our epistemic 1
“Original Intentionality” (1998), my italics.
2
Indeed, near the end of TRF, he offers a tentative analysis of the constitutive standards for such objects.
218
Rebecca Kukla
dealings with everyday objects cannot turn out to be less objective than are the more specialized dealings, otherwise we have lost contact with an authoritative world all over again, at a more global level – the only level that ultimately matters. Nor can Haugeland say that in this everyday case, we don’t need a story about threatenable constitutive standards to get objectivity off the ground, since he offers the story as an exposition of the nature of objectivity itself. And yet, it is hard to pin down the sense in which the constitutive standards for everyday objects count as threatenable. The very pursuit of the project of understanding and coping with the everyday world, and making meaningful claims about it, already proves the success of the constitutive standards which are the transcendental condition for the disclosure of this world. In the case of specialized domains, we have plenty of background standards through which to encounter an anomalous phenomenon, though not, perhaps, as a phenomenon in that specialized domain. For instance, a phenomenon which is anomalous with respect to particle physics can still be experienced as a reading on an instrument, and a phenomenon which is anomalous with respect to biology could still be experienced as a weird moving fleshy thing. But a phenomenon which genuinely violated the standards for being an everyday objects could not show up as an empirical phenomenon at all. (And here, it is appropriate to notice, we are beginning to tread firmly on Kantian territory). So what substantive form could the threat of the failure of these standards take? I think that we need to understand the threat of skepticism as something which must practically govern our actions, but we must not understand the governing force of this threat as grounded in a real empirical possibility. In order to be epistemic agents in objective contact with the world, we must be responsible for our constitutive standards, and this responsibility is essentially accompanied by a practical imperative to protect our standards against the threat of refutation, and to be willing to give them up should they be refuted. This practical imperative functions as a necessary normative principle which regulates our epistemic practices. We must be committed to and bound by this principle in order to count as having objective standards at all. But the legitimate grip of this principle is not, and cannot be, grounded or determined by the empirical facts of the matter. For, on the one hand, the facts will always reconfirm our basic standards and the impossibility of radical wrongness. And on the other hand, the disclosure of empirical ‘facts of the matter’ already presupposes the very objectivity which has the grip of this principle as a condition for its possibility. Instead, the principle’s legitimacy stems from the very structure of objectivity and epistemic practice themselves, insofar as it is a transcendental condition for their existence – it thus has a grounding which
How to Get an Interpretivist Committed
219
comes as close to being a priori in the traditional sense as anything philosophically fashionable does these days. The threat of skepticism, and the responsibility for this threat, do have practical content even though the threat cannot be empirically realized. This is because even though it is impossible that our constitutive standards for mundane objects and phenomena will be empirically overthrown, the rightness of the standards is contingent in the sense that there is no necessary guarantee, in advance of the disclosure of the world, that an objective world that we can cope with should disclose itself to us. affiliation I believe that my account here of both the empirical impossibility and the practical threat of radical wrongness are substantially true to Kant’s account of the transcendental necessity of the harmony between our cognitive faculties and the world.1 In a subtle move, begun in the First Critique but spelled out fully only in the Third Critique, Kant explains that our belief in this suitedness of our faculties to the world (and vice-versa) is necessary. This belief cannot turn out to be wrong, since every experience will in fact verify it, in virtue of the fact that this harmony is a transcendental condition for the possibility of any experience at all. Radical misunderstanding of the objective world is not a possibility, for Kant. But at the same time, our belief in this harmony, precisely because of its transcendental status, cannot be based in experience. Nor can it be proven a priori, independent from actual experience, for there is nothing about our contingent nature that conceptually guarantees this fit with the world.2 So what sort of necessity does this belief have? As becomes progressively clearer throughout Kant’s critical works, this necessity is not that of a provable truth, but of an unavoidable demand. We must insist upon the suitedness of the world to our faculties, in order to let the world disclose itself. We must take this suitedness as a governing assumption that lies at the basis of every encounter with the world. But this assumption is not just a passive hypothetical, but one for the truth of which we must take responsibility: So if we say that nature makes its universal laws ... in a way commensurate with the human understanding ... then we are neither prescribing a law to nature, nor learning one from it by observation (although observation can confirm the aforementioned principle). For it is a principle not of determinative but merely of reflective judgment. We insist only that, however nature may be arranged in terms of its universal laws, any search for its empirical laws should follow this principle of purposiveness... This harmony of nature with our cognitive power is presupposed a priori by judg1
See, in particular, the B-version of the “Transcendental Deduction” in the Critique of Pure Reason.
2
Critique of Judgment, ac. 195.
220
Rebecca Kukla ment... Without presupposing this harmony we would have nothing to guide us in using empirical laws so as to experience and investigate nature...1
We demand of the world that it be suited to our epistemic capacities, and of ourselves that we live up to its claims. To demand the disclosure of the world we must be able to conceive of the possibility that this demand fails, otherwise we rob the demand of its normative bite. Thus the demand that our constitutive standards be right, and the practical responsibility for the possibility that they may not be, are two sides of the same coin. We must be practically concerned with the threat of epistemic collapse, not because anything could provide determinate evidence of such a collapse, but because we can and must understand the conceivability of radical wrongness, much as we must understand the conceivability of the excluded zones of more specialized domains, and because it is only our concern with this conceivable wrongness which allows our standards to be objective. Hence, while we do not live with the theoretical possibility of radical wrongness, we do live with the practical burden of its possibility. This is not an empirical possibility, but a governing limit point which gives content to our commitment to truth. Thus, in Kantian terms, the conceivability of radical wrongness is an idea of practical reason, not an empirical concept. It is this practice of taking responsibility for this limit point of radical wrongness which allows the pursuit of truth to be a project. Concern with radical skepticism is thus appropriately taken up as an ethical attitude, not a theoretical worry. We are responsible for treating the task of correctly disclosing the objective world, and that of vindicating the legitimacy of our epistemic practices, as intertwined projects at risk of radical failure. This responsibility is a condition for the possibility of a genuinely epistemic relationship to the world in the first place.
References Davidson, D. (1980): “Mental Events”, in Essays on Actions and Events, Oxford University Press, Oxford. Davidson, D. (1984a): “Radical Interpretation”, in Inquiries into Truth and Interpretation, Oxford University Press, Oxford. Davidson, D. (1984b): “Thought and Talk”, in Inquiries into Truth and Interpretation, Oxford University Press, Oxford. Davidson, D. (1986): “A Coherence Theory of Truth and Knowledge”, in Truth and Interpretation: Perspectives on the Philosophy of Donald Davidson, ed. E. Lepore, Basil Blackwell, Oxford. Davidson, D. (1990): “The Structure and Content of Truth”, Journal of Philosophy 87:6, 279-328. Davidson, D. (1995): “Pursuit of the Concept of Truth”, in On Quine: New Essays, ed. L. Paolo, Cambridge University Press, New York. 1
Critique of Judgment, ac. 186-7, my emphasis.
How to Get an Interpretivist Committed
221
Davidson, D. (1996a): “The Folly of Trying to Define Truth”, Journal of Philosophy 93:6, 263-278. Davidson, D. (1996b): “Subjective, Intersubjective, Objective”, in Current Issues in Idealism, ed. P. Coates, Thoemmes, Bristol. Davidson, D. (1997): “Seeing Through Language” in Thought and Language, ed. J. Preston, Cambridge University Press, New York. Haugeland, J. (1997): “Truth and Rule-Following”, in Having Thought, Harvard University Press, Cambridge, MA. Haugeland, J. (1998): “Originary Intentionality”, talk given at Carleton University, November 1998, Ottawa, Canada. Heidegger, M. (1996): Being and Time, trans. J. Stambaugh, SUNY Press, Albany, NY. Hegel, G. W. F. (1977): Phenomenology of Spirit, trans. A. V. Miller, Oxford University Press, Oxford. Hornsby, J. (1997): Simple Mindedness, Harvard University Press, Cambridge, MA. Kant, I. (1929): Critique of Pure Reason, trans. Norman Kemp Smith, Macmillan, London. Kant, I. (1987): Critique of Judgment, trans. W. S. Pluhar, Hackett Press, Indianapolis. Manning, R. (1998): “All Facts Great and Small”, Protosociology 11, 18-40. McDowell, J. (1994): Mind and World, Harvard University Press, Cambridge, MA.
222
David Pitt
DAVID PITT Nativism and the Theory of Content Abstract Externalism is the view that the intentional content of a mental state supervenes on its relations to objects in the extramental world. Nativism is the view that some of the innate states of the mind/brain have intentional content. I consider both “causal” and “nomic” versions of externalism, and argue that both are incompatible with nativism. I consider likely candidates for a compatibilist position – a nativism of “narrow” representational states, and a nativism of the contentless formal “vehicles” of representational states. I argue that “narrow nativism” is either too implausible to appeal to the nativist – because it entails that innate representational states are lost as the mind becomes more experienced, or too costly to appeal to the externalist – because a reasonable version of it requires the analytic-synthetic distinction. Finally, I argue that “syntactic nativism” is indistinguishable from classical anti-nativist empiricism, given the latter’s broad tolerance for innate implementation of psychological principles and mechanisms.
Recent controversy over psychological externalism has centered on the issues of its compatibility with privileged access and the a posteriority of empirical knowledge.1 In this paper I wish to add to the controversy by raising another issue for externalism, viz., its compatibility with an important and widely held version of the innateness hypothesis. I argue that externalism does not, on generally accepted terms, sit comfortably with a plausible and philosophically interesting nativism. Given the existence of influential arguments for both externalism and nativism,2 I take it a serious tension between them would be a matter of some concern. If I am right, one or another of several popular philosophical doctrines about language, thought and meaning would have to be significantly revised – or abandoned.
1
See, e.g., Boghossian 1989, Brueckner 1990 and 1994, Heil 1988 and 1992, McKinsey 1991, Falvey and Owens 1994, Brown 1995.
2
Viz., Putnam 1975 and Burge 1979 for externalism, and Chomsky 1965 and Fodor 1975 and 1981 for nativism.
Nativism and the Theory of Content
223
1. Externalism and Nativism Externalism is the view that the intentional content of a mental state supervenes on its relations to objects in the extramental world.1 The version of externalism I consider in this paper is the informational theory of content. I focus on this view both because I take it to be, for present purposes, representative of externalism generally, and because I think it is best equipped to cope with the pressures I exert.2 The informational theory is reductive. It says that having content is being informationally (i.e., causally or nomologically) related to extramental objects. On such nonreductive externalisms as McDowell’s or McCulloch’s (McDowell 1986, McCulloch 1995), for example, the content-conferring relation between thoughts and their objects is inclusion (the mind and its contents are not within the individual). Yet, on both of these non-reductive theories, being in informational relations with non-mental objects is a necessary condition on being in a contentful state. Thus, McDowell makes essential use of the notion of acquaintance (typified by perception) in his account of object-dependent thoughts (op. cit.: 138-41); and McCulloch adverts to “embodied interactions” between thinkers and non-mental objects (op. cit.: 202). It is hard to see what such relations as acquaintance and embodied interaction could be, if not causal. On either version of externalism, then, informational relations between a thinker and the objects of his thought determine which thought he is thinking, where thoughts are individuated by their contents.3 I consider, in particular (again because I take them to be both representative and best equipped), two versions of the informational theory, which I call (following Fodor 1990) the causal theory and the nomic theory. According to causal theories, the content of a mental state type is determined by what has caused its tokenings. On Fodor’s version (Fodor 1990: 119-122), for example, a concept-type expresses the property f if its tokenings have been reliably causally correlated with instantiations of f (if, that is, they have been caused by 1
Internalism is the view that mental content supervenes on an individual’s non-relational properties or on intramental relations.
2
I take informational theories to include teleofunctional theories, such as those advocated in Millikan 1984 and Papineau 1987. Teleofunctions are supposed to solve the disjunction problem (Millikan 1993, 7; Papineau 1987: 63-64; Fodor 1990: 64ff), which arises specifically for theories on which content-conferring relations are causal-informational. I will not explicitly discuss teleofunctional theories in what follows, since the problems I will raise concern what’s in the head, not how it got there, or which sorts of causal-nomic connections to external objects determine content, or how it is that only one among a state’s many informational relations to external objects is content-constitutive.
3
I am grateful to Tim Crane for comments that led to clarification of my position here.
224
David Pitt
Fs), and if its non-F-caused tokenings are counterfactually dependent on its Fcaused tokenings. According to nomic theories, the content of a mental state type is determined by what has or would cause its tokenings. On (again) Fodor’s version (ibid.: 100-101), concepts have their contents in virtue of their counterfactual relations to property instantiations. This version allows for concepts that express uninstantiated properties as well as instantiated ones. One has the concept MERMAID,1 for example, in spite of never having been suitably situated with respect to one, if one is so constituted that, were mermaidhood to be instantiated in the right sort of way in one’s vicinity, something or other (some dedicated something or other) would go off in one’s head; and, excepting that mermaids would bring this about, nothing would. (The mental representation type whose tokening is counterfactually, though not actually, connected to the instantiation of mermaidhood is the concept MERMAID.) Nativism is, broadly speaking, the view that the mind/brain is innately endowed with structures that predetermine in some way or other the cognitive capacities humans are capable of exercising. As a purely general point, such structures are either representational (i.e., have intentional content), or not. In the present context, the relevant version of the nativist thesis is one on which some of the innate states of the mind/brain have intentional content (of the kind appropriate to concepts). This sort of view, which I call “Semantic Nativism,” has been at the center of traditional and contemporary controversy over innateness in the philosophy of mind and epistemology. Philosophers have been interested in whether or not there are concepts (“ideas”) humans are capable of entertaining that are unlearned (especially, concepts which might enable us to think thoughts with non-empirical, or non-experiential,2 content). In any case, it is clearly Semantic Nativism that poses a prima facie problem for the externalist. I will, however, also consider another sort of thesis, on which any innate states of the mind/brain are not representational, but merely structural, as possibly affording a philosophically interesting nativism compatible with externalism. On this view, which I call “Syntactic Nativism,” it is
1
In this paper I use capitalized expressions to denote concepts, italicized expressions to denote properties, and single-quoted expressions to denote the expression types so quoted.
2
I take the difference to be this: empirical concepts are of empirical objects, properties, states of affairs, etc.; experiential concepts can be acquired on the basis of experience. Note that empirical concepts needn’t be experiential (as for us, I suppose, the concept SUPERSTRING is not).
Nativism and the Theory of Content
225
only the “syntactic vehicles” (to use Fodor’s term) of concept-sized representational states that are innate.1 The relevant (again, for present purposes) construals of the innateness of a mind/brain state are (1) its possible presence (“tokening”) in the mind/brain independently of actual interactions with the world (experience) and (2) its possible presence (tokening) in the mind/brain independently of actual and potential interactions with the world. On the first (weaker) construal, a state is innate iff it can be tokened in the mind/brain even if it has not been caused to be tokened by experience. On the second (stronger) construal, a state is innate iff it can be tokened in the mind/brain even if there are no conditions under which it has been or would be caused to be tokened by experience. Since it does not matter to my arguments which way innateness is construed (see section 2.1), I will not try to decide here on one of these characterizations. I will also not be concerned with the question of whether or not any concepts are innately entertainable. Concepts need not be innately available to their possessor to be innately possessed; a nativist can allow that experience may be necessary for a mind to begin to think. In the following section, I consider the compatibility of externalism and four forms of Semantic Nativism. In section 3, I consider the compatibility of externalism and Syntactic Nativism.
2. Semantic Nativism 2.1. The Conflict The causal theory claims that content is determined by past interactions between the mind/brain and the rest of the world. The incompatibility here is obvious. If the nativist thesis is that some of the innate states of the mind/brain are intentional, then any view on which intentional content is acquired through causal interaction with the world – presumably through the senses (i.e., by experience), will be ipso facto committed to anti-nativism. Indeed, it seems the very idea of causal theories that content is not intrinsic to the mind. Though most information theorists are committed to the nomic version of their theory,2 Fodor (1990: 119-122) has identified some reasons for preferring the causal theory. On the nomic theory, you cannot have different concepts for properties it is nomologically impossible for you to distinguish. This, 1
What I mean by ‘merely structural’ and ‘syntactic vehicle’ will be explained below (see note 27).
2
See, e.g., Dretske 1981 and 1986, Fodor 1987 and 1994, Stalnaker 1984 and Stampe 1977.
226
David Pitt
according to Fodor, amounts to a form of verificationism, which one may well wish to avoid. On the causal theory, in contrast, if a mental representation is, as a matter of fact, caused to be tokened by instantiations of only one of such properties, then that representation expresses the property instantiations of which have caused its tokenings. (In other words, if Fs and Gs would both cause tokenings of a concept C as a matter of nomological necessity, then on the nomic theory C must express the property f or g, whereas on the causal theory it is possible for C to express only f, if only Fs have actually triggered its tokenings.1) Anyone impressed by this argument and sufficiently wary of verificationism must contend with the obvious inconsistency of nativism and the causal theory. Whether or not there is a tension between nativism and nomic externalism might seem to depend on how innateness is construed. If innateness is independence of actual causal transactions between the mind/brain and the rest of the world, then there is no incompatibility (mental states can possess nomic content independently of actual interactions); whereas if innateness is construed as independence of nomic relations (i.e., actual and potential tokening) between the mind/brain and the rest of the world, then there is an incompatibility: mental states cannot possess nomic content independently of nomic relations. The issue between nativism and externalism would then devolve upon how innateness should be understood – an issue for nativists to debate (unquestion-beggingly) among themselves. I maintain, however, that nomic externalism is incompatible with a plausible Semantic Nativism however innateness is construed. The tension between them depends not on the distinction between actual and counterfactual tokening, but on the distinction between the representation (i.e., expression) of content and the implementation of content. Mental contents are generally agreed to be abstract (non-spatiotemporal) objects – properties, relations and propositions, most often construed as functions. For a mental state to have a function as its content is for that state to represent (express), and not just implement, that function. The distinction between representation and implementation is nicely made by an example in Fodor 1987. The planets in their orbits implement Kepler’s laws of planetary motion (roughly, functions from planets to trajectories) – that is, they behave in accordance with them; but (obviously) they do not represent them. The 1
Though there is, Fodor notes, some residual verificationism even here: on the causal theory you cannot have distinct concepts for indistinguishable properties instantiations of all of which you have encountered.
Nativism and the Theory of Content
227
general moral is that for a function to be represented it is not sufficient that it be implemented.1 The moral in the instant case is that whether or not the nomic theory is compatible with a Semantic Nativism depends on whether or not what it places in the head is sufficient for representation (as opposed to mere implementation). And here the nomic theorist is faced with a dilemma. Either he allows that what is in the head is sufficient for content representation, or he does not. If he does, then though he need not be in conflict with the nativist, he is no longer an externalist. For, recall, it is a fundamental tenet of externalism that the head’s contribution to mental states is not sufficient for content representation. (Molecularly identical heads in different environments represent differently; and it does not make sense to say that the state is representational, but represents no particular content.) This is, after all, the point of the Putnam and Burge thought experiments. But if what is in the head is not sufficient for representation, then, well, it is not; at best it is sufficient for implementation, and the resulting nativism is not semantic. Thus, externalism in both its causal and nomic versions is (so far), incompatible with Semantic Nativism. There are, as I see it, only two moves an externalist might make here, both of which involve opting for a weaker nativist thesis. Adopting a distinction between “narrow” and “wide” content, an externalist can argue for a nativism on which only narrow concepts are innate. Alternatively, an externalist could adopt Syntactic Nativism, the view on which only the syntactic vehicles (physical bases) of concepts are innate. In the next two sections I argue that neither of these alternatives is likely to yield a plausible and philosophically interesting nativism, on generally accepted terms. 2.2. Narrow Nativism There seem to be two ways of thinking about “narrow” content in the literature. Sometimes (e.g., Putnam 1975, Fodor 1994: 39ff, and Block 1986: 627ff) it is spoken of as something like de dicto (as opposed to de re) content, or character (à la Kaplan 1989), which is directly expressible.2 Sometimes (e.g., Fodor 1987) it is spoken of as radically inexpressible – as some kind of protocontent, or mere content-determinant (e.g., Block 1986). On the former construal, innate possession of states with narrow content would be incompatible with externalism, for the reasons given above. Thus, it is narrow content in the latter sense that will be at issue here. 1
Nor is it necessary: Kepler’s mind, for example, represented the laws of planetary motion without implementing them.
2
The narrow (in this sense) content of Oedipus’s desire is that he marry Jocasta; its wide content that he marry his mother.
228
David Pitt
Now I don’t think a Narrow Nativism of this sort would be acceptable to most nativists. Here, for example, is Fodor (1983: 4-5) on what the nativist wants: It [is] important to the [nativist’s] story that what is innately represented should constitute a bona fide object of propositional attitudes; what’s innate must be the sort of thing that can be the value of a propositional variable in such schemas as ‘x knows (/believes, /cognizes) that P’.
Why should the nativist insist on states with wide content? Because, according to Fodor (id.): Chomsky’s account of language learning is the story of how innate endowment and perceptual experience interact in virtue of their respective contents. ... This sort of story makes perfectly good sense so long as what is innate is viewed as having propositional content .... But it makes no sense at all on the opposite assumption.1
and narrow contents (on the relevant construal of ‘narrow’) are not propositional.2 So I think this move is actually blocked, and the conclusion of the previous section stands: externalism in both its causal and nomic versions is incompatible with Semantic Nativism. Still, an externalist might try to argue either that Fodor’s claims are false – that in fact wide contents are not needed to do the work innate concepts are supposed to do;3 or that a Narrow Nativism, though perhaps not what the nativist wanted, is still sufficiently interesting to console him in his loss of Wide Nativism. So it will be worthwhile to see what the commitments and consequences of such a nativism would be. Cross-classifying the causal-nomic and wide-narrow distinctions yields four possible theories of content, the wide causal theory, the wide nomic theory, the narrow causal theory and the narrow nomic theory.4 The wide 1
I should mention that Fodor no longer accepts this assumption – though not because he accepts a nativism of narrow concepts; but, rather, because he no longer accepts Semantic Nativism (see Fodor 1998; also cf. note 29).
2
One might also doubt that Putnam- or Burge-type cases can be constructed for the sort of logico-grammatical concepts (e.g., NOUN, CATEGORY, WELL-FORMED FORMULA) the contemporary nativist typically claims to be innate. If they cannot, then going narrow would not be an option (because the wide-narrow distinction does not apply). I believe that they can, however; see section 3.3, below, for argument.
3
Specifically, for Fodor (id.), they are needed because the interaction of experience and innate endowment is computational, and “a computation is a transformation of representations that respects ... semantic relations,” such as implication, confirmation and logical consequence. It could be argued that computations need only respect narrow semantic relations.
4
On the wide causal theory, mental state types have the same wide content in every context, and what that content is is fixed by actual causal interactions; whereas on the wide nomic theory mental state types have the same wide content in every possible context independently
Nativism and the Theory of Content
229
causal and nomic theories have already been discussed; and the narrow causal theory may be dismissed for the same reasons as the wide causal theory. So the version of Narrow Nativism we need to explore is Narrow Nomic Nativism. 2.2.1. Narrow Nomic Nativism Though it is defined functionally in terms of externalistically individuated wide content, narrow content is (or is represented by or supervenes on something) “in the head,” and so is perhaps just what the committed externalistnativist needs to fashion a compromise position. The nomic narrow content of a mental state is determined by a property of that state in virtue of which it would get caused by particular bits of the world (objects, propertyinstantiations, states of affairs, or whatever), given particular circumstances – i.e., some dispositional property. To say this much, however, is not yet to say anything a non-semantic nativist would be constrained to deny. In particular, in the present context, more needs to be said about what, precisely, is in the head in virtue of which a given mental state has the disposition characteristic of its having a particular narrow nomic content. From the point of view of such “mapping function”-style theories of narrow content, such questions may seem irrelevant: one has only to say what the constitutive conditions for (narrow) concept possession are, not how the dispositions are realized. As Georges Rey (personal communication) has put it, the mapping theorist “claims that this dispositional level of abstraction is all we ... need: semantics no more needs the details of the manner of computation than it needs the details of the particular chemicals the brain deploys in going through its computations.” But since the innateness hypothesis concerns what is actually in the mind/brain (and when), it is highly relevant to ask what sort of story one could tell about what the satisfaction of a certain function by a brain state consists in. Semantics might not need the details, but nativism does; and what’s at issue here is the compatibility of the latter with a particular version of the former. Short of taking such dispositions to be primitive, the most likely story consistent with naturalistic constraints is that they are realized by preset connections to sensory transducers/processors and to other state types – i.e.,
of any actual causal interactions. On the narrow causal theory, the wide content of mental state types may vary depending on context, and the pattern of variance (the narrow content) is set by actual causal interactions; whereas on the narrow nomic theory, the wide content of mental state types may vary with context, but the pattern of variance is set independently of any actual causal interactions. I take it the last sort of theory is what “mapping function” theorists such as Fodor 1987 and 1990 and White 1982 have in mind.
230
David Pitt
by, generally, functional, or computational, role.1 Fodor (1991: 263-271; 1994: 27), for example, says that narrow contents supervene on computational role (two narrow concepts differ in their content only if they differ in their computational role). Thus, two mentalese “word” tokens have the same narrow content only if they have computational roles that effect reliable, asymmetrical (for Fodor) connections to the same property instantiations in the same contexts. If computational roles are realized by causal relations among neural structures, then they are specifiable individualistically, and there could be mental states that have such roles innately, the roles themselves being “hardwired.”2 But is possession of states with individualistically specified functional properties sufficient for narrow content representation? The dilemma posed above for the wide nomic theorist does not apply here, since it is stipulated that narrow content is not externalist (the narrow theorist is happily impaled on the first horn of that dilemma). Our narrow nomic theorist is faced with a different dilemma, however. Computational role is either sufficient for narrow representation, or it is not. If it is not,3 then at best a state’s having a computational property is sufficient for it to implement a narrow content. But such a nativism would not be semantic. If on the other hand computational role is sufficient for narrow content, the resulting nativism, though semantic, is, I maintain, either too implausible to appeal to a nativist, or too costly to appeal to either a nativist or an externalist. The computational role of a mental state type consists in its causal relations to other mental state types. The individuation of such roles is, therefore, at least to some degree non-atomistic: they are determined by the computational “position” of the state-type in a network of causal relations. The degree of 1
Functional-role theories, which come in many varieties (cf., e.g., Block 1986, Field 1977 and 1978, Harman 1987, Lycan 1984, McGinn 1982), are the most popular theories of the internalistic aspects of mental content. I should add that teleofunctional theories, being stories only about how states acquire their content-determining functions, do not require separate consideration here.
2
Cf. Lloyd 1987, 59: What can be represented by a representational system depends on the functional interanimation of its (preexisting) representational devices. Some stimuli may affect transducers but not produce representations for want of the appropriate connections and mechanisms. In a sense, then, what can be represented by a representational system is innate, the product of the hard-wired connections available to the system as loci for representing devices. (Note that Lloyd is not claiming that functional role is sufficient for representation, but only that what it is possible to represent is limited by available functional roles.)
3
As is the case for Fodor, as noted above (computational role is only necessary for narrow content). Fodor wishes to avoid content holism (see below).
Nativism and the Theory of Content
231
non-atomism involved depends on whether the state’s position is determined by its relations to all other positions in the field (resulting in holism) or to only some of them (resulting in “regionalism”1). Let’s consider these in turn. If narrow content individuation is holistic, then the introduction of any new state type into the network changes the contents of all of its constituents. Thus, the acquisition of any new narrow concept would change an individual’s innate endowment: innate narrow concepts would either be lost, or (if this is indeed something different) change their content. If wide content is determined by narrow content together with context, the loss of a narrow content, with no context change (the situation the vast majority of us find ourselves in!), would amount to the loss of a wide content. Given the uses to which nativists typically put innate concepts, I doubt seriously they would find this result acceptable.2 A Chomskian, for example, would be faced with the prospect of a child’s losing, as it might be, the concept GRAMMATICAL upon acquiring the concept DOG. Inasmuch as innate concepts (including grammatical ones) are supposed to be fundamental in explanations of enduring cognitive (including linguistic) capacities, I take it any nativist would wish to avoid this situation. One might, on the other hand, try for a general distinction between innate concepts and acquired concepts such that innate concepts are preserved through the acquisition of new ones – because, in particular, their conceptual roles are “insulated” from those of subsequently acquired concepts (though not, perhaps, vice versa).3 But this too has costs, and ones that it seems to me unlikely most philosophers would be willing to pay. Since in order for innate concepts to figure in enduring cognitive capacities applicable to novel experiences they will have to form liaisons with acquired concepts, regionalism would require a general distinction between contentconstitutive and noncontent-constitutive computational roles. If, that is, innate concepts are to retain their content while entering into new causal-computational roles, those new roles must not figure in the determination of their content. However, most parties to this discussion would, I believe, agree that 1
Cf. Devitt’s “molecular localism” (Devitt 1996).
2
It would be at least controversial (see, e.g., Woodfield 1993, and references therein).
3
This sort of view is discussed in Sterelny 1989 (135): Sterelny entertains the possibility of a nativism on which an innate endowment of computational roles determines the content of basic concepts, from which the computational roles of nonbasic concepts are derived. The parenthetical (in the text) anticipates two possible objections, viz., (1) if acquired roles are added one at a time, the first could have no content unless the roles of the innate concepts were accessible, and (2) unless the innate/acquired barrier is permeable (from the acquired side), the contents of innate concepts will not be available to be constituents of acquired ones (a desideratum for many nativists).
232
David Pitt
the distinction between content-constitutive and noncontent-constitutive computational roles is coextensive with the distinction between analytic and synthetic inferences. If you buy the former, you buy the latter. Thus, on a computational role view of narrow conceptual content, the analytic-synthetic distinction (or its narrow ersatz) would be required to distinguish innate concepts from acquired ones. The (narrow) analytic-synthetic distinction would also be required to distinguish the content-constitutive conceptual roles of innate concepts from one another. For, if the contents of innate concepts are determined by all of their interconnections (cf. Sterelny, op. cit.), then no two minds can have any innate narrow concepts in common unless they have all their innate narrow concepts in common. But then it would be logically impossible for me to have any innate concept that you have unless I have all of them; which seems highly implausible.1 Moreover, if the innate repertoire is constant (and accessible; cf. note 21), and the acquired conceptual repertoire is based on it, then it would follow that no two people differing in as much as one innate conceptual role could think any of the same thoughts. But surely these are counterintuitive entailments. Couldn’t I be missing an innate concept – say, GOVERNING CATEGORY – that you have, but still be capable of thinking many of the thoughts that you can think – couldn’t we nonetheless both think, for example, that Madonna can’t dance? These counterintuitive consequences could be blocked if a distinction were made between content-constitutive and noncontent-constitutive conceptual roles within the innate base. The contents of innate contents would then be determined only by their relations to other innate concepts, though not by all of them. Thus, I would argue, a plausible Narrow Nomic Nativism requires acceptance of the analytic-synthetic distinction (or something very much like it). But I think this is a price many philosophers of language and mind will find prohibitive. Block (1986), in particular, holds that “C[onceptual] R[ole] S[emantics] cannot make use of the analytic/synthetic distinction” (629); and I think the consensus is still that Quine was right in holding that there is no objective (i.e., naturalistic) sense to be made of it.2 In the absence of an argument that narrow content can do for the nativist everything wide content can do, the abandonment of wide nativism in combination with the adoption of the (narrow) analytic-synthetic distinction makes Nomic Narrow Nativism a 1
Though a consequence that someone like Field (1978) seems prepared to accept.
2
Though there does seem to be a growing number who are not willing to follow Quine in this. See, e.g., Boghossian 1996, Katz, 1988, 1990, Peacocke 1992, Rey 1993.
Nativism and the Theory of Content
233
position I think most philosophers will find rather unattractive. Faced with a choice between Nomic Narrow Nativism and No Nativism, many would, I suspect, opt for the latter. A likely conclusion here would be that externalist arguments show nativism to be false. 2.3. A Hybrid View? It might be objected that all my arguments show is that one cannot comfortably be a nativist and an externalist about the same concepts, but that it is possible to be a (wide) nativist about some concepts (e.g., NOUN, AND, OBJECT), and a non-nativist externalist about the rest. So there’s a perfectly plain sense in which externalism and nativism are compatible. Clearly this is not an option for someone like the Fodor of Fodor 1981, for whom all (lexical) concepts are innate. But that sort of nativism seemed so implausible anyway (MODEM? GRUNGE?), perhaps this is just another reason to abandon it. The position of choice would be one on which there is an innate base of wide concepts (all those that are needed for, say, Chomskian reasons), while all acquired concepts are individuated externalistically.1 But this seems ad hoc. If (e.g.) a conceptual role theory can (modulo the analytic-synthetic distinction) be made plausible for some concepts, then why not adopt it for all concepts? It might be replied that the distinction is not ad hoc, but is in the nature of things: it’s just a brute fact that some concepts are not individuated externally. But if Putnam- or Burge-type cases are constructable for the concepts the nativist needs – if externalism for concepts like NOUN, SENTENCE, AND and OBJECT is motivated as much as it is for such concepts as WATER, TIGER and ARTHRITIS – then, though a conceptual role semantics might be workable for such concepts, the distinction is ad hoc. Are such cases constructable? I think they are. Since the concepts at issue here are not natural-kind concepts, the relevant sort of scenario would be Burgean. (Burge’s examples are of technical (e.g., ARTHRITIS – Burge 1979) and artifactual (e.g., SOFA – Burge 1986) concepts.) Consider someone who, as a result of somewhat limited linguistic experience in our linguistic community, believes that the category NOUN includes only terms referring to con1
The proposal in this section is different from the “insulated” view discussed above. Though both postulate a principled distinction between the innate and acquired concepts, the distinctions they postulate are different. The present suggestion is that externalism is false of innate concepts but true of acquired ones. The suggestion of the previous section was that externalism is true of all concepts, but that the narrow content of innate concept is determined only by its relations to other innate concepts, and not to acquired ones.
234
David Pitt
crete objects. Would we say that he does not have the concept NOUN, or that he had a mistaken belief about it? Compare him to his twin in a language community in which the category NOUN does include only terms referring to concrete objects. Does his twin have our concept NOUN? It seems to me that if you have the Burgean intuitions about ARTHRITIS and SOFA, you ought to have them about NOUN. It would be easy enough to construct analogous examples using concepts such as SENTENCE (can a sentence have an embedded sentential clause, or not?), CATEGORY (is RIGID DESIGNATOR a grammatical category?), and GRAMMATICAL (does the concept cover sentences with pleonastic pronouns, or not?). Likewise, I should add, for (wide) logical concepts. It is possible to give inferential-role characterizations of them (à la Gentzen – cf. Peacocke 1992), but if they are subject to Burge-style counterexamples, then why not say that they are individuated externally too? Consider a member of our language community who believed that ‘and’ has an intrinsic temporal meaning, so that for him sentences of the form p and q are not equivalent to sentences of the form q and p (though, say, he also believes that both p and q follow from p and q). Would we say that he did not have the concept AND, or that he had false beliefs about it? Compare him to his twin in a language community in which AND does have this meaning. It seems to me that if you accept the intuitions about the inferences Al is prepared to draw from ‘x has arthritis’, you ought to accept them about the inferences he is prepared to draw from sentences of the form p and q – and, hence, that if you don’t accept them about NOUN or AND, you shouldn’t accept them about ARTHRITIS and SOFA, in which case you’ve abandoned externalism. I conclude, therefore, that the hybrid view is unmotivated. But then either all concepts are individuated externalistically – in which case the nativist is unhappy, or they are all individuated internalistically (via inferential role) – in which the externalist should be unhappy.
3. Syntactic Nativism I have argued that going narrow is not going to help the (orthodox) compatibilist. Perhaps, then, the thing to do is to go syntactic – to argue that what is innate is not representations, but only their vehicles. By thus rejecting the innateness of content of any sort, the compatibilist would abandon (what I take to be) traditional nativism; but it is open to him to argue that his thesis is still philosophically interesting.
Nativism and the Theory of Content
235
In addition to the considerations mentioned above (section 2) concerning the (wide) content requirement, a prima facie objection to Syntactic Nativism is that it is merely a hypothesis about the innate physical structure of the brain: and why should a philosopher care about that? Why is it any more philosophically interesting than a thesis about the innate physical structure of the stomach? Recall my characterization of nativism as the view that there are innate mental/brain structures that predetermine in some way or other the cognitive capacities we are capable of entertaining. I take it that it is this predetermination that gives the nativist thesis its philosophical interest. The Syntactic Nativist could argue that the innateness of the structured objects1 that subserve conceptual content (plus whatever transducing mechanisms are needed to establish the relevant head-world hookups) are sufficient unto such predetermination (whereas nothing about the physical structure of the stomach has such effects). In particular, since the representational capacity of the brain underwrites our ability to think (by enabling us to entertain propositions), questions about the availability of mental syntax have a direct bearing on philosophically interesting issues – the question of what it is possible to think being clearly more interesting than the question of what it is possible to digest. The conceptual vehicles themselves do not have content, but since the grasping of concepts involves (let us suppose) representing them mentally, one is constrained to think only what can be “written” in the brain’s “language.”2 The Syntactic Nativist says that at least some (perhaps all) mentalese terms are innately tokened, and, therefore, that what it is possible to think is, in a sense, not (or not entirely) determined by experience.3 So the nativist thesis as construed by the Syntactic Nativist is philosophically interesting after all. Well, maybe not. It depends, for one thing, on whether or not you take lexical concepts to be semantically structured. If they are not, then the sense in which it is true that thought is limited by syntax is irrelevant to the debate over the innateness of (lexical) concepts, which is, I take it, what is at issue 1
Just what these objects would be is perhaps not completely clear. Assuming, however, that a distinction can be made between (say) neural structures that subserve concept possession and those that do not, there is at least some reason to think that the notion of a vehicle is not without objective sense.
2
I am not committing myself to a language-of-thought story here, though I will find it convenient to speak of mental representations as if I were.
3
The Syntactic Nativist may be contrasted with the Syntactic Anti-nativist, who says that the brain’s representational vocabulary – hence, what can be written in mentalese (hence, what can be thought) – is entirely determined by experience. An odd view, to be sure; but certainly a possible position in the logical space of the innateness debate.
236
David Pitt
here. The availability of conceptual vehicles of some particular “shapes” (for this is what the issue comes to) can have no effect on the availability of mental contents: the sense in which the limitations of brain-writing are the limitations of thought does not concern the mental vocabulary. It certainly seems correct to say that, for example, a mind/brain that could not “write” mentalese sentences with subordinate clauses (i.e., could not token structures interpretable as such) or could not form (mental structures interpretable as) disjunctive predicates, would be incapable of thinking many thoughts that we can think. But if the Syntactic Nativist takes the innate conceptual endowment to consist of the vehicles of simple concepts, there does not seem to be a way in which what can be written in mentalese could limit what can be thought. One would have to say that the limitations of mental orthography are the limitations of the content the vehicles can acquire. But this seems just false. If conceptual vehicles are syntactic objects whose expressive power is not determined by their significant internal structure (because they have none), they should no more be constrained to have any particular content than words in public languages; one will be as good as another for the representation of any “lexical” property. If, on the other hand, the syntactic nativist assumes that some lexical concepts are complex, then, presumably, whatever innate limitations there may be on available constructive principles or structures would likely have consequences for what concepts may be entertained. If, for example, the concept MOTET includes the concept COMPOSITION as a superordinate (a motet is a kind of (musical) composition), then a mind/brain incapable of tokening structures with formal characteristics appropriate to the representation of superordination could not entertain the concept MOTET. So Syntactic Nativism could be an interesting thesis after all, though in an unexpected way. The philosophical interest of Semantic Nativism lies, in part, in its claim that we might be capable of entertaining thoughts whose content could not be gleaned from causal relations between our minds and the external world (i.e., experience). On the other hand, this version of Syntactic Nativism implies that there might even be experiential concepts we are incapable of entertaining. This probably isn’t what the Syntactic Nativist had in mind; but, it seems to me, it is all he is likely to get by way of philosophical interest for his theory. For, how, given an externalist picture of content determination, could the internal structure of a conceptual vehicle enable it to have a content that could not be fixed in experience? In any case, as noted above, it seems likely that most philosophers would be unwilling to pay the price of semantic decomposition and the analyticsynthetic distinction for a philosophically interesting nativism.
Nativism and the Theory of Content
237
For these reasons, I do not think syntactic nativism is a live option, and (for the orthodox, at least) something does have to give. One cannot (without accepting the analytic-synthetic distinction) have both a plausible or philosophically interesting nativism and an externalist theory of content.
4. Conclusion I have presented the externalist-cum-nativist with a budget of nested dilemmas. He will be either a Semantic Nativist or a Syntactic Nativist. If he is a Semantic Nativist, then (qua externalist) he is either a causal theorist or a nomic theorist. If he is a causal theorist, then his nativism is blatantly inconsistent with his externalism. If he is a nomic theorist, then he is either a wide nomic theorist or a narrow nomic theorist. If he is a wide nomic theorist, then either he allows that what is in the head is sufficient for content representation – in which case he is not an externalist, or he does not – in which case he is not a Semantic Nativist. Narrow Nativism is not a view that would appeal to most nativists. If, however, the externalist-nativist opts for a Narrow Nomic Nativism, he (most likely) will assume a computational-role view of narrowcontent-function satisfaction. But computational role is either sufficient for narrow representation or it is not. If it is not, then he is not a Semantic Nativist. If it is, then he must accept either the analytic-synthetic distinction or a highly implausible version of the (semantic) nativist thesis. Syntactic Nativism is also not a view that would appeal to most nativists. If, however, the externalist-nativist opts for it, then (whatever his views about content determination) he must be either a primitivist or decompositionalist about lexical concepts. If he is a primitivist, then his thesis is not philosophically interesting. If he is a decompositionalist, then his thesis is philosophically interesting, but at the price of the analytic-synthetic distinction. Thus, the arguments in this paper show that any plausible or philosophically interesting nativism compatible with psychological externalism presupposes the analytic-synthetic distinction. I argued, further, that this is a cost many (though by no means all) philosophers would be unwilling to pay. In the end, it probably should not come as a surprise that externalism and nativism are not easily reconciled. Externalism undermines the Cartesian picture of the mind as metaphysically independent of the non-mental world, and seems an inevitable component of any effort to naturalize intentionality by making it causally based. The nativist, in contrast, is typically interested in explaining the possession of intentional states, including knowledge of nonnatural (or even natural) objects, whose content, he claims, couldn’t be deter-
238
David Pitt
mined on the basis of causal/nomic relations between the mind and the world. Whether those objects are considered to be constituents of thought or not, such states must be inherent in a world-independent mind. Perhaps, for some, the choice between nativism and externalism – or between nativism/externalism and the denial of the analytic-synthetic distinction – is an easy one to make. Many mainstream contemporary philosophers and cognitive scientists, however, should have a harder time of it. If the arguments in this paper are sound, they must choose between two of their central commitments.1,2
References Block, N. 1986. “Advertisement for a Semantics for Psychology,” in Midwest Studies in Philosophy, Vol. X: Studies in the Philosophy of Mind, P.A. French, T.E. Uehling and H.K. Wettstein, eds., University of Minnesota Press, Minneapolis, 615-678. Boghossian, P. 1989. “Content and Self-Knowledge,” Philosophical Topics 17, 5-26. ——— 1996. “Analyticity Reconsidered,” Noûs 30, 360-91. Brown, J. 1995. “The Incompatibility of Externalism and Privileged Access,” Analysis 55, 149-156. Brueckner, A. 1990. “Skepticism About Knowledge of Content,” Mind 99, 447-51. ——— 1994. “Knowledge of Content and Knowledge of the World,” Philosophical Review 103, 327-343. Burge, T. 1979. “Individualism and the Mental,” Midwest Studies in Philosophy, Vol. IV: Studies in Metaphysics, P.A. French, T.E. Uehling and H.K. Wettstein, eds., University of Minnesota Press, Minneapolis, 73-121. ——— 1986. “Intellectual Norms and Foundations of Mind,” Journal of Philosophy LXXXIII, 697-720. Chalmers, D. 1996. The Conscious Mind, Oxford University Press, New York. Chomsky, N. 1965. Aspects of the Theory of Syntax, The MIT Press, Cambridge. Devitt, M. 1996. Coming To Our Senses, Cambridge University Press, Cambridge. Dretske, F. 1981. Knowledge and the Flow of Information, The MIT Press, Cambridge. ——— 1988. Explaining Behavior, The MIT Press, Cambridge. Falvey, K. and Owens, J. 1994. “Externalism, Self-Knowledge and Skepticism,” Philosophical Review 103, 107-137. Field, H. 1977. “Mental Representation,” in N. Block, ed., Readings in Philosophy of Psychology, Vol. II, Harvard University Press, Cambridge, 78-114. ——— 1978. “Logic, Meaning and Conceptual Role,” Journal of Philosophy LXXIV, 379-408. Fodor, J.A. 1975. The Language of Thought, Crowell, New York.
1
Jerry Fodor (in propria persona) has indicated that he, for one, would be willing to pay the price of a philosophically uninteresting nativism for an informational (and atomistic) psychosemantics.
2
For helpful comments on drafts of this paper I would like to thank Brad Armour-Garb, Tim Crane, Jerry Katz, Jeff Poland, Georges Rey, Daniel Stoljar, Adam Vinueza and Barbara Von Eckardt.
Nativism and the Theory of Content
239
——— 1981. “The Present Status of the Innateness Controversy,” in Representations, The MIT Press, Cambridge, 257-315. ——— 1983. The Modularity of Mind, The MIT Press, Cambridge. ——— 1987. Psychosemantics, The MIT Press, Cambridge. ——— 1990. A Theory of Content and Other Essays, The MIT Press, Cambridge. ——— 1991. “AReplies,” in Meaning in Mind: Fodor and His Critics, B. Loewer and G. Rey, eds., Basil Blackwell Ltd., Oxford, 255-319. ——— 1994. The Elm and the Expert: Mentalese and Its Semantics, The MIT Press, Cambridge. ——— 1998. Concepts: Where Cognitive Science Went Wrong, Oxford University Press, Oxford. Harman, G. 1987. “A(Nonsolipsistic) Conceptual Role Semantics,” in Lepore, ed., New Directions in Semantics, Academic Press, Inc. London Ltd., London, 55-81. Heil, J. 1988. “A Privileged Access,” Mind 97, 238-51. ——— 1992. The Nature of True Minds, Cambridge: Cambridge University Press. Kaplan, D. 1989. “Demonstratives,” in Themes from Kaplan, J. Almog, J, Perry and H. Wettstein, eds., Oxford University Press, Oxford, 481-614. Katz, J. 1988. “The Refutation of Indeterminacy,” Journal of Philosophy LXXXV, 227-52. ——— 1990. The Metaphysics of Meaning, The MIT Press, Cambridge. Lloyd, D. 1987. “A Mental Representation from the Bottom Up,” Synthese 70, 23-78. Lycan, W. 1984. Logical Form in Natural Language, The MIT Press, Cambridge. McCulloch, G. 1995. The Mind and Its World, Routledge & Kegan Paul, London. McDowell, J. 1986. “Singular Thought and the Extent of Inner Space,” in Subject, Thought and Context, J. McDowell and P. Pettit, eds., Oxford University Press, Oxford, 137-168. McGinn, C. 1982. “The Structure of Content,” in Thought and Object, A. Woodfield, ed., Clarendon Press, Oxford, 207-257. McKinsey, M. 1991. “Anti-Individualism and Privileged Access,” Analysis 51, 9-16. Millikan, R.G. 1993. White Queen Psychology and Other Essays for Alice, The MIT Press, Cambridge. Papineau, D. 1987. Reality and Representation, Basil Blackwell Ltd., Oxford. Peacocke, C. 1992. A Study of Concepts, The MIT Press, Cambridge. Putnam, H. 1975. “The Meaning of ‘Meaning’,” Language, Mind and Knowledge: Minnesota Studies in the Philosophy of Science, Vol. 7, K. Gunderson, ed., University of Minnesota Press, Minneapolis, 131-193. Rey, G. 1993. “The Unavailability of What We Mean I: A Reply to Quine, Fodor and Lepore,” Grazer Philosophische Studien 46, 61-101. Stalnaker, R. 1984. Inquiry, The MIT Press, Cambridge. Stampe, D. 1977. “Toward a Causal Theory of Linguistic Representation,” Midwest Studies in Philosophy, Vol. I: Contemporary Perspectives in the Philosophy of Language, P.A. French, T.E. Uehling and H.K. Wettstein, eds., University of Minnesota Press, Minneapolis, 81-102. Sterelny, K. 1989. “Fodor’s Nativism,” Philosophical Studies 55, 119-141. Stich, S. (ed.) 1975. Innate Ideas, University of California Press, Berkeley. White, S. 1982. “Partial Character and the Language of Thought,” Pacific Philosophical Quarterly 63, 347-65. Woodfield, A. 1993. “Do Your Concepts Develop?” Philosophy 34, 41-67.
240
Raffaella De Rosa
RAFFAELLA DE ROSA On Fodor’s Claim that Classical Empiricists and Rationalists Agree on the Innateness of Ideas1 Abstract Jerry Fodor has argued that Classical Empiricists are as committed to the innateness of (at least some) ideas as Classical Rationalists. His argument, however, is proven inconclusive by an ambiguity surrounding “innate ideas”. Textual evidence for this ambiguity is provided and the “Dispositional Nativism” that, prima facie, makes Empiricist and Rationalist views similar dissolves into two distinct views about the nature of both the mind’s and the environment’s contribution in the process of concept acquisition. Once the Empiricist’s Dispositional Nativism is not conflated with the Rationalist’s, it becomes evident that the Empiricist can accept the premises of Fodor’s argument without accepting his conclusion and, hence, remain unmoved in her conviction that no ideas are innate.
That Classical Rationalists and Empiricists agree on concept nativism – or on the nativism of (at least some) ideas – is a leitmotif in Fodor’s writings.2 I will evaluate this Fodorian theme, concluding that it is untenable. Here in brief is how I will proceed. Section I outlines Fodor’s argument for the view that, contrary to established opinion – according to which Rationalists regard concepts as innate whereas Empiricists regard them as derived from experience – the latter are, willy-nilly, committed to the innateness of, at least, primitive (i.e., sensory) ideas. Section II presents the gist of my criticism of Fodor: the validity of Fodor’s argument is threatened by an ambiguity in “innate idea”,
1
I thank Martha Bolton, Mary Frances Egan, Jerry Fodor, Ernie Lepore, Kirk Ludwig, Robert Matthews and Roger Woolhouse for helpful comments and encouragement while writing this paper.
2
See [Fodor, 1975], [Fodor, 1981] and [Fodor, 1998]. The present paper focuses only on the “standard argument” for nativism given by Fodor in [Fodor, 1981]. However, it is worth noticing that [Fodor, 1998] rejects the view that Empiricists and Rationalists agree that “the initial state for [concept] acquisition needs to be intentionally specified” ([Fodor, 1998], p.143) – viz., the view that Rationalists and Empiricists agree on innateness of ideas. He now prefers the view that Rationalists and Empiricists equally acknowledge the existence of an innate mechanism underlying concept acquisition. However, it is unclear how different [Fodor, 1998]’s views are from the content nativism of [Fodor, 1975] and [Fodor, 1981]. Notice that for the purposes of this paper, I use “concept” and “idea” interchangeably.
On Fodor’s Claim that Classical Empiricists and Rationalists ...
241
and related expressions. Sections III and IV offer textual support for the ambiguity in the Modern debate on the acquisition of ideas. The aim of this paper is not to defend an empiricist theory of concept acquisition, but to show that Fodor has not conclusively established that Empiricism is committed to nativism about primitive ideas, and hence, has not established that it is no real alternative to a Nativist theory of concept acquisition.
I. Fodor’s argument1 In [Fodor, 1981], Fodor argues that the issue between Modern Rationalist and Empiricist theories of concept acquisition “has been widely misinterpreted and that when one sees what the issue really is one sees that the Nativist account is distinctly in the running” (p.259). His upshot is that classical theories of concept acquisition, Empiricist and Rationalist, regard concepts as innate. Fodor’s argument can be reconstructed in four steps: Step I. Classical theories of concept acquisition. Classical theories of concept acquisition come in two varieties: the Empiricist and the Nativist. Their theories of concept acquisition share the following traits: (i) both are a species of a Representational Theory of the Mind, since both regard mental representations as mental objects (p.259); (ii) both regard a concept as a (sort of) mental representation that expresses a property, and is expressed by an open sentence (p.259); (iii) both assume infinitely many phrasal concepts are acquired by applying combinatorial procedures to a basis of lexical concepts (p.262)2; and, most importantly, (iv) both maintain that the concepts an organism acquires are a function of the environmental stimulations it encounters (p.265).3
1
The page references in this section are to [Fodor, 1981], unless otherwise indicated.
2
By “lexical concept” Fodor means a concept “that is expressible, in English, by an open sentence with a morphologically simple [...] predicate term” (p.260). So, for example, RED as well as TABLE are lexical concepts, because they are expressible, respectively, by the open sentences “...is red” and “...is a table”. By “phrasal concepts”, he means a concept that is expressible, in English, by an open sentence with a morphologically complex or syntactically structured term (p.261). So, for example, READS THE NEWSPAPER AND LISTENS TO SCHUBERT is a phrasal concept expressible by the open sentence “...reads the newspaper and listens to Schubert”.
3
At first glance, it might seem odd to list as a common trait of both Nativist and Empiricist theories of concept acquisition the view that the concepts a subject possesses are a function of the environmental stimuli it encounters. However, listing of this common trait is justified by the fact that Rationalists like Descartes and Leibniz never denied the significance of experience for the “acquisition” of concepts. More on this later.
242
Raffaella De Rosa
Step II. The Empiricist Story. According to Empiricism, the space of potentially available concepts is fixed as follows: every concept is either primitive or acquired from primitive concepts by applying combinatorial rules; primitive concepts are sensory; therefore, every concept is either sensory and primitive, or acquired from sensory concepts by applying combinatorial rules, and so is complex.1 As to concept acquisition, Empiricism postulates two mechanisms, one for (i) primitive concepts and one for (ii) complex concepts. I begin with (i). According to Empiricism, the mechanism that occasions the availability of primitive concepts is merely causal. Environmental stimuli cause (occasion or trigger) primitive concepts in an organism’s mind. Since a causal process (such as occasioning and triggering) is not rational and since learning is rational, the acquisition of primitive concepts is not a learning process. So, primitive concepts are unlearned. Moreover, Fodor argues, what determines what kind of stimuli is necessary and sufficient to occasion (or cause) concepts in an organism’s mind is the innate structure of the sensorium (p.265). Since then, on an Empiricist account, primitive concepts are unlearned and their availability is determined by the innate structure of the sensorium, we must conclude that, according to Empiricism, (at least) primitive concepts are innate. As Fodor puts it, the Empiricist agrees with the Nativist that (at least with regard to sensory-primitive concepts) “the concept isn’t coming from the environment, it’s coming from the organism. All the environment does is provide the triggers that release the information” (p.280, emphasis added). As to (ii), Empiricism postulates a rational relation between a complex concept and the experiences that issue in its acquisition. Fodor contends that any Empiricist theory of concept learning must postulate a process of hypothesis formation and confirmation to explain how the subject learns the concept of being an X from experiencing instances of X ([Fodor, 1975], pp.34-42). So, the rational relation Empiricism postulates between complex concepts and the experiences that issue in their acquisition is specified as a process of hypothesis formation and confirmation. To illustrate briefly, take the experimental paradigm of concept learning. A subject is faced with the task of learning a concept FLURG from some sequence of experiential data. FLURG is the same concept as GREEN OR SQUARE; the subject is shown a pack of stimulus cards each of which represents a colored geometrical figure (yellow circles, green triangles, red squares and so on); the experimenter guides the subject’s learning 1
Since the Empiricist regards sensory concepts as primitive, the class of primitive concepts is not identical to the class of lexical concepts. Many lexical concepts inevitably turn out to be internally structured (i.e., complex), on the Empiricist account.
On Fodor’s Claim that Classical Empiricists and Rationalists ...
243
by informing her when she guesses correctly, where the subject’s response is correct iff the card is either green or square. The subject’s task of learning FLURG is a process of inductive extrapolation from a set of experiential data. The subject’s learning is a function of a confirmation relation between the data and the hypothesis: the concept FLURG is the concept of that which is green or square. If concept learning is a hypothesis testing process, we must conclude with Fodor that the Empiricist theory of concept learning is no theory of concept acquisition at all. With complex and primitive concepts, Fodor argues, an Empiricist theory of concept learning must assume the availability of the concept being learned in order for the subject (a) to express the hypothesis “the concept of being an X is the concept of something that is such and such” (where the concept of being an X is the concept being learned); and (b) to represent the data as instances of the concept of being an X. In the case of complex concepts, since concept learning is essentially a process of hypothesis formation and confirmation, complex concepts must already be available to the subject, because the hypothesis contains the complex concept to be learned. (That is, only (a) above applies to complex concepts. Notice that Fodor is not committed to the view that the complex concept is available in the sense that it is innately given. What is available to the subject are the primitive concepts and the combinatorial rules out of which the complex concept is built). In the case of primitive concepts, the concept must already be available to the subject, because the hypothesis contains the concept to be learned and the data are necessarily represented in terms of that concept being learned (That is, both (a) and (b) above apply to primitive concepts). To illustrate, complex concept learning “needn’t presuppose the availability of that concept for the canonical representation of the data base” (p.271). The hypothesis that FLURG is the concept of something green or square is confirmed by observations such as “Stimulus 1 is green and is positive” but does not require for its confirmation observations such as “Stimulus 1 is green or square and is positive”. Learning a primitive concept is different. Whereas “hypotheses which contain complex concepts can (maybe) be inductively confirmed by a data base in which only primitive concepts figure”, hypotheses with primitive concepts are inductively confirmed by a data base with those very same primitive concepts – on pain of an empty data base which confirms nothing (p.272). Consider learning the concept GREEN. The hypothesis to be confirmed is the concept being learned is the concept of something that is GREEN and the confirming observation is “Stimulus 1 is green and positive”.
244
Raffaella De Rosa
Consequently, primitive concepts cannot be learned;1 and so, Empiricism has no account of concept learning, since what it offers at best applies to complex concepts – and even then, as we have seen, any concept being learned must already in some sense be available to the subject (p.272). It follows, according to Fodor, that the Empiricist is unwittingly committed to concept nativism. Since, according to Empiricism (i) the disjunction “every concept is either sensory and primitive or acquired from primitive concepts by application of some combinatorial rule and complex” exhausts the space of potentially available concepts; and (ii) primitive concepts are unlearned (or simply occasioned by experience) and complex concepts are composed from primitive concepts by the application of an innate combinatorial apparatus; then, (iii) the space of potentially available concepts is innately fixed. Step III. The Rationalist/Nativist Story. Nativism also requires concepts to be either primitive or built up out of primitive concepts by application of combinatorial rules. However, for Nativism lexical concepts are primitive concepts (p.274). As to concept acquisition, Nativism regards the acquisition of lexical concepts as causal. All primitive/lexical concepts are simply occasioned in the mind by experience. “[T]he Nativist says that the triggering of the sensorium is, normally, causally necessary and sufficient for the availability of all concepts except those that are patently phrasal” (p.273). Step IV. The Real Issue between Nativism and Empiricism. Steps (I)-(III) are intended to establish that the issue between Nativism and Empiricism is misinterpreted if it is put (as it is normally) as follows: the Empiricist regards primitive concepts as deriving from experience, whereas the Nativist regards them as innate. Steps (II) and (III) in particular show that Nativists and Empiricists agree that the availability of primitive concepts is merely occasioned in the mind by some triggering experience. As Fodor writes: Both [the Empiricist and the Nativist] theories assume that the availability of primitive concepts is a function of environmental stimulation. In particular, they both assume that the acquisition of such concepts is contingent upon the activation of the sensorium. As Descartes frequently emphasized, it is not part of the Rationalist program to claim that your repertoire of primitive concepts is independent of the character of your experience. On the contrary, just as there are presumably triggering experiences that are nomologically necessary for the acquisition of sensory-primitive concepts like RED, so are there presumably triggering experiences that are 1
This confirms and explains what was said above about the Empiricist explanation of the acquisition of primitive concepts. Primitive concepts are simply triggered or occasioned by external stimuli.
On Fodor’s Claim that Classical Empiricists and Rationalists ...
245
nomologically necessary for the acquisition of non-sensory primitive concepts like TRIANGLE. In either case, if you don’t encounter the trigger, you don’t get the concept (pp.274-5, a mixture of mine and original emphases).
If Nativism and Empiricism agree that primitive concepts are unlearned, viz., that they are occasioned in the mind by experience, it follows, according to Fodor, that they also agree primitive concepts are innate; then the real dispute between Empiricism and Rationalism “is over which concepts are primitive” (p.281), with the Nativist claiming many primitive concepts are non-sensory. For clarity, I summarize Fodor’s argument as: (1) Concept learning is a rational process of hypothesis formation and confirmation. So, (2) in learning the concept of being an X a subject must be able to express the hypothesis (e.g., The concept of being an x is the concept of something which is Y or V) and represent the data against which that hypothesis is tested (e.g., “Stimulus 1 is Y and is positive”). (3) The concept being learned then must be in principle available to the subject for her to be able to express a hypothesis and represent data ([Piattelli-Palmarini, 1980], p.151 and [Fodor, 1981], p.280). (4) These concepts can be available to the subject in either of two ways: learned or unlearned. (5) Not all of them can be learned on pain of circularity – viz., at least some (i.e., the primitive ones) are unlearned. Therefore, (6) Concept learning presupposes the innateness of (at least) some concepts, i.e., the primitive ones. Since, according to Fodor, any theory of concept learning endorses premise (1), the Empiricist is committed to concept nativism.
II. On the Status of Fodor’s argument and the ambiguity of “innate idea” Before evaluating Fodor’s view, preliminary considerations on the status of his argument are in order. There are at least two ways in which it can be challenged: One strategy would be to challenge the alleged necessary connection between concept learning and concept nativism by challenging Fodor’s assumption that concept learning is essentially a process of hypothesis formation
246
Raffaella De Rosa
and confirmation (i.e., by challenging premise (1)).1 I will not pursue this challenge. Instead, my strategy will be to grant premise (1) but challenge the validity of the inference from the premise that Empiricism must acknowledge that primitive ideas are unlearned and in principle already available to the subject to the conclusion that Empiricism is as committed to the innateness of primitive ideas as Rationalism.2 In particular, I will argue that the inference is threatened by an ambiguity in both “innate ideas” and (an idea being) “in principle available to us”. “Innate ideas’ is ambiguous between two readings, a weak and a strong one: (1) The weak reading. An idea is innate in a subject just in case the subject is born possessing an innate capacity or disposition to attain that idea on the occasion of encounters with the environment. Mutatis mutandis, endorsing a weak reading of “innate ideas” amounts to endorsing the negative (and rather innocuous) claim that we are not infinitely flexible in the range of concepts we may attain. (2) The strong reading. An idea is innate in a subject just in case the subject is born possessing innately specified mental objects; these innately specified mental objects are present in the mind in a real way although they may not be available to the subject until triggered by an appropriate encounter with the environment. Both the weak and strong readings emphasize the relevance of experience in the “acquisition” of ideas though, as we shall see in Section IV below, “acquisition of ideas”, in turn, takes up a different meaning contingent on whether a strong or weak reading of “innate ideas” is in play. Moreover, these distinct readings of “innate ideas” correspond, respectively, to distinct understandings of (ideas being) “in principle available to us”: (3) Ideas are in principle available to us in the sense that all ideas that (compatible with the way our perceptual apparatus is innately constituted) can be acquired on the occasion of encounters with the 1
One could argue, contra Fodor, that the Empiricist held a “habit strength” model of associative learning, that is, that the Empiricist sees the rational process underlying the acquisition of concepts as a purely associative rather than hypothesis-testing process. It is worth noting that Fodor is well aware of this possible objection and, in [Fodor, 1981] (see, pp.268-9 and fn.4), rejoins that associative processes are better described as hypothesis-testing processes.
2
This criticism was first voiced by [Samet, 1896] and [Samet & Flanagan, 1989]. I essentially agree with them that the inference from “being unlearned” to “being innate” is problematic, but, as we shall see presently, I will expand on their criticism. The importance of this criticism has been somewhat acknowledged by [Fodor, 1998], ch.6.
On Fodor’s Claim that Classical Empiricists and Rationalists ...
247
environment are in principle available to us. Ideas are in principle available to us in the sense that “any logically possible concepts can, given the appropriate experience, be realized in the mind […]”. ([Katz, 1966], p.146, emphasis added). (4) Ideas are in principle available to us in the sense that though really present in our mind, they are not being used (viz., they are latent). The difference between claiming that an idea is in principle available and innate to us in the strong and in the weak senses is the same as the difference between claiming that The Thinker was present in the veined block of marble in Rodin’s studio and claiming that The Thinker was present in the blank hunk of marble in Rodin’s studio. Only in the former case can we reasonably say the statue of The Thinker was present in the block of marble before Rodin started carving; in the latter case, we can say the statue of The Thinker was present in the block of marble only after Rodin finished carving it out. One may object that the analogy is not fully adequate since the weak reading of “innate ideas” admitted a degree of predisposition of the mind to acquire certain ideas on the occasion of certain stimuli – i.e., we admitted that the mind is not infinitely flexible to experience. In response, we can fix the analogy as follows: the blank piece of marble can be disposed to be carved in certain ways rather than others (e.g., it could break under certain blows) – without ipso facto containing the shape of the thinker itself. As I will argue presently, though Modern Rationalists like Descartes and Leibniz did acknowledge, as Fodor rightly notes, the relevance of experience in the “acquisition” of ideas, they endorsed nativism of ideas and thought of ideas being “in principle available to us” in the robust sense mentioned above. Namely, according to Modern Rationalists, experience is necessary for the subject’s acquisition of ideas only in the sense that experience is necessary to trigger or release what was already present (though latent) in the subject’s mind. This should come as no surprise since the interesting issue in the debate over innate ideas is whether innate ideas exist, viz., whether innate contents are in our minds. Likewise, although Modern Empiricists like Locke, as Fodor correctly notes, acknowledged the innateness of the sensorium and, hence, (given the innateness of the sensorium) recognized that some ideas must be in principle already available to us (in the weak sense mentioned above), they never endorsed a “nativism of ideas” beyond the weak reading. This too should come as no surprise since, as Empiricists were well aware, (1) above is neutral between Empiricists and Rationalists and, hence, is not a strong enough nativism to support the nativism of ideas.
248
Raffaella De Rosa
With this said, we can see why Fodor’s inference from premises (1)-(5) to the conclusion (6) that Empiricism is committed to the innateness of ideas is dubious. In the Empiricist case, what is validly inferred from conceding that some ideas must in principle be available to the subject, is that ideas are innate in the weak sense – viz., we have an innate disposition or capacity to acquire ideas upon encounters with our environment. However, since the interesting sense in which ideas are innate is the sense in which a subject is born possessing innate mental contents, Fodor’s conclusion that Empiricists are committed to the innateness of ideas must assume the strong reading. If this is so (as I think it is since inferring nativism in a weak sense would not give Fodor the conclusion he wants, i.e., an Empiricist commitment to nativism of ideas), Fodor’s argument is a fallacy of equivocation, turning on an equivocal reading of “innate ideas” and of an idea being “in principle available to us”. In short, the inference from premises (1)-(5) to (6) is fallacious when applied to the Empiricist view because: Being faithful to Empiricism requires interpreting (5), namely, (i) that some ideas must in principle be already available to the subject, in the weak sense.1 Fodor’s claim that Empiricism is committed to the nativism of (ii) ideas is interesting (i.e., the claim that Empiricists were unwittingly committed to the innateness of at least some mental contents) only if we interpret the conclusion (6) – that at least some ideas are innate to us – in the strong sense. (iii) From the premise that ideas are in principle available to the mind in the weak sense, it obviously does not follow that ideas are innate to the mind in the strong sense. The Empiricist can grant Fodor’s argument against concept learning without jeopardizing Empiricism, viz., he can grant Fodor’s premises without accepting his conclusion. 2 The ambiguity surrounding “innate ideas” emerges clearly in the debate between Modern Rationalists and Empiricists. In sections III and IV, I will provide textual evidence for this ambiguity in defense of my criticism of Fodor. In Section III, I consider the sort of textual evidence Fodor apparently
1
I will provide evidence for this interpretation in the sections below.
2
The Empiricist can grant Fodor that a theory of concept learning based on a hypothesis formation-confirmation model presupposes (on pain of circularity) that some concepts are unlearned without buying into his conclusion that unlearned concepts are innate. The burden of proof is then on the Empiricist who must provide an account of the mechanism that explains the acquisition of unlearned concepts.
On Fodor’s Claim that Classical Empiricists and Rationalists ...
249
has in mind and that is often offered in support of the claim that no substantial disagreement exists between Rationalism and Empiricism about concept acquisition. This evidence is found in both the Rationalist and Empiricist endorsement of what I shall call “Dispositional Nativism” (DN). In section IV, I argue that even if the Rationalist endorsed DN, that endorsement would still not be conclusive evidence for Fodor’s claim that Modern Rationalists and Empiricists agreed on the innateness of ideas. Firstly, DN is either neutral between Rationalism and Empiricism or, even worse, may equally well have the effect of turning the Rationalist into an Empiricist. Secondly, and more importantly, a more attentive reading of the passages offered in support of the Rationalist endorsement of DN and other textual evidence show that, in fact, the Rationalist endorsement of DN (despite the seemingly similarities in the wording of the view) differs in important respects from the DN the Empiricist is prepared to endorse. The difference consists in their different views on the nature of both the mind and the environmental contribution to the “acquisition” of ideas.
III.“Dispositional Nativism” (DN): the (alleged) evidence in support of Fodor’s argument In this section, I will locate textual evidence usually advanced in support of Fodor’s claim. First, I will discuss passages in Descartes and Leibniz where both advocate a kind of Nativism (which I shall call “Dispositional Nativism” (DN) and define below) that makes relevant concessions to Empiricism and, consequently, attenuates the degree of disagreement between Rationalist and Empiricist theories of concept acquisition. Second, I will discuss passages from Locke’s Essay that implicitly endorse DN. In brief, I suggest that both the Rationalist and Empiricist endorsement of DN may be offered in support of Fodor’s claim that no substantial disagreement exists between Rationalist and Empiricist accounts of concept acquisition. I begin with the Rationalist endorsement of DN. If the Empiricist critique of Nativism is targeted to a certain sort of Nativism and if there exists a different kind of Nativism which can make relevant concessions to the Empiricist, then the Empiricist critique loses its force and then it becomes unclear what the disagreement between the Nativist and the Empiricist is. I suggest that this may be so with Locke’s critique of Nativism.1 1
See [Locke, 1975], Books I-II. It is often remarked in the literature that Locke portrays the Nativism under attack so crudely it is legitimate to ask who his adversaries were and whether they held the view he attributes to them. Excellent documentation on the issue is provided by
250
Raffaella De Rosa
In Book I.2.1 of his Essay, Locke depicts Nativism as the “established opinion amongst some men” that innate ideas are “as it were stamped upon the mind of man, which the soul receives in its very first being; and brings into the world with it” ([Locke, 1975], p.48).1 So far, according to Locke, the Nativist agrees with Plato that the mind possesses ideas in its pre-carnate state and, hence, comes into the world with these ideas imprinted on itself. In Book I.2.5, it becomes clear that Locke interprets Nativism as the view that the mind must know these ideas. It is a contradiction, Locke writes in Book I.2.5, “to say that there are [ideas] imprinted on the soul, which it perceives or understand not [...]” since “to imprint any thing on the Mind without the mind’s perceiving it, it seems to me hardly intelligible” (Ibid., pp.49-50). Contra Nativism, Locke writes in Book I.2.1, he (the Empiricist) argues that “Men, barely by the Use of their Natural faculties, may attain all the knowledge they have, without the help of any Innate Impressions” (i.e., without assuming the existence of innately specified mental objects). That men do not possess innate ideas in addition to their natural capacities to attain knowledge is shown, according to Locke, as follows. Assume, Locke writes in Book I.2.5, that ideas are imprinted upon the mind; then, since it is contradictory that ideas are both imprinted upon the mind and the mind does not know them, the case of children and “ideots” (who clearly do not know certain ideas) forces the Nativist to derive a contradiction from her own assumption; it follows that the assumption that ideas are imprinted on the mind is false. As Locke puts it, “to say that a notion is imprinted on the mind, and yet at the same time to say that the mind is ignorant of it [...] is to make this impression nothing” (my emphasis). So, the Nativist should grant the Empiricist that the mind has only the (innate) capacity to attain ideas (from experience). However, no advocate of innateness ever maintained that children were born possessing ideas in the sense of actually thinking ideas or knowing these ideas at birth. Locke’s contemporaries advocated a Dispositional Nativism (DN), according to which, the mind is not born actually thinking ideas; rather,
[Yolton, 1956] ch.2; see, also, [Woolhouse, 1983], p.43; [Greenlee, 1972]; and [Armstrong, 1969]. It is generally agreed that Locke’s polemic targets were (rather than Descartes) the British theologians and moralists in the XVIIth Century, who used the doctrine of innate ideas and principles to underpin moral and religious beliefs whose truth (they thought) was threatened by the materialism of the “new science”. Among them, suffice it here to mention Henry More (the Cambridge Platonist) and the Bishop of Worcester, Edward Stillingfleet. 1
This passage (and the whole of Book I) is mainly concerned with Locke’s critique of innate principles. However, since, as Locke himself points out (p.58 and pp.84-5), no proposition can be innate unless the ideas it is about (or out of which it is composed) are innate too, I freely translate Locke’s talk about innate principles into talk about innate ideas.
On Fodor’s Claim that Classical Empiricists and Rationalists ...
251
it is innately disposed to attain certain ideas on the occasion of certain environmental stimulations. I define DN as follows: (DN):
The mind has an innate disposition (or faculty) to attain certain ideas on the occasion of certain environmental stimulations. Accordingly, the idea of being an X is innate to the organism Y iff it is true that were Y to encounter instances of X, Y would acquire the idea of being an X.1
Consequently, a Nativist who endorses DN can agree with Locke that “natural impressions” on the mind are “nothing” (i.e., no mental objects always present to the mind); and that what men need in order to acquire ideas is an innate disposition of the mind to attain certain ideas on the occasion of certain environmental stimulations. Among the many Rationalists who advocated DN, James Lowde, one of Locke’s first critics, objected that the doctrine of innate ideas is not necessarily committed to the child being born with actual and occurrent ideas. In A Discourse Concerning the Nature of Man, Lowde writes: These natural notions are not so imprinted upon the soul, as that they naturally and necessarily exert themselves (even in children and idiots) [...] without any assistance from the outward senses or without the help of some previous cultivation ([Lowde. 1694], quoted in [Yolton, 1956] and [Woolhouse, 1983].
The locus classicus of Descartes’ endorsement of DN is a passage from Comments on a Certain Broadsheet 2. There, in response to Regius’ remark that the mind has no need of innate ideas besides its innate faculty of thinking, Descartes writes: [...] [Regius’] disagreement with me seems to be merely verbal. When he says that the mind has no need of ideas [...] which are innate, while admitting that the mind has the power of thinking (presumably natural or innate), he is plainly saying the same thing as I, though verbally denying it. I have never [...] taken the view that the mind requires innate ideas which are something distinct from its own faculty of thinking. I did, however, observe that there were certain thoughts within me which neither came to me from external objects nor were determined by my will, but which came solely
1
I deliberately leave the meaning of “acquire” vague in this definition since I believe that the alleged agreement between Empiricicts and Nativists on concept acquisition trades precisely on the ambiguity of the Rationalist formulation of (DN). As we shall see in the next section, the Rationalist (DN) entails, a more robust version of nativism.
2
All the references to Descartes’ works are to The Philosophical Writings of Descartes, volumes I-II, translated by Cottingham, Stoothoff and Murdoch (hereinafter, CSM); and volume III, translated by Cottingham, Stoothoff, Murdoch and Kenny (hereinafter, CSMK).
252
Raffaella De Rosa from the power of thinking within me [....]; so I applied the term ‘innate’ to [...] ideas [in] the same sense [...] in which we say that generosity is ‘innate’ in certain families, or that certain diseases such as gout or stones are innate in others: it is not so much that the babies of such families suffer from these diseases in their mother’s womb, but simply that they are born with a certain ‘faculty’ or tendency to contract them [CSM I, pp.303-4, my emphasis].
By the end of the Comments, Descartes insists that by “innate ideas” he never meant anything other than what Regius admitted as true, viz., “that ‘there is present in us a natural power which enables us to know God’” (Ibid., p.309); and that he never maintained “that such ideas are actual, or that they are sort of ‘forms’ which are distinct from our faculty of thinking” (Ibid., p.309). Rather, as he puts it, ideas are innate to us as certain diseases are innate to some families. As children of those families do not suffer from the disease in their mother’s womb but are genetically disposed to contract it later in life, so too, infants are not born possessing ideas but possessing dispositions to attain them later in life. Since the issue of concept acquisition comes down, at least partly, to the issue of how primitive ideas are acquired, I now illustrate Descartes’ DN by reviewing how he explains the “acquisition” of sensory ideas, viz., the ideas of primary and secondary qualities. However, it is worth noticing that Descartes lists among simple ideas not only ideas of bodies but also all intellectual ideas (God, all universals, geometrical and mathematical notions). (See, for example, Principles, in CSM I, pp.208-213). As to ideas of secondary qualities (i.e., ideas of color, taste and so on), Descartes calls them “adventitious” (i.e., coming from outside us), although he openly regards them as innate.1 His reason is that ideas of secondary qualities are occasioned in the mind by external things, although it is the innate structure of the sensorium (as opposed to things in the environment) which, in the end, explains why the subject attains certain ideas under certain stimuli. In other words, according to Descartes, something coming from external objects (i.e., local motions) occasions the mind to form ideas of color (and so on) because the mind is disposed by nature to form those ideas under appropriate stimulations. The view that the structure of the sensorium (rather than the environment) determines what ideas are occasioned in the mind under appropriate stimuli is related to Descartes’ view that there is no resemblance between ideas of secondary qualities and their causes. 1
For Descartes’ classification of ideas of secondary qualities as “adventitious” see CSM II, p.26. Moreover, the very proof of the existence of bodies, in the Sixth Meditation (CSM II, p.55), assumes that sensory ideas come from an outside world on pain of God being a deceiver. See also Principles II.1 (CSM I, p.223). For Descartes’ explicit claim that sensory ideas are innate, see Comments on a Certain Broadsheet, in CSM I, p.304.
On Fodor’s Claim that Classical Empiricists and Rationalists ...
253
As Descartes writes in Optics, the causes of ideas of secondary qualities are local motions: [...] it is the movements [in the brain] which, acting directly upon our soul in so far as it is united to our body, are ordained by nature to make it have such sensations. [...] regarding light and color [...] we must suppose our soul to be of such a nature that what makes it have the sensation of light is the force of the movements taking place in the regions of the brain where the optic nerve-fibers originate, and what makes it have the sensation of color is the manner of these movements. [...] there need be no resemblance between the ideas which the soul conceives and the movements which cause these ideas [CSM I, p.167, my emphasis.]
So, put crudely, ideas of secondary qualities cannot come to the mind from the environment because such ideas resemble nothing in the environment. The link between the innateness of the ideas of secondary qualities and their lack of resemblance with their causes is drawn by Descartes explicitly in his Comments on a Certain Broadsheet. He writes that once we realize all that reaches our faculty of thinking by way of the senses is a corporeal motion which “gives the mind occasion to form [...] ideas [of bodies] by means of the faculty innate to it” (CSM I, p.304), then, a fortiori, we must concede that ideas of secondary qualities are innate to us because “there is no similarity between these ideas and the corporeal motions” (Ibid. For discussion see [Wilson, 1991], pp.303-4). The situation is more complicated for ideas of primary qualities of body (motion, figure, number and so on). In Principles II.1, Descartes writes, reminiscent of Aristotle’s De Anima, that “we appear to see clearly, that the idea of [body as extended matter] comes to us from things located outside ourselves, which it wholly resembles” (CSM I, p.223, my emphasis). However, Descartes clearly denies that ideas of primary qualities are transmitted from material things to the mind via sense experience or that they are formed by the mind via an abstractive process based on observation of particulars. On the contrary, Descartes thought ideas of primary qualities innate. In Principles II.4, Descartes writes that the idea of the real nature of body (i.e., the idea of body as being extended in length, breadth and depth) is “implanted” in the intellect by nature (i.e., is innate) (CSM I, p.224). So, the above passage from Principles II.1 is to be interpreted as follows: experience of particular bodies is necessary to occasion in the mind the (already present) idea of body in general (as extended matter). This interpretation is confirmed by another passage from Principles II.1, where Descartes writes that “we have sensory awareness of, or rather, as a result of sensory stimulation we have clear and distinct perception of, some kind of matter, which is extended in length, breadth and depth, and has vari-
254
Raffaella De Rosa
ous differently shaped and variously moving parts [...]” (see [Wilson, 1991] and [Wilson, 1993]). Furthermore, in response to Gassendi’s objection that, in the Second Meditation, Descartes had abstracted the general idea of body as something extended, flexible and changeable from the observation of the particular properties of the piece of wax, Descartes replied that Gassendi misunderstood him because he “did not abstract the concept of the wax from the concept of its accidents.” Rather, he wanted “to show how the substance of the wax is revealed by means of its accidents” (CSM II, p.248, emphasis added). In conclusion, according to Descartes, no particular sensory given (particular instances of extended bodies) could cause the mind to form the idea of body unless the mind were already innately disposed to attain that idea upon experiencing particular extended bodies. Experiencing extended bodies is necessary for the mind’s attaining the idea of body but not sufficient to explain its origin and content. I briefly turn to Leibniz. In his Preface to New Essays on Human Understanding, Leibniz writes that ideas are innate to us as inclinations and dispositions of the mind.1 Interestingly, Theophilus, in response to Philalethes’ remark that he (Theophilus) seems to maintain we never learn anything new, replies: [...] You attribute to me, sir, a doctrine which I am far from accepting; for I quite agree that we learn innate ideas and innate truths, whether by paying heed to their sources or by verifying them through experience. [...] And I cannot accept the proposition that whatever is learned is not innate. The truths about numbers are in us; but still we learn them [...] ([Leibniz, 1981], p.85, a mixture of original and my emphases).
It follows from the dispositional version of Nativism that to say an idea is innate does not mean it cannot be acquired through experience. On the contrary, both Descartes and Leibniz emphasize the importance of experience in our “acquisition” of (innate) ideas2. A positive role for experience complements a dispositional account of innate ideas. So far I have provided evidence that the Nativist agrees with Locke that: (a) ideas are not innate to us as items of knowledge always present to our mind and distinct from our faculty of thinking; (b) all we need to admit as innate (to explain how we attain concepts) is an innate disposition (or faculty) of the mind to attain certain ideas on the occasion of certain environmental stimulations; and, hence, (c) environmental stimulation is necessary for concept acquisition. I move to show that Locke too endorses DN. 1
See [Leibniz, 1981], p.52. The New Essays are written in the form of a dialogue between Theophilus and Philalethes. The former is the spokesman for Leibniz, the latter for Locke.
2
Of course, as we shall see, this shifts the focus on what Dispositionalist Nativists mean by “concept acquisition”.
On Fodor’s Claim that Classical Empiricists and Rationalists ...
255
As we have seen, Locke is prepared to admit the existence of innate faculties to attain ideas from experience. The existence of such capacities is the very reason he does not see a need to postulate innate ideas in our mind; all we need to attain ideas, he argues against the Nativist, is the “use of [our] natural [innate] faculties” ([Locke, 1975], p.48). Locke concedes that (a) the innate structure of the sensorium plays an essential role in the acquisition of simple ideas of secondary qualities of bodies; and (b) some innate combinatorial power of the mind plays an essential role in the formation of complex ideas. Locke’s account of concept acquisition is generally presented as follows: ideas divide into simple and complex ones. Simple ideas are sensory, i.e., are furnished to the mind by experience; our experience of external objects (sensation) and of the internal operations of the mind (reflection) supplies the understanding with the simple ideas of, respectively, bodies and perception, thinking and so forth. Once the understanding is stocked with simple ideas, it forms novel complex ideas by repeating, comparing and uniting simple ideas in an infinite variety of ways (See [Locke, 1975], Book II.1.1-5; Book II.2.2 and Book II.11-12). So, all ideas come from experience because simple ideas derive from experience of both external and internal objects and complex ideas are composed out of simple ones by the application of innate combinatorial powers of the mind. Closer examination of how Locke explains the mind’s acquisition of simple ideas shows that the fact that simple ideas “come from experience” does not prevent their being “innate” (in the above dispositional sense). Let me explain.1 Although Locke claims that it is contrary to reason and experience to suppose that ideas of colors, tastes, sounds and so on are innate, there is abundant textual evidence that Locke (willy-nilly) regarded simple ideas of secondary qualities as innate. We possess certain ideas of secondary qualities (rather than others), according to Locke (as well as according to Descartes) because our sensory apparatus is disposed by nature to form those ideas on the occasion of certain external stimuli. In Book II.1.24, Locke writes “the first capacity of humane intellect is, that the mind is fitted to receive the impressions made on it [...] through the senses, by outward objects”. So, the first natural faculty that enables man to attain ideas is the innately fixed structure of the sensorium. That it is the innate structure of the sensorium which determines what ideas the mind can form under certain stimulations is evidenced by the following passage: [...] though we cannot believe it impossible to God, to make a creature with other organs, and more ways to convey into the understanding the notice of 1
I will limit my analysis to Locke’s account of simple ideas of secondary qualities.
256
Raffaella De Rosa corporeal things, than those five [...] which are usually given to man: yet I think it is not possible, for any one to imagine any other qualities in bodies, howsoever constituted, whereby they can be taken notice of, besides sounds, tastes, smells, visible and tangible qualities. And had mankind been made with but four senses, the qualities then, which are the object of the fifth sense, had been as far from our notice, imagination and conception, as now any belonging to a sixth, seventh or eight sense, can possibly be [...]. ([Locke, 1975], p.120)
In Book II.3.1, Locke writes of ideas of secondary qualities that they could not be formed if the sense organs “peculiarly adapted to receive them” were malfunctioning. That is to say, the acquisition of these ideas is determined by the natural/normal functioning of sense organs. Further evidence for Locke’s commitment to the “innateness” of simple ideas of secondary qualities can be found in Locke sharing with Descartes the view that ideas of secondary qualities have no resemblance with external causes occasioning them in the mind (See [Adams, 1975]; [Wendler, 1996] and [Fodor, 1981], pp.276-7). In Book II.8.1, Locke writes that simple ideas are caused in the mind by something so constituted in nature that, upon impressing itself on our sense organs, it causes the mind to form certain ideas. This “something” is a motion that, coming from external objects is “continued by our nerves” and then arrives “to the brains or the seat of sensation”([Locke, 1975], p.136). Consequently, for Locke there is no resemblance between the cause of the idea and the idea itself; secondary qualities “are [...] nothing in the objects themselves, but powers to produce certain sensations in us and depend on those primary qualities, viz., Bulk, Figure, Texture and Motion of Parts [...]” (Ibid., p.137).The similarity with Descartes’ view on ideas of secondary qualities is apparent. Like Descartes, Locke maintains that some local motions are “ordained by nature” to cause the mind to have certain ideas of color (and so on) and that these ideas do not resemble, or correspond to, any real quality in the environment. Therefore, Locke must also agree with Descartes that ideas of secondary qualities are not conveyed into the mind from the environment (since they do not exist in the environment) and, hence, are innate. I conclude that for Locke ideas of secondary qualities – at least in a dispositional sense – are innate because the innate structure of the sensorium explains why certain ideas are occasioned by certain external stimuli. Mutatis mutandis, ideas of secondary qualities are “innate” because the sensorium is disposed by nature to form them on the occasion of certain stimuli.1 1
The problem of understanding why, despite their substantial agreement on the acquisition of ideas of secondary qualities, Locke maintained that they come from experience and Descartes maintained that they are innate is raised by [Schmaltz, 1997].
On Fodor’s Claim that Classical Empiricists and Rationalists ...
257
What about complex ideas? Locke postulated some innate combinatorial power of the mind in the process of forming complex ideas. In Book II.2.2, he writes that once understanding is stored with simple ideas (as “the materials of knowledge furnished to the mind by sensation and reflection”), it (the understanding) “has the power to repeat, compare and unite them” in an infinite number of ways. In Book II.12.1, he writes “as the mind is wholly passive in the reception of all simple ideas, so it exerts several acts of its own, whereby out of its simple ideas, as the materials [..] of the rest, the other are framed”. These natural/innate combinatorial operations of the mind are essentially three: combination of simple ideas into a compound one; relation of ideas and abstraction of general ideas from particular ones. It follows that complex ideas cannot be regarded (at least not completely) as coming from experience. Complex ideas cannot come from the environment because an innate combinatorial power of the mind plays a crucial role in their formation. As Descartes replied to Gassendi’s objection that the idea of a chimera is an adventitious idea, concluding that a complex idea is not constructed by the mind but comes from experience is the same as concluding “that Praxiteles never made any statues on the grounds that he did not get from within himself the marble from which he sculpted them” (CSM II, p.250; For Gassendi’s objection, see CSM II, p.195). In conclusion, the textual evidence supports the view that there are more similarities between Rationalist and Empiricist accounts of concept acquisition than one might have expected. On the one hand, Rationalists like Descartes acknowledge that the mind does not come into the world possessing actual ideas but is rather disposed to acquire ideas on the occasion of external stimulations. On the other hand, an Empiricist like Locke is prepared to grant that we acquire ideas from experience in virtue of innate faculties of the mind. So, on the basis of this textual evidence, Fodor’s claim may be defended as follows: Rationalists and Empiricists agree the mind is provided with a natural endowment, viz., the mind is so structured by nature that environmental stimulation is causally necessary and sufficient to occasion ideas in it; they agree that the ideas so occasioned are primitive; the ideas so occasioned are, according to the Empiricist, sensory ideas; the ideas so occasioned are, according to the Nativist, lexical ideas; therefore, the only disagreement between Nativist and Empiricist is over which concepts are primitive. In brief, since Rationalists and Empiricists agree that the mind is naturally disposed to attain
258
Raffaella De Rosa
ideas on the occasion of stimulation, but disagree on which ideas are so occasioned, they agree on DN.1 Doubts may begin to arise in the reader’s mind as to whether the Empiricist and Rationalist mean the same thing by “DN”. I address that question in the next section. In the present section, in order to highlight similarities between Rationalist and Empiricist accounts of concept acquisition (viz., in order to understand what kind of evidence Fodor may have in mind in support of the claim that Modern Empiricist theories of concept acquisition are committed to concept nativism), I use “dispositionalism” broadly to mean the unqualified view that the mind is naturally inclined, by its own structure, to attain certain ideas on the occasion of encounters with the environment. As I will show in the next section, further analysis of what Rationalists, on the one hand, and Empiricists, on the other, packed into their “dispositionalism” ultimately creates problems for the full tenability of Fodor’s claim that no substantial disagreement exists between Empiricists and Rationalists about the origin of ideas. Despite their apparent agreement on (DN), Empiricists and Rationalists disagree about the kind of role both the mind and the environment play in concept acquisition.
IV. “Dispositional Nativism” Reconsidered In this section I will proceed as follows: first, I argue that even if Modern Rationalists and Empiricists agree on DN, their agreement does not establish agreement on the innateness of ideas because DN is too weak a nativism to defend nativism of ideas. Secondly, and most importantly, the evidence that supports Fodor’s claim (viz., the Rationalist and Empiricist endorsement of DN) is not conclusive evidence Rationalists and Empiricists agree on the innateness of ideas because Modern Rationalists and Empiricists agree on no common interpretation of DN. A careful reading of the passages in support of a Rationalist endorsement of DN and further textual evidence show that the Rationalist DN (despite similarities in wording) differs in relevant respects from the DN the Empiricist is prepared to endorse. The difference, as we shall see presently, consists in different views on the nature of both the contribution of the mind and the environment in the “acquisition” of ideas. 1
I note in passing that even a recent critic of Fodor’s defense of conceptual nativism, i.e., [Cowie, 1999], concedes that Empiricists and Nativists disagree about “precious little” since “[…] nativists agree with the empiricists’ ‘externalist’ insistence that […] interactions with the environment are necessary for the acquisition [of ideas]” and that “[…] empiricists agree with the nativists’ ‘internalist’ claim that were it not for our possession of some rather special inborn equipment, we [….] would have [no ideas at all]” ([Cowie, 1999], p.17).
On Fodor’s Claim that Classical Empiricists and Rationalists ...
259
So, consider the following objection to DN: by making relevant concessions to Empiricism, DN turns into Empiricism; namely, Nativism, in its dispositional version, claims so little it no longer merits the label “Nativism”. In response to Descartes’ and Leibniz’s account of innate ideas, one might object that by admitting as innate nothing besides the mind’s innate disposition to form certain ideas on the occasion of certain stimuli, Leibniz and Descartes agree with the Empiricist that at birth the mind is like a tabula rasa (in that no actual characters are imprinted on it) and that we attain what we know from experience. In other words, so this objection goes, since DN boils down to no more than infants having an innate disposition to acquire ideas from their environment, DN provides no support for Nativism, since it is compatible with the Empiricist claim that ideas come from experience. Both Regius and Locke advanced this sort of objection against the dispositional account of the innateness of ideas. Regius infers Empiricism about concept attainment from the premise (that Descartes admits, in his Comments on a Certain Broadsheet he shares) that the mind has no need of innate ideas besides its innate faculty of thinking: […] The mind has no need of innate ideas [...] which are innate: its faculty of thinking is all it needs for performing its own acts. […] Thus, all common notions which are engraved in the mind have their origin in the observation of things or in verbal instruction [CSM I, p.295, my emphasis].1
In a passage from his Essay, Locke writes that if the claim for innate ideas reduces to a claim for innate dispositions of the mind to attain ideas (“if the capacity of knowing be the natural impression contended for”, [Locke, 1975], p.50), then: [...] this great point will amount to no more, but only to a very improper way of speaking; which whilst it pretends to assert the contrary, says nothing different from those, who deny innate principles. [...] The capacity, they say, is innate, the knowledge acquired. But then, to what end such contest for innate maxims? [Ibid., my emphasis]
To Lowde’s dispositional version of Nativism, Locke replied that Lowde’s suggestion that the mind does not know ideas at birth but comes to know them afterwards with the help of the “outward senses” and some “previous cultivation” does not differ from what he had affirmed in Book I of his Essay. As Yolton points out, Locke’s remark is that if “the test for dispositional innateness can only come after experience (my emphasis)”, then “the claim for 1
Henry de Roy (or Henricus Regius, 1598-1679), towards the end of 1647, published anonymously a broadsheet under the title “An account of the Human Mind, or Rational Soul, which explains what it is and what it can be”. The text is republished in CSM I together with Descartes’ “Comments on a Certain Broadsheet” (early 1648) which is a response to Regius.
260
Raffaella De Rosa
innateness becomes pointless” ([Yolton, 1956], p.53). It follows that since DN is equally compatible with Rationalism and Empiricism, the dispositional account of innate ideas accepted by Rationalists and Empiricists is not conclusive evidence for Fodor’s claim that Empiricists and Rationalists agree on the innateness of ideas. Since DN is neutral between Empiricism and Rationalism, the fact that both endorse DN can be exploited as evidence that they agree on the innateness of ideas and also as evidence for the contrary, viz., that they agree on the non-innateness of ideas. Clearly, no progress is made in assessing either what being a Nativist means or who the Nativist is (is she an Empiricist? Is she a Rationalist? Or both?). This seems to suggest that for the Rationalist to defend innate ideas, she must defend a stronger version of Nativism. This remark takes us to my second point: The textual evidence that supports Fodor’s claim is ambiguous in that it is susceptible of being interpreted as suggesting both a DN as defined in the previous section and a DN that entails a more robust nativism (or, as I shall call it and define below, “Cognitive Nativism”). Consequently, DN as defined in the previous section does not provide a conclusive interpretation of the Rationalist DN. In particular, an examination of the passages quoted in Section III (along with further evidence) establishes that what Rationalists and Empiricists pack into “the innate disposition of the mind” is different in nature and, hence, that (contra Fodor and more along the lines of established opinion) Rationalists and Empiricists attribute a different role to experience in the acquisition of concepts. I start with the Rationalists. Textual evidence abounds that Leibniz and Descartes advocated a “Cognitive Nativism” (CN) along with DN: (CN): Ideas as such (viz., their intentional content) are innate to the mind. According to CN, the mind possesses ideas themselves – qua dispositions – prior to, and independently of, experience. Innate ideas are innate (though latent) mental objects. Accordingly, the idea of being an X is innate to the organism Y iff it is true that were Y to encounter instances of X, Y would become conscious of X. Before adducing evidence that Modern Rationalists held CN together with DN or, better yet, that the Rationalist DN entailed CN, let me clarify the difference between them. One might object that both in the case of ideas being nothing besides the innate disposition of the mind to attain certain ideas on the occasion of certain external stimulation and in the case of ideas being latent contents of the mind in need of external stimulation to become actual, the organism will not have the idea until it encounters instances. This worry is legitimate only if the test
On Fodor’s Claim that Classical Empiricists and Rationalists ...
261
for the difference between CN and DN comes after the organism has experienced instances of what it has ideas of. However, the difference between DN and CN consists in what the state of the organism is prior to encounters with its environment. The difference is as follows: whereas CN postulates the existence of something in the mind with intentional content (i.e., ideas), DN, strictly speaking, does not. All DN requires to be innate is a disposition of the mind to attain ideas on the occasion of encounters with its environment. Accordingly, the difference between DN and CN is the same as that the mind has a (general) disposition towards attaining certain ideas on the occasion of certain external stimulations and the mind has preformed ideas (qua dispositions) that await external stimulation to become actual (or conscious to the mind). In the former, the content of experience determines the repertoire of concepts the organism is going to have; in the latter, the content of experience is determined by (or structured by) the repertoire of concepts the organism already has.1 In short, the difference between DN and CN (or a DN that implies CN) amounts to the difference between, respectively, weak and strong nativism, as described in Section II above. I now move to provide evidence that the Rationalist DN is underpinned by CN. Leibniz, in many writings, expresses sympathy for Plato’s theory of innate ideas and compares his own views to Plato’s. In Discourse on Metaphysics, e.g., Leibniz writes that “nothing can be taught to us whose ideas we do not already have in our mind” ([Leibniz, 1989], p.58). He adds: This is what Plato so excellently recognized when he proposed his doctrine of reminiscence, a very solid doctrine, provided that it is taken rightly and purged of the error of preexistence and provided that we do not imagine that at some earlier time the soul must already have known and thought distinctly what it learns [...]. [..] our soul knows all these things virtually and requires only attention to recognize truths, and that, consequently, it has at very least, the ideas upon which these truths depend [Ibid., emphasis added. See also [Leibniz, 1981], p.52; pp.78-9 and p.106]
So, dormant or latent ideas lie within our mind waiting to be aroused (or brought to consciousness) by an external stimulus (“Experience is necessary [...] if the soul [...] is to take heed of the ideas that are within us”, [Leibniz, 1981], p.111). Consequently, to a certain extent, Leibniz agrees with Plato that learning is a process of rendering something already there explicit (although he does not concede to Plato that we must have been aware of these ideas in a prior life).
1
Needless to say, the difference between DN and CN (or a DN that implies CN) is connected to the difference between, respectively, a posteriori and a priori knowledge.
262
Raffaella De Rosa
If an idea lies dormant within the mind, (a) the mind has an innate disposition to attain that specific idea and also (b) an external stimulus is necessary to bring that specific idea, so to speak, to the surface. This is how dispositional and cognitive accounts of innateness of ideas merge in Rationalist writings. In point of clarification, I will review passages from Leibniz’s New Essays where it is clear he assumes the two accounts complement one another. In Book I.i, Theophilus (Leibniz’s spokesman) replies to Philalethes’ (Locke’s spokesman) acknowledgement that there could be something in the soul not perceived by it (since one does not always remember everything one has learned) as follows: Why couldn’t you also admit the mind contains ideas as “items of unacquired knowledge” of which it is not aware? ([Leibniz, 1981], p.78). Leibniz has in mind Plato’s account of innate ideas qua items of innate (although latent) knowledge (Ibid.). Philalethes does not admit this, rejoining that ideas within the mind the mind does not perceive can only mean, as far as origin is concerned, the mind has an innate capacity of coming to know them (and this degree of nativism is no threat to Empiricism). To this Empiricist challenge, Theophilus responds: The mind is capable not merely of knowing them, but also of finding them within itself. [...] [Our faculty of understanding does not consist] in a mere possibility of understanding [...] truths: it is rather a disposition, an aptitude, a preformation, which determines our soul and brings it about that they are derivable from it. Just as there is a difference between the shapes which are arbitrarily given to [...] a piece of marble and those which its veins already indicate or are disposed to indicate if the sculptor avails himself of them [Ibid., p.80, my editing and emphasis].
In this last passage, Leibniz slips into his dispositional account of innate ideas, as the analogy with the veined block of marble suggests (For the analogy, see [Leibniz, 1981], p.52). However, the Empiricist challenge (that the only other way in which ideas can be in our mind without being perceived consists in admitting that the mind is innately capable of coming to know ideas) led Leibniz to underpin his dispositionalism with a more robust conception of innateness: the mind does not merely possess the capacity to attain ideas on occasion of experience but also of finding them within itself. Consequently, Leibniz’s analogy of the mind as a veined block of marble (often quoted in support of Leibniz’s DN) takes up a new meaning (This point has been suggested by [Cooper, 1972] and [Kitcher, 1978]). This analogy suggests that, according to Leibniz, ideas themselves are innate dispositions (i.e., latent contents of the mind) that need external stimulation to come to our consciousness (see [Leibniz, 1981], p.74 and p.7). According to this reading of the analogy, ideas are already in us and experience serves only to highlight what was al-
On Fodor’s Claim that Classical Empiricists and Rationalists ...
263
ready there all along (only unconscious) – as the sculptor’s chipping reveals veins already in the marble. As the veins are not yet the figure of Hercules because the chipping of the sculptor has not yet brought that figure to the surface, so too ideas are not yet in the mind (i.e., they are dispositions) because experience has not yet brought them to awareness. This is how to interpret Leibniz’s claim that ideas are innate to us as dispositions of the mind.1 In conclusion, according to Leibniz, the mind is disposed to acquire ideas from experience because the mind already possesses these ideas (qua dispositions) prior to experience. Whether Leibniz dispositionalism reduces to his view of innate ideas as latent items of knowledge is unclear; however, Leibniz did not see an inconsistency between the two accounts of innate ideas. CN and DN co-exist unproblematically in Leibniz’s theory of innate ideas. The same is true for Descartes. Many passages in Descartes, along with those that advocate DN, endorse the view that the child is born possessing ideas (though not occurrent ones). Here are a few examples: In a letter to Hyperaspistes, he writes that a child in its mother’s womb “has in itself the ideas of God, of itself [...] in the same way as adult human beings have these ideas when they are not attending to them, for it does not acquire these ideas later on, as it grows older” (CSMK III, p.190). At the beginning of the Fifth Meditation, Descartes writes with regard to the ideas of extension, number, size: [...] the truth of these matters is so open and so much in harmony with my nature, that on first discovering them it seems that I am not so much learning something new as remembering what I knew before; or it seems like noticing for the first time things which where long present within me although I had never turned my mental gaze on them before [CSM II, p.44, my emphasis].
Again, in the Fifth Set of Replies, Descartes replies to Gassendi: You say you want to [...] ask whether I think the soul always thinks. But why should it not always think, since it is a thinking substance? It is no surprise that we do not remember the thoughts that the soul had in the womb [...], since there are many other thoughts that we equally do not remember, although we know we had them when grown up [...]. So long as the mind is joined to the body, then in order for it to remember thoughts which it had in the past, it is necessary for some traces of them to be im1
In the Preface to [Leibniz, 1981], Leibniz writes: “Although the senses are necessary for all our actual knowledge, they are not sufficient to provide it all [...]”, p.49. See also Book II, where Theophilus (Leibniz) replies to Philatethes’ (Locke’) remark that Theophilus has to grant him that all our ideas come into the mind through either sensation or reflection: “[...] let me say in advance, sir, that when you say that ideas come from one or other of those causes, I shall take that to mean the actual perception of the ideas [....]”, p.111.
264
Raffaella De Rosa printed on the brain; it is by turning to these or applying itself to them, that the mind remembers [CSM II, pp.246-7, my emphasis].
In response to Hobbes’ remark that for an idea to be innate it must be always present to the mind, Descartes replies that for an idea to be innate to the mind means that “we have within ourselves the faculty of summoning up the idea” (CSM II, p.132, emphasis added). Responding to Gassendi, who objected that the idea of a triangle is formed by the intellect by abstraction from particular triangles, Descartes replies: I do not [...] concede that the ideas of these figures ever came into our mind via the senses [...] when in our childhood we first happened to see a triangular figure drawn on paper, it cannot have been this figure that showed us how we should conceive of the true triangle studied by the geometers, since the true triangle is contained in the figure only in the way in which a statue of Mercury is contained in a rough block of wood. But since the idea of the true triangle was already in us, and could be conceived by our mind more easily than the more composite figure of the triangle drawn on paper, when we saw the composite figure we did not apprehend the figure we saw, but rather the true triangle. [...] we could not recognize the geometrical triangle from the diagram on the paper unless our mind already possessed the idea of it from some other source [CSM II, p.262, my emphasis].
We should interpret these passages as follows: the acquisition of the idea of triangle is not the result of a rational process (as Gassendi wants) but of a causal process. Experiencing instances of “triangle” gives the mind occasion to attain the idea of triangle that is already present within it qua (a latent) disposition. All these passages indicate that Descartes, like Leibniz, saw no inconsistency between dispositional and cognitive accounts of innate ideas. The two accounts complement, and entail, one another. Like Leibniz, Descartes holds that the mind is innately disposed to attain certain ideas on the occasion of certain stimuli because ideas themselves are present within the mind qua latent contents that await external stimulation in order to be brought to the mind’s attention. In sum, once it is clear that Descartes’ and Leibniz’s dispositionalism is underpinned by a more robust (or cognitive) Nativism, it follows that, according to the Rationalist: (i)
Ideas are innate in the sense that their intentional content is; moreover, ideas are dispositionally innate in the sense that they are latent contents of the mind awaiting external stimulation in order to become conscious.
On Fodor’s Claim that Classical Empiricists and Rationalists ...
(ii)
(iii)
265
Experience (or encounters with the environment) has the function of triggering ideas already in the mind. Triggering experience is necessary to excite the mind to “form” as an actual conscious state an idea that previously existed in it (though dormant). The mind encounters the environment saddled with a stock of determinate ideas.
So, the Rationalist DN must be seen in light of an endorsement of a more robust version of the innateness of ideas (or CN). According to the Rationalist, the mind is innately disposed to acquire ideas from experience because it possesses ideas (qua dispositions) prior to experience. This move is necessary for the Rationalist if she wants to distance herself from the dispositionalism endorsed by Empiricists and, hence, to defend the claim that (some) ideas are innate. So much for clarifying what Rationalists pack into “the innate disposition of the mind”. I turn now to the Empiricist DN. As we saw in Sections II and III, Locke is prepared to concede that, given the innate structure of the sensorium, the mind is disposed to form certain sensory ideas (rather than others) and that there are innate inclinations to behavior. However, what does acknowledging dispositional nativism amount to? It fails to go beyond acknowledging the innateness of the structure of the sensorium and of combinatorial mechanisms of the mind. In particular, Locke does not pack into “the innate disposition of the mind” innate intentional content. It follows that according to the Empiricist: (i)’
(ii)’
(iii)’
Ideas are not innate in the sense that their intentional content is; rather, some innate mechanism (or capacity) in the mind leads the organism from experiencing Xs to forming the concept of being an X; experience (or interaction with the environment) activates the organism’s innate mechanism (and this activation results in the addition of an idea to the subject’s cognitive stock) rather than triggering the ideas themselves; and the mind comes to encounter the environment provided only with innate mechanisms and capacities.
In short, the difference between the dispositional nativism endorsed by the Rationalist and by the Empiricist is the same as the difference between saying (respectively) A has an idea and is disposed to employ it in conscious thought in condition C and A is disposed to come to have an idea and employ it in conscious thought in condition C.
266
Raffaella De Rosa
Let me draw some conclusions. The difference between (i)-(iii) and (i)’-(iii)’ illustrates the difference in nature (on Rationalist and Empiricist accounts, respectively) of the mind’s contribution in the acquisition of ideas from experience. Rationalist and Empiricist do not disagree, as Fodor wants, only about which ideas are primitive and innate; they disagree on the nature of what is to be admitted as innate in the mind. The Empiricist admits some mechanism or capacity to be innate in the mind, whereas the Rationalist admits that ideas as such (their intentional content) are innate to the mind as well. As a result of the different nature of the mind’s contribution, Rationalists and Empiricists also assign different roles to environmental stimulation in the process of concept acquisition. For the Rationalist, experience triggers (or brings to consciousness) ideas already present within the mind. For the Empiricist, since no intentional content is postulated besides the innateness of a mechanism, the very formation of an idea (i.e., its content) is coeval with the triggering of the mechanism on the occasion of environmental causation. The difference between Empiricist and Rationalist views on the role of environmental stimulation is swept under the rug in Fodor’s claim that both theories “assume that the acquisition of [primitive] concepts is contingent upon the activation of the sensorium”. Fodor is right that for both Rationalists and Empiricists, experience (or “the triggering of the sensorium”) is necessary for the acquisition of concepts. However, there is a difference of sense in which experience is necessary for concept acquisition. According to the Rationalist, the acquisition of primitive concepts is contingent upon the activation of the sensorium in the sense that the environment occasions the availability of the concept to the organism. According to the Empiricist, the acquisition of primitive concepts is contingent upon the activation of the sensorium because the environment contributes the very content of the concept.
Conclusion In Sections III and IV, I established an ambiguity surrounding “innate ideas” and on which, as anticipated in section II above, my criticism of Fodor is based. This ambiguity emerges especially in Section IV, where I clarify what Empiricists and Rationalists mean by DN. As we saw, whereas the Empiricist DN is nothing more than a weak reading of Nativism (and a weak reading of “ideas being in principle being available to us”), the Rationalist DN amounts to a strong reading of “innate ideas” (and of ideas being “in principle available to us”). This evidence cripples Fodor’s inference to the conclusion that Empiricism is committed to the innateness of primitive ideas from his premise that a
On Fodor’s Claim that Classical Empiricists and Rationalists ...
267
hypothesis-testing model of concept learning must acknowledge primitive concepts as in principle available to the subject (i.e., as unlearned). Here is a revised list (in light of Sections III and IV) of my reasons: (i)
(ii)
(iii)
To be faithful to the Empiricist, we must interpret his claim that primitive ideas are in principle available to us in a weak sense only – that is, as admitting that any idea realized in the mind upon activation of an innate mechanism (i.e., the sensorium) in consequence of encounters with the environment is available to the mind. For Fodor’s claim that Empiricists are unwittingly committed to the nativism of primitive ideas to be interesting, we must interpret his conclusion (6) above – viz., that Empiricists are as committed to the innateness of primitive ideas as Rationalists – in the sense that Empiricists are as committed to the existence of innate contents as Rationalists. However, the strong reading of (6) does not follow from the weak reading of (5).
In conclusion, the Empiricist can grant Fodor premises (1)-(5) and agree with him that primitive concepts must be in principle available to the subject without relinquishing Empiricism – that is, without committing himself to a strong Nativist claim that there exist latent innate contents in our mind. “Being unlearned” or “being in principle available to us”, the Empiricist can insist, do not mean “being innate”. All one needs to explain the acquisition of primitive ideas is an innate mechanism that realizes ideas in our minds upon encounters with the environment. On this picture, since the environment does not activate preformed (latent) ideas but rather an innate mechanism, primitive ideas are unlearned (since the activation of the sensorium is not a rational process), but are nonetheless acquired. At least this is how the story goes. However, one may argue that if the innate mechanism triggered by external stimulation is, according to the Empiricist, a merely causal (as opposed to rational) one – that is, if the sensorium causally produces ideas of color, taste and so on (i.e., states with bona fide intentional content) upon external stimulation – how does the Empiricist view really differ from the Rationalist view that experience triggers latent contents? Is this merely a verbal disagreement? Well, maybe the Empiricist ought to have granted the Rationalist that admitting an innate mechanism which on activation immediately causes the subject to have primitive concepts (such as RED) is tantamount to granting that primitive concepts are innate. Leaving aside the issue of whether Empiricists would be prepared to concede that the innate
268
Raffaella De Rosa
mechanism upon whose activation we acquire primitive concepts is brute causal, a priori arguments could be provided for why Empiricists ought to have conceded that (at least some) ideas are innate. However, it is useful to keep in mind that Empiricists did not, in fact, make this concession and to make this clear was the aim of the present paper.
References Adams R. M., “Where do our Ideas Come From? Descartes vs Locke”, in [Stich, 1975], pp.71-87. Armstrong R. L., “Cambridge Platonists and Locke on innate ideas”, in Journal of the History of Ideas, 30, 1969, pp.187-202. Cooper D., “Innateness: Old and New”, in The Philosophical Review, 81, 1972, pp.465-483. Cowie F., What’s Within. Nativism Reconsidered, Oxford University Press, 1999. Descartes R., The Philosophical Writings of Descartes, translated by Cottingham, Stoothoff, Murdoch and Kenny, Cambridge University Press, 1985, Volumes I-III. Fodor J., The Language of Thought, Harvard University Press, 1975. Fodor J., Representations, The Mit Press, Cambridge, Massachussetts, 1981a. Fodor J., “The Present Status of the Innateness Controversy”, in [Fodor, 1981a], 1981b, pp.257316. Fodor J., Concepts. Where Cognitive Science went wrong, Clarendon Press, 1998. Katz J., “Innate Ideas”, in [Stich, 1975], pp.145-163. Kitcher P., “The Nativist’s Dilemma”, in The Philosophical Quarterly, 1978, vol.28, pp.1-16. Greenlee D., “Locke and the controversy over innate ideas”, in Journal of the History of Ideas, 33, 1972, pp.251-264. Leibniz G. W., New Essays On Human Understanding, translated and edited by P. Remnant and J. Bennet, Cambridge University Press, 1981. Leibniz G. W., Philosophical Essays, edited and translated by R. Ariew and D. Garber, Hackett, 1989. Locke J., An Essay concerning Human Understanding, edited by P. H. Nidditch, Clarendon Press, Oxford, 1975. Lowde J., A Discourse Concerning the Nature of Man, (London, 1694; reprinted Garland, New York and London, 1979). Piattelli-Palmarini M., (ed), Language Learning, Cambridge, 1980. Rescher N., “A New Look at the Problem of Innate Ideas”, in British Journal for the Philosophy of Science, 17, 1966, pp.205-218. Samet J., “Troubles with Fodor’s Nativism”, Midwest Studies in Philosophy, X (1986), pp.575-594. Samet, J. and O. Flanagan, “Innate Representations”, in [Silvers, 1989], pp.189-210. Silvers S., (ed.), Rerepresentations, Philosophical Studies Series, 40, Kluwer Academic Publishers, 1989. Schmaltz, T. M., “Descartes on innate ideas, sensation, and scholasticism: the response to Regius”, 1997, in Stewart M. A., (ed.), Oxford Studies in the history of philosophy, vol.2. Stich S. (ed.), Innate Ideas, University of California Press, 1975. Wendler D., “Locke’s acceptance of innate concepts”, in Australasian Journal of Philosophy, 74, 1996, pp.467-483. Wilson M., “Descartes on the Origin of Sensation”, in Philosophical Topics, 1991, pp.293-323.
On Fodor’s Claim that Classical Empiricists and Rationalists ...
269
Wilson M., “Descartes on the Perception of Primary Qualities”, in Voss S (ed), Essays on the Philosophy of Science of Rene Descartes, Oxford University Press, 1993, pp.162-176. Woolhouse R., Locke, University of Minnesota Press, 1983. Yolton J. W., John Locke and the way of Ideas, Oxford University Press, 1956.
270
Consuelo Preti
CONSUELO PRETI Belief and Desire Under The Elms Abstract This paper begins with an exposition the apparent tension between externalist theories of content and common-sense belief/desire psychology, with a view to resolving the conflict between these two views. The second part of the paper is a criticism of Fodor’s reformulation of Twin Earth type cases. I argue that this attempt to mitigate the damage such cases do to the tenability of folk psychological explanation cannot work, because Twin Earth cases pose a metaphysical problem for content and explanation, not a nomological one, as Fodor argues. I discuss this in detail, by arguing that Twin Earth cases are illustrations of a metaphysical appearance/reality distinction for (some kinds of) content. The paper concludes with a strategy for blending content externalism with a robust folk psychology: one that centers on the distinction between causation and causal explanation.
1. Introduction This paper is broadly concerned with the implications of semantic externalism for mental causation and for the nature of folk psychology. I have argued elsewhere that concepts subject to weak externalism pose a threat to folk psychology1. Here I will show that the tension between concepts subject to strong externalism and folk psychology remains, notwithstanding recent arguments marshaled by Fodor (1994). The threat of externalism to folk psychology may therefore appear to be intractable, but I will close by suggesting a strategy for a defense of the mutual tenability of externalism and folk psychology, one that deploys a familiar line of argument in a novel and promising way.
1
See Preti (1998). The difference between so-called strong and weak externalism can be understood by way of a supervenience claim. Strong externalism says that the supervenience base consists of the environment and the causal relations the environment bears to a psychological subject. Weak externalism says that the supervenience base is restricted to the environment impinging on a psychological subject and does not include causal relations.
Belief and Desire Under The Elms
271
2. Content and Causation, or, “If Psychological Processes are Computational, How Can Psychological Laws be Intentional?”(2) Call the following the mutual tenability thesis (MTT): Mental content has casual powers, and mental content is externalistically individuated. Familiarly, there is pressure on MTT for two reasons. First: a naturalistic account of the casual powers of intentional properties is elusive. Second – to add to the difficulty – the fundamental intuitions about causation and the basic tenets of externalism appear to be in direct contradiction. Causation is commonly understood as a relation that holds between entities in virtue of their intrinsic properties, so a causal relation between mental states and action will have to be a relation between intrinsic properties of mental states and the intrinsic properties of action. It is reasonable to suppose that these properties will be the intrinsic physical properties of the central nervous system, which are what they are independently of the world extrinsic to the psychological subject. And the first puzzle is how propositional attitudes such as belief could have causal powers that affect (and effect) action. Any property will need to meet certain conditions of efficacy (instantiation, for example), in order to play a causally relevant role in a system. Our best theories about causation purport to be accounts of how causal connections obtain between naturalistic properties of the entities in a causal relation, and there is still widespread controversy in the philosophy of mind over how it is that intentional properties are materially implemented. If a mental state is conceived of wholly physically, then how can it have intentionality? But if it isn’t, then how can it have causal powers? Now apart from the problem of explaining how any wholly material entity or process can be understood to implement semantic properties like content, there is the further additional difficulty generated by the externalistic individuation of content. According to externalism, the content of a mental state is a relation to the particulars and properties in the world external to the psychological subject. The thesis is provocative precisely because it claims that the mind itself is (partly) constituted by the mind-independent world of particulars and properties. But if the existence-and-identity conditions of a propositional attitude are those of its objects, then it difficult to see how a mental state satisfies the plausible conditions on causal efficacy. Externalism in effect pulls
272
Consuelo Preti
content apart from causation, by showing that sameness of intrinsic properties is not sufficient for the sameness of intentional ones1. The puzzle of accounting for the causal properties of the intentional is only one aspect of the general issue. The other concerns the character of psychological explanation. The claims of folk psychology, for example, refer to what a person believes to explain why she did what she did. But the attraction of the claims of folk psychology is that the etiology of action devolves on these properties, and that to cite them in explanation is to cite causally relevant explanatory properties of action, a claim that is threatened by the externalistic individuation of content.
3. Robust Folk Psychology One point of formulation: I think there is a way to spare folk psychology from the threat of externalism, and, most important, to do this in such a way that is consistent with a stalwart commitment to folk psychology as genuinely intentional, genuinely causal, and genuinely explanatory. I emphasize this to make the point that some formulations of folk psychology could (and do) forswear one or more of those elements. What I will defend as robust folk psychology (hereafter RFP), however, is to be understood as follows: RFP is set of naturalistic/scientific causal generalizations whose explanatory features are modeled on the covering-law or deductive-nomological model of scientific explanation. The claims of RFP are (held to be) causally explanatory and predictive in a naturalistic sense. Most importantly, the explanatory claims of RFP advert to the content of mental states – content bears the explanatory burden. Now, given that the intuition the RFP has explanatory virtues worth defending, notwithstanding the tension generated by externalism, the so-called causal role problem is a considerable obstacle. But I believe it appears more intractable than it really is, and I will argue below that the problem is generated by the tendency to formulate the necessary conditions of RFP too inclusively. How much do we really need to stay true to the robust conception folk psychology as genuinely intentional science? We may need less than we think. I shall argue in closing that a new look at the old debate concerning reasons and causes may help to resolve the problem of content in explanation2. But first we must make clear the severity of the threat to RFP from externalism. 1
At least for concepts subject to strong externalism, which is our concern here. The situation is more complex for concepts subject to weak externalism (cf. Preti, 1998).
2
This is not the place to argue for the independent plausibility of folk psychology. This is a subject treated extensively in literature; see (Guttenplan, 1994) for an overview.
Belief and Desire Under The Elms
273
4. Nomological Supervenience and Content-Individuation Fodor’s claims (1994) are a useful perspective from which to focus on the threat to RFP posed from strongly externalist content individuation1. I will argue here that his arguments do not blunt the threat to folk psychology coming from Twin Earth cases, by showing that his attempt to defuse Twin Earth cases by emphasizing the nomological conditions for their obtaining does not succeed. This is because (as I will defend) Twin Earth cases are ways of highlighting that an appearance/reality distinction is what is at work in the content-individuation of strongly externalist concepts. The threat to RFP is a metaphysical one, which Fodor’s reformulation does not address. In (Fodor 1994), Fodor claims that there are no Twin Earth cases of the sort that could genuinely threaten the claims of folk psychology. For Fodor, twin cases are cases that, among other things, show that content can come apart from causation, since they show that the same implementing mechanism (some internal computational state, say) is metaphysically insufficient to guarantee the same intentional state. As Fodor rightly sees it, Twin cases separate the key properties that underpin the virtues of the explanatory claims of RFP. So for RFP to be true is for these cases to fail to obtain (Fodor 1994, 24-25). This way of putting it, of course, makes the argument straightforwardly vulnerable, by contraposition, to the rebuttal that should Twin Earth cases be shown to proliferate after all, then RFP remains vulnerable. I argue in what follows that in fact they do proliferate quite naturally. What Fodor does is to reformulate the Twin Earth thought-experiments with an emphasis on the nomological, rather than metaphysical, conditions that govern their obtaining. His position is that there are no laws that require or necessitate Twins, because there are no laws that require conceptdistinguishability even under conditions where kinds may be distinct from one another. So, he claims, if there are no laws necessitating Twin cases then Twin cases are accidents. No respectable folk psychology is under any obligation to cover my mental states and any accidentally divergent mental state of any of my molecular duplicates. And as a result, folk psychology can withstand the purported threat of externalism2. 1
It goes without saying that Fodor is the most ardent defender of RFP. Stich (1991), among others, is a useful start for the case against RFP.
2
It could be that Fodor is conflating: (1) ~ (∃L) (There are twins) with (2) (∃L) ~ (There are twins) The stronger claim is (2), but doesn’t follow from (1), nor is it obvious that it could be
274
Consuelo Preti
The upshot of what Fodor calls Putnam’s story (Fodor 1994, 28) is that [it] shows something about our concept of content: it shows that the supervenience of the broadly intentional upon the computational isn’t conceptually necessary. But is doesn’t argue against the nomological supervenience of broad content on computation…
But Fodor’s claims with respect to Twin Earth cases, as formulated in terms of nomological supervenience of the intentional on the computational are too weak1. Recasting the issue in terms of nomological supervenience of the intentional on the computational cannot meet the genuine problem, because, as I show below, it is precisely the point of Twin Earth cases to show that your concept C and that of your twin can be distinct irrespective of the nomological/empirical conditions that govern your dispositions to apply C (Fodor 1994, 30). It may be true that Twin Earth cases may show something about our concept of content, or about the conditions that apply to our abilities for concept-discrimination. But the relevant issue with respect to the role of content in explanation is what Twin Earth cases entail about the nature of mental content (in Russellian terms, about propositional constituency). The heart of the issue remains a metaphysical one, Fodor’s redirection of emphasis notwithstanding2.
defended. 1
It looks a bit as if Fodor is just smuggling in narrow content in disguise, in spite of his claims in (1994) that narrow content is now unnecessary – by defending a kind of narrow nomological content. Fodor claims that there is no obstacle to claiming that nomological similarity between me and my Twin fixes our intentional similarity. But this seems to turn depend on her presence in my environment or one where (as Fodor sees it) the same nomological regularity of concept discrimination and deployment rules. Note the familiar counterfactual formulation that he used in his defense of narrow content in (1988). It’s still an argument to causal powers, but this time the causal powers attach to the nomological regularity with which we make concept discriminations.
2
I should say here the erroneous picture may be due in part to Fodor’s view about informational semantics, about which he is inconsistent. On the one hand, he claims that (Fodor 1994, 4-5): The basic idea is this: The content of a thought depends on its external relations; on the way that the thought is related to the world, not the way it is related to other thoughts… But on the other (Fodor 1994, 37): Semantics, according to the informational view, is mostly about counterfactuals; what counts for the identity of my concepts is not what I do distinguish but what I could distinguish if I cared to (interalia, what I could distinguish by exploiting instruments and experts). I think Fodor tends to slide between nomological/epistemological considerations and semantic ones in his zeal to counter holism in semantics (and maybe for other reasons, too). And while I agree with him about holism, I think this preoccupation steers him wrong on externalism and mental causation.
Belief and Desire Under The Elms
275
5. Twin Earth: Appearance and Reality It bears repeating that Twin Earth cases are a device in an argument that more generally concerns what the determining conditions might be for meaning and, ultimately, the nature of mind – the externalist thesis that it is the world outside the subject’s head necessarily determines mental content. I wish to suggest that the true essence of a Twin Earth case – its relevance to the issues arise concerning content-individuation – is what it illustrates about the relation between mind and world. And this is that the way the world appears is not always the way the world is – but that, significantly, it is the way the world is that governs the existence-and-identity conditions of mental content. Twin Earth cases, in short, should be thought of as a vivid way of showing how the head-independent environment has a life of its own with respect to contentindividuation. The head-independent environment amenable to Twin Earth cases is so amenable because it is subject to an appearance/reality distinction, a pervasive and systematic characteristic of all worlds that contain psychological subjects. (i) Natural kind concepts The natural kind concept Twin Earth cases – water/xyz; jade/jadeite; aluminum/molybdenum; elm/beech – are the simplest illustrations of the point that part of the head-independent environment is subject to an appearance/reality distinction which can affect the individuation of mental content. Recall that Putnam formulates the thought experiment so as to undermine that Fregean semantic theory, which he achieves by showing that, to take one example, even identical descriptions associated with the term “water” on both Earth and Twin Earth, generated by the appearance of the stuff in each environment, is insufficient to guarantee that both Earthlings and Twin Earthlings mean the same thing by their term “water.” And this is because what looks like water is not necessarily water. XYZ presents the same phenomenological appearance to me, and to my twin, as H20 does, but does not have the same chemical composition, so it isn’t the same stuff as the stuff I call “water.” My twin and I use the same term (“water”) to apply to what surrounds us in our respective environments, but it doesn’t mean the same thing because it doesn’t apply to the same stuff1. This point holds likewise for the jade/jadeite case (Putnam 1975, 241). Putnam describes this case as involving two very different microstructures that produce the same surface properties; here the exploitation of an appear1
See Putnam (1975).
276
Consuelo Preti
ance/reality distinction is clear. Neither the aluminum/molybdenum case nor the elm/beech case is fundamentally different, although these introduce the notion of deference to experts that we will discuss separately below. For now, note that, as stipulated by Putnam (1975, 225-6), the only people who can tell that ‘aluminum’ refers to molybdenum on Twin Earth are (both sets of) metallurgists, who are in possession of additional facts about metallurgical reality; the rest of us, by stipulation, cannot distinguish the two elements. And similarly with the elm/beech case, as described by Putnam (1975, 226): “Suppose you are like me and cannot tell an elm from a beech tree…we still say the extension of ‘elm’ in my idiolect is the set of all elm trees.” (ii) Artifact concepts Artifact concepts like “pencil” (Putnam 1975, 242-43), also fall neatly in line. Earth-pencils can’t be organisms, because earth-pencils are necessarily artifacts. Another world, however, might contain pencil-counterparts whose microstructure is that of an organism, but whose macrostructure is phenomenologically indistinguishable from the artifacts that we use in this world. In that world, the term “pencil” refers to pencil-counterparts which are organisms; in our world, “pencil” refers to an artifact. (iii) Indexical concepts Putnam (1975, 234) explicitly considers the distinction between the phenomenological properties that arise in employing the concept “I” to refer to oneself, and the particular to whom one refers. Indexicals such as “I” have context-dependent reference; each use of “I” by a different person refers to a different person and so yields a different meaning, according to externalism. Yet each time one uses the concept to refer to oneself, one presents oneself with the same appearance as does one’s twin; that of oneself. Same appearance; different reality. (iv) Deferential concepts (expert cases)
Fodor’s (1994) position on deferential concepts is somewhat murky, in my view because he may be implicitly grappling with the consequences of externalism for the intuition that we have privileged access to our mental states, and the upshot of this issue for the retention of RFP. We can assess Fodor’s claims here without going too far afield, however, by noting that there is indeed an issue here as to what commitments folk psychology makes (or
Belief and Desire Under The Elms
277
ought to make) with respect to whether I can introspectively distinguish my own concepts, and how externalism bears on this. Remember that according to the usual story, I myself am in the same computational/intrinsic state whether I am thinking about elms or beeches, but elm-thoughts are about elms, not beeches, and thinking things about elms is not thinking things about beeches. So there appears to be a problem: I am not able to distinguish by introspection my elm-thoughts and my beechthoughts, because what distinguishes them is a question of what botanical environment my thoughts are directed onto, something to which I do not have introspective access. Fodor claims instead that my failure to discriminate concepts in cases like these is an epistemological failure, and claims that such a failure is no threat to folk psychology. His reasoning looks to be as follows: semantically speaking, elms individuate elm-thoughts according to the principles of externalistic content-individuation. If I can’t tell elms from beeches, however, the problem of concept-distinguishability arises: I can’t distinguish my own concepts of elm and beech, but externalism claims they are distinct. Generalizations fail; and RFP is under siege. But Fodor claims I can tell the difference between elms and beeches, and so I do have the ability to distinguish my elm-thoughts from my beech-thoughts: what I do is, I consult the experts. It is just false, on his view, to claim that I can’t distinguish concepts that are distinct. My own particular abilities may be limited, but in the wider community there are experts whose distinguishing abilities I can help myself to, or, as he puts it, co-opt (Fodor 1994, 36). Deferential concept cases, for Fodor, are cases where I can make the requisite conceptual distinctions, in the sense that I get the experts to do the work for me. So externalist principles of content-individuation do not bear so threateningly on the claims of folk psychology when it comes to deferential concepts, precisely because they are deferential1. There are a number of objections to this. Arguably, one of the things that is subject to pressures from externalism is my ability to discriminate my own thoughts with no reference to the external environment. As many have noted2 the issue of privileged access is thrown into relief in the debates on externalism 1
Fodor could be following Putnam’s line of argument here (Putnam, 1975), which introduces the so-called socio-linguistic hypothesis about the division of linguistic labor with respect to recognizing the reference of such terms as “elm,” and “aluminum.” My objection to this is that to the shift the emphasis to method of recognizing reference is ultimately of no use against the problem posed by externalistic individuation of content.
2
See (Burge 1988) for a start; and additional discussion by, among others, Boghossian (1989); Brueckner (1992); McKinsey (1991); Warfield (1992).
278
Consuelo Preti
and its consequences. A further element of this issue, as Fodor is aware, bears on the intuitive case for the robust conception of folk psychology. Part of that conception – possibly implicit – seems to be we do have privileged access to our own mental states. We (traditionally) know when we are in pain, we know what kinds of intentional perceptual states we are in, and, further, that we know that we have beliefs and desires and we know what they are1. Fodor counters the putative threat with an emphasis on the deferential concept cases. My own (externalistically individuated) thought might be introspectively closed to me, he claims, but I can thwart this difficulty by consulting experts. My distinguishing abilities are more than just enhanced by theirs; my mental states expand to include those of the experts to which I defer2. But now we face a dilemma (horns (H1) and (H2)). Externalism threatens privileged access and folk psychology. To save it, I build into my mental states deference to the mental states of experts (among other things). But then this suggests that the mental states of others are somehow individuatively part and parcel of my own. So (H1): either I don’t have privileged access to my own mental states or, remarkably (H2): I have, in effect, privileged access to the mental states of others. If the mental states of experts are “co-opted” by mine in the strong way that Fodor claims – their ability to make distinctions is now mine – then my mental states include the mental states of others. In short, either we have first-person acquaintance with the mental states of others as well as our own or we don’t have first-person acquaintance with our own mental states3. The concession to make to this, it seems to me, is to admit that “co-opting” the mental states of experts is not so strong a claim as to lead to H1 and H2. But to weaken the claim puts it in line with the claims we made above on the nature of the individuation of natural kind, indexical, and artifact concepts. If the mental states of experts impinge on the existence-and-identity conditions 1 The question as to whether a robust formulation of folk psychology need be based on Cartesian intuitions about the mind is one to pursue elsewhere. 2
This suggestion is also open to a reductio in that it appears to trespass illegitimately on the notion of ability. For instance, it seems absurd on the face of it to claim that my partner has the ability to have a baby because he can defer the baby-having to a baby-having expert (i.e. a mother, even the mother of his child). For Fodor to be right here, my ability to make conceptual discriminations has to be different from this sort of ability, and it has to be different enough to make my deference to experts plausible in the right kind of way.
3
This point, incidentally, would apply to Burge’s claims that the social environment also plays a role in the individuation of my mental states; not surprisingly, since Burge’s point is an extension of Putnam’s socio-linguistic hypothesis (Burge 1975). But Burge (wisely, it turns out) makes the weaker point, and so evades the dilemma.
Belief and Desire Under The Elms
279
of my mental state whether or not I can tell, then the situation is precisely that of canonical externalist content-individuation. It is the world outside my head that independently settles the question of what mental states I possess1. There are other objections to Fodor’s treatment of deferential concept cases. For instance: there is no reason to believe that experts are less subject to the pressures of externalist content-individuation than the rest of us. For one thing, they are subject like anyone else to appearance/reality distinctions. Suppose there are entities whose macrostructure is identical but whose microstructure is opaque in principle – then there would be entities for which there could be no expert discriminators2. So by what means could experts mitigate the problem of distinct entities that can’t be distinguished? And what of the mechanics of deference? Is there any reason to suppose that we can in principle rule out a chain of deference between me and my expert? Suppose the chain were to telescope into an infinite regression of deference? These points seem to undermine the kind of deferential relation that is bearing the weight here.
6. Summary If I am right, then Fodor’s attempt to refashion Twin Earth cases by way of nomological supervenience is insufficient to blunt the threat of externalism to folk psychology. That threat, as I spelled out, instead comes from the underlying metaphysical state-of-affairs that Twin cases illustrate. The conditions that bear on the application of our concepts, or our discriminatory capacities, may well be subject to all sorts of interference, and some of the interference may well be “accidental, fortuitous, or wildly exceptional”3. But the disruptive 1
The nature of the world outside of my head should be understood to encompass properties, particulars, and the mental states of others.
2
See (McGinn 1991).
3
I think there is an interesting analogous point to be made with respect some of the discussions of personal identity familiar in the literature. It is sometimes claimed that it is tiresomely speculative to erect thought-experiments-interruptions of consciousness, split-brain experiments and the like-in order to fix intuitions about personal identity (Cf. For example Wiggins, 1980, who claims that the self is a natural kind). This kind of point could be formulated as follows: To fix the nature of the self we ought to take advantage of the fact that selves supervene on human beings with local nomological regularity. On this folk-personalidentity-theory, our taxonomic criteria are commonsense, local and empirical. This approach is not entirely unattractive. I think, however, that it is unphilosophical. The whole point about selfhood is whether such a thing as a self can be attributed to entities in the light of criteria that transcend the contingencies – regular and general though they may be – that obtain in the actual world.
280
Consuelo Preti
conditions that may in principle negatively affect cases of concept-application and concept-distinguishability are ultimately irrelevant to the problem of content in explanation. Those latter are epistemological concerns; the problem of content in explanation is a metaphysical one. We can support this claim by way of a broader question – relevant to all sciences – as to what kinds are to be relevantly referred to in an explanatory theory. RFP, as we know, requires its kinds to be intentional. But then it is reasonable to expect that the taxonomical principles that will be at work to deliver intentional kinds in the right way will be more than a local nomological matter. When content-individuation is externalist, we must ask: can contentindividuation be simply a matter of the nomological properties (or lack thereof) at work in local/contingent examples of head-world relations? Or does content individuation need more genuinely constitutive conditions to be satisfied? On Fodor’s account, my Twin and I are internally/computationally identical, and what varies, as he sees it, are the nomological conditions that govern the application of our respective concepts. These conditions are fortuitous, depending as they do on the accidental local variations in the substance environment impinging on my mental states. So, as Fodor would have it, the important issue is not really that content varies between twins; but that what varies are the conditions of content-deployment, so to speak. And folk psychology is under no obligation to generalize over such local contingent matters. We can go back to Putnam’s original example in (1975) to support, in contrast, the view that the significance of externalist content-individuation (and its consequences for the nature of the explanatory concepts of RFP) is in its character as a metaphysically constitutive thesis. Putnam (1975, 224-25) is clear that the natural kind environment (among others) bears a constitutive relation to the mental states of those in relation to it, and it does so whether or not they can tell (cf. his remarks on what ‘water’ meant in 1750). When Fodor claims that there are no laws that require concept distinguishability (even) where concepts are distinct, I have argued that he is missing the point. It may well be that we will be able to tell that our concepts are distinct only under certain contingent conditions, but the point is irrelevant to the problem at I can’t help but suspect something like this may be at work in Fodor’s position here. And what I’ve tried to show, on the contrary, is that there is nothing farfetched, anomalous or unconstrainedly speculative about Twin Earth though-experiments. What underlies them, and what leads to the problem for content in explanation, is a perfectly straightforward and uncontroversial appearance/reality distinction, an attribute of the very nature of worldliness that is at work in the issue.
Belief and Desire Under The Elms
281
hand. We have to face the externalist thesis squarely: our concepts are individuated by factors independent of our capacities to distinguish, discriminate and apply them. Are externalism’s consequences for the claims of folk psychology therefore insurmountable?
7. Retaining Folk Psychology Suppose that we are not convinced that either robust folk psychology or externalism should be abandoned in spite of the problems that so far seem to plague MTT. It may be that to solve the problem we shall need to approach it from another angle. My concluding remarks here will take the form of two brief discussions: one on the nature of the proposed strategy; the other, on how it could work. (i) Causal relations and Causal Explanations We can start by looking squarely at the source of the impasse. The difficulty of reconciling wide content with RFP appears to arise from taking content-based psychological explanation as a species of causal explanation, in virtue of its reference to the intentional properties of mental states. But, as we have seen, content-adverting explanation fails to capture the causal-explanatory properties of content. So what we need is a strategy which will enable us to consistently hold these theses:
& Intentional mental states have causal powers & Intentional mental state are externalist & Psychological explanation is intentional without falling afoul of the problem. And the way to do this may be to abandon the condition that the intentional vocabulary of RFP need be casually explanatory. It is a familiar point that there are (at least) two explanatory roles that an intentional mental state might bear vis-à-vis an action: rationalization and causation. Rationalization, a normative/justificatory enterprise, situates the mental state in question within a logical network of other mental states possessed, and purports to reveal their logical interconnections. But, familiarly, rationalization is not enough, because we may (say) possess all the beliefs and desires that would make a particular action reasonable and still do not do it (akrasia); or ultimately, act for different reasons. Rationalization alone can fail to fully explain why an action was performed, so there appears to be need for
282
Consuelo Preti
a further link between mental states and actions. That link is commonly thought to be a causal one. Now Davidson’s claim that rationalization is a species of causal explanation (Davidson 1980) was an attempt to render compatible the distinct explanatory properties attributed to reasons. But Davidson’s defense of that thesis is obscured by the fact that it appears to proceed by way of the claim that reasons cause action. The former, however, doesn’t follow from the latter. There are familiar logical gaps in the move from the thesis that reasons cause actions to the thesis that reasons causally explain actions. We know that: & Reasons must cause action to explain action; but just because reasons cause actions, it won’t follow that citing reasons explains action; & If a reason causes an action and justifies it, it won’t follow that citing that reason causally explains the action; & Reasons might explain actions by justifying them, and reasons can cause actions; but it doesn’t follow that reasons causally explain actions. Causation is a relation between events however described, but causal explanation is a relation between events under a description. So the way events are described will affect the truth-conditions of the explanation-sentence, quite independently of the causation relation borne by those events as they are in themselves. Now the distinction between causation and explanation is commonly thought to engender more problems than it solves. I think instead that that very distinction is potentially our best hope to preserve the mutual consistency of externalism and robust folk psychology. The proposed strategy – call it neo-Davidsonian – exploits the difference between causal relations and causal explanations head on. The claim is that mental states cause actions, but the intentional properties of mental states, while explanatory, are not causally explanatory. Intentional mental states enter into genuine causal relations, but we cannot causally explain their doing so by adverting to their intentional properties. The significant consideration in favor of this strategy is that it has more promising implications for the threat posed by externalism to the claims of folk psychology. There seems to be no initial prohibition on this view to the claim that all causally relevant properties in cases of mental causation are intrinsic to the agent. Intentional properties, however, can be individuated relationally or externalistically all the same.
Belief and Desire Under The Elms
283
My concluding remarks will sketch how this strategy could work without sacrificing RFP. The aim, again, is to make a case for the mutual tenability of externalism and a robust folk psychology, the latter formulated as a set of lawlike claims that are at once intentional, causal, and explanatory, in spite of the fact that those conditions appear to be the source of mutual inconsistency. But are they? Let’s consider how strong a claim the defender of MTT must make to preserve RFP. The pretheoretical intuitions to be preserved – the ones doing the work – appear to be two: first, that propositional attitudes are causes with contents (Fodor 1986, 24); second, that the intentional vocabulary of folk psychology provides a naturalistic and ineliminable taxonomy for genuinely explanatory psychological statement. What is doubtful is that the preservation of a robust conception of folk psychology really needs the condition that its intentional claims be causally explanatory. To suppose so, I would suggest, might be the result of taking the threat of epiphenomenalism (too) seriously and as a result, conflating two claims: (1) that intentional properties are “causally responsible;” and that1 referring to these properties under intentional descriptions is causally explanatory. Epiphenomenalism can’t really threaten robust intentional causal realism on the score of intentional description. We cannot take seriously an argument to the effect that intentional properties do not or cannot have casual powers unless referring to them qua intentional yields causally explanatory claims. If intentional properties have causal powers they will have them quite independent of choices of vocabulary. What makes this plausible – I would claim even irresistible – for a defense of RFP, is a commitment to distinct levels of conceptual explanatory taxonomy. Now Fodor (for instance) has never quailed from the position that the intentional properties of mental states will need, by the principles of natural causation, implementation in a casually plausible mechanism. But that suggests that we can get what we’re looking for. Again, strictly speaking, all RFP needs is: (1) generalizations that advert to content; (2) generalizations that advert to content and are explanatory; and (3) causally responsible intentional properties. The implementation of intentional properties in, say, computational ones grants to them the desired causal role. So to advert to intentional properties is genuinely explanatory. Adverting to them qua intentional may block causally transparent reference to the implementing mechanism, but the prior condition on RFP – the implementation of the intentional in the causal/intrinsic – makes
1
The words are Fodor’s (Fodor 1994, 1). For more discussion of this issue, see Mind and Language (Vol. 12, Numbers 3 and 4), September/December 1997.
284
Consuelo Preti
this no sacrifice that hasn’t already been made. It may be that this is a small price to pay for a plausible defense of the mutual tenability thesis. I am grateful to Jerry Fodor and Colin McGinn for discussion of the ideas in this paper.
References Bogosshian, P., 1989. “Content and Self-Knowledge,” Philosophical Topics 17 (1989): 5-26. Brueckner, A., 1992. “What an Anti-individualist Knows A Priori,” Analysis 52 (1992): 111-18. Burge, T. 1979. “Individualism and the Mental,” Midwest Studies in Philosophy, vol. IV., (Minneapolis: University of Minnesota Press): 73-121. Burge, T., 1988. “Individualism and Self-Knowledge,” Journal of Philosophy 85 (1988): 649-63. Davidson, D. 1980. “Actions, Reasons, and Causes,” Essays on Actions and Events (Oxford: Oxford University Press): 3-19. Fodor, J., 1994. The Elm and the Expert (Cambridge, MA: MIT Press). Fodor, J., 1988. Psychosemantics (Cambridge, MA: MIT Press). Guttenplan, S. 1994. A Companion to the Philosophy of Mind (Oxford: Basil Blackwell). McGinn, C., 1991. The Problem of Consciousness (Oxford: Basil Blackwell). McKinsey, M., 1991. “Anti-Individualism and Priviledged Access,” Analysis 51 (1991): 9-16. Preti, C., 1998. “The Irrelevance of Supervenience,” in Protosociology: Cognitive Semantics II (Externalism in Debate). Dordrecht: Kluwer (160-172). Putnam, H., 1975. “The Meaning of ‘Meaning,’” in Mind, Language and Reality, Philosophical Papers, Vol. II (Cambridge: Cambridge University Press): 214-271. Stich, S. 1991. From Folk Psychology to Cognitive Science: The Case Against Belief (Cambridge: MIT Press). Warfield, T. 1992. “Priviledged self-knowledge and externalism are compatible,” Analysis (Vol.52, No.4), October, 1992: 232-37 Wiggins, D. 1980. Sameness and Substance (Oxford: Basil Blackwell).
On David Lewis’ Philosophy of Mind
285
ERWIN ROGLER On David Lewis’ Philosophy of Mind Abstract Are there eliminativist tendencies in Lewis’s theory of mind? Prima facie one would like to give a negative answer to this question. Lewis (1994) conceives his theory as “Reduction of Mind”. Certainly, both reduction and elimination of mental states are regarded as materialist, yet nevertheless as competive strategies. Relying on folk psychology (FP), as Lewis does, is objected to by eliminativists who denounce FP mainly because they think it is a theory that is essentially wrong. Yet, Lewis sees the importance of FP residing only in its causal schemes that explain behavior and not in its role of specifying internal states by specific mental properties. Should therefore the reduction of mental states not be considered as eliminativist in a certain sense? There is an ambiguity in these concepts that forbids an immediate answer. The problem will be elaborated in section II and III with regard to intentional and qualitative mental states. As a preliminary tho that I discuss in section I central aspects and problems of Lewis’ theory of mind.
I
Ramsification, Reduction, Multirealization
a) Some Fundamentals of Lewis’ Theory of Mind Lewis’ theory is made of two theoretical steps: a semantic-functionalist analysis of mental concepts as the first step and (faithful prospects for) psychophysical identifications as the second (Lewis 1966, 1970, 1972, 1980). The point of departure of his theory is FP. Its “platitudes” concern mainly causal relations among mental states, mental processes or events1 as well as sensory stimuli (inputs) and behavior (outputs). In the sense of Sellars’ myth of our Rylean ancestors, FP is conceived as a term-introducing theory for mental concepts. Mental concepts are taken to be theoretical (T-) terms which FP implicitly defines. Input and output vocabulary, as well as the notion of causality, make up the already understood and unproblematic O-terms. FP sentences containing mental terms M1,...,Mn – or perhaps, disjunctions of those sentences for neutralizing false sentences – are conjoined via logical conjunction into one single sentence: T(M). Replacing all M-predicates M1,...,Mn 1
Differences between these concepts can be neglected here.
286
Erwin Rogler
uniformly by variables m1,...,mn and existential quantification (with respect to all those variables) leads to the Ramsey sentence }mT(m) and the modified Ramsey sentence }!mT(m) of T(M) (abbr.: RT(M), R!T(M)). The Ramsey sentence says that there is at least one, the modified Ramsey sentence there is exactly one, model of T(M). By means of R!T(M) Lewis construes explicit functional definitions of the M-terms: (1)
M1 =def k1 }k2 ,...,kn ~m1,...,mn[T(m1,...,mn) kn=mn] and accordingly for M2, M3,...,Mn.
k1=m1
Y ,..., Y
At the heart of Lewis’ analysis lies a “functional specification view” (Block 1980a:179ff). According to this view functional state types are to be identified with physical, especially neuronal state types. This is step 2 in Lewis’ theory. If one concedes the possibility of physical multirealization (MR), then Mconcepts defined in the style of (1) cannot, due to the transitivity of identity, be rigid designators. Block (ib.) on the other hand defines mental concepts according to a “functional state identity view”: (2)
M1* =def y}m1 ,..., mn[T(m1,...,mn) Y y has m1] and accordingly for M2*,..., Mn*.
Here, in contrast to scheme (1) a mental term “Mi*” denotes a class of objects (actual and possible objects) which occupy a specific causal role. As far as I can see, Lewis (1970: 87, 1994: 420) expresses functional concepts according to the schemes (1) and (2) by means of the predicates “M” and “being in M” for “M*”; the latter are interpreted as rigid designators.1 Therefore, if one allows the possibility of MR, the classes denoted by them cannot be identical with single nonfunctional state types. The first step in Lewis’ theory is often dubbed “a priori” or “analytical” functionalism (Block 1978: 271ff). But this can be understood in two ways. In one sense Lewis’ introduction of mental terms via functional definitions is taken in contrast to mere functional characterizations of psychological states. These then are conceived as empirical hypotheses, whereby psychological states are already determined by means of other entities, for example their phenomenological properties. Understood that way, Lewis’ analyses clearly
1
Definitions according to scheme (2) should be distinguished from another type of functional definitition:
(2’) M =GHI k }m, ..., mQ[T(m, ..., mQ) Y k=m]
and accordingly for M ,..., MQ
The M -terms in scheme (2) are of order two and of level one, whereas the M -terms in scheme (2’) are of order two and of level two as well.
On David Lewis’ Philosophy of Mind
287
belong to analytical functionalism. But this dubbing is also chosen to indicate that the underlying psychological theory is taken to be analytical in contrast to psychofunctionalism, a proposal which directly starts from scientific psychology as an empirical theory. If Lewis were an analytical functionalist in this second sense as well, he would have to interpret FP as a set of analytical truths. In (Lewis 1972: 213) he writes: “There is a strong odor of analyticity about the platitudes of commonsense psychology.” But presumably this means that the platitudes of FP should on the whole be taken as largely secure factual truths, and not as truths without factual content. Briefly stated not the FP-sentences themselves are considered to be analytic but “that either pain etc. do not exist or most of our platitudes about them are true.” (ib.). To express this a little more generally, the analytical component of FP (T(M)) is what is usually called its Carnap sentence (RT(M) T(M)). The factual content of T(M) on the other hand is given by RT(M). T(M) then is L-equivalent with the conjunction of both RT(M) and RT(M) T(M). Therefore it seems to be clear that Lewis is not an analytical functionalist in the second sense. Functional definitions of mental state types by means of their causal roles are considered to be ontologically neutral, i.e. they tell nothing about the nonfunctional properties of their tokens. It’s a contingent fact, whether, and if yes, what sort of entities take in certain causal roles. But if it turns out that e.g. the pain role is occupied by states of a certain typ (e.g. C-fibre firing), then by way of its functional definition this type must be identical with the functional type of pain. Briefly put: the realization of functional state types is by logical reasons identification. (op.cit.:211/12).1 From both premisses, (a) theoretical terms M1,...,Mn of T(M) are interpreted according to (1), and (b) exactly one ntupel of non-functional state types (P1,...,Pn) uniquely satisfies T(M), the contingent statements of identity P1=M1,...,Pn=Mn can be deduced. These theoretical identifications are specified by Lewis in several ways: (I) Motivated by the development of latter day physics Lewis endorses the metaphysical hypothesis of contingent materialism for the actual world. That means “roughly speaking (...) that physics – something not too different from present-day physics, though presumably somewhat improved – is a comprehensive theory of the world, complete as well as correct. The world is as physics says it is, and there’s no more to say.” (Lewis 1983a:361). Functional types therefore can only be realized by physical types in our world. Exchanging in T(M) P-terms for the M-terms yields a theory stated solely in physical terms: T(P). Conceptual reduction of that sort does not automatically answer 1
Another proposal looks for merely methodological identifications, for instance, if mental terms are not interpreted functionally, as for example in H. Feigl’s view.
288
Erwin Rogler
the further question, whether T(P) can additionally be reduced completely to physical laws. However, Lewis’ materialist hypothesis postulates that, in taking physics as the universal science, at least in principle also a reduction of laws is envisaged. (II) The theoretical terms of a theory T only denote entities which will satisfy T uniquely, whereby approximate satisfaction will be sufficient. They don’t denote in cases where there is no model of T nor do they denote in cases where there is more than one model. (III) The unique realization of T(M) should be understood relative to certain classes of structures; therefore MR is not precluded. Lewis considers several candidates for such structures, among which only empirical investigations might decide: possible worlds, populations within a world, especially species, sometimes even particular individuals (and some counterparts thereof in possible worlds) (Lewis 1969:25, 1974:120, 1994:420). It is always restricted type-type identities that matter for Lewis.1 (IV) As a consequence of both steps in Lewis’ theory there are two different ways of describing mental phenomena: descriptions in functional and descriptions in physical terms. Assuming type-type identities to be restricted to species e.g. humans, then functional pain for humans (Mi) will be realized by a specific neuronal state type (Pn). Lewis (1980) considers special cases. One deals with the possibility of humans who are in state Pn but not in state Mi (mad pain). The other deals with the possibility of creatures having a hydraulic mind which differ biologically but not functionally from humans. For such a being, therefore, the causal role by which Mi is defined is not realized by Pn but instead is realized by a different state type Ph, for example by “the inflation of certain cavities in the feet” (Martian pain). Since in both cases pain could be felt, a theory of mind should give both their place. In these cases it is advisable however not to take literally terms like “Feeling of pain”, otherwise a third aspect of pain would come into play, complicating the entire approach. In Lewis’ proposal – which is exclusively designed for functional and physical pain – such an aspect of pain is not adequately accounted for (I will come back to this problem in section III). “Feeling of pain” should rather be seen here as a facon de parler which indicates a conceptual problem: the ascription of pain only by means of functional and physical data in accordance with liberal criteria. These may not lead to unique results in all imaginable cases. The principal idea is to think of type-type identities like Mi = Pn as restricted to 1
Even if psycho-physical identities are restricted to individuals the issue is not merely to identify singular state tokens, but to see within an individual, respectively a time segment of an individual, whether neuronal states of a specific type occupy specific causal roles.
On David Lewis’ Philosophy of Mind
289
species, holding only for typical or paradigm cases. Strange cases like mad pain can’t be excluded apriori because of Lewis’ regularity view of laws of nature. Nevertheless one would first try to explain them in accordance with natural laws established so far, that is by searching for inhibitory factors which are not present in typical or paradigm cases. For Lewis, pain may even be ascribed to a Martian who is in the physical state Ph but quite untypically not in the functional state Mi (mad Martian pain). Such a claim is implausible for J. Owens (1986:171), for this state is in no way similar to human pain, neither functionally nor physically. From a “chauvinistic” human point of view this seems to be a sound objection. From a more neutral perspective, however, mad pain and mad Martian pain can be considered as functional anomalies, moreover, as functional anomalies in different species. Both cases are then equally eligable candidates for the term “pain”; accordingly Martian pain would not appear to be such a strange case. Due to symmetry reasons it therefore seems appropriate to leave room not only for functional but even for physical anomalies.1 b) Some questionable objections Several critics of Lewis’ theory of mind point to an incompatibility of having psycho-physical type-type identities conjoined with reductionism on the one side with allowing the possibility of MR on the other. But their arguments often rely on premisses Lewis is not willing to accept. Owens’ (1986) objections are based on Kripkean views: that natural kind terms, especially natural kind terms of neurophysiology, are rigid designators and on Kripke’s general objections against contingent identity statements. According to such views it is by discovering the physical structure of pain states that one acquires knowledge of their “intrinsical nature”; therefore functional properties of these states are seen to be merely accidental properties. For Owens this clearly shows that “pain” should be taken as a rigid designator. Consequently, the necessity of identity statements like “Pain = C-fibre firing” would be established, and a “radical physicalism” would have to be accepted, a position which definitely excludes the possibility of MR.2 This line of critique parallels Kripke’s well known argument for the necessity of identity statements with natural kind 1
For humans this would be a case where Mi is not realized by the neuronal state Pn and also not by the hydraulic state Ph – this is the difference to the case of Martain pain – but rather by a different neuronal state P*n.
2
Actually, Owens doesn’t accept the position of “radical physicalism”, a view hardly anyone adheres to. Instead, he takes sides with objections to identity theories brought up by Kripke, for whom the intrinsical character of pain can’t be revealed by neurophysical research but instead by immediate experiences of pain.
290
Erwin Rogler
terms like in “N(heat = molecular movement)”. But assertions concerning the rigidity or non-rigidity of designators can only be justified by appeal to intuitions, for example, intuitions regarding the use of natural language etc. Unanimous agreement can’t be guaranteed and is de facto not always achieved. Lewis (1983a: 369) considers heat to be simply the entity “which occupies the heat role”. In the actual world this role is occupied contingently by molecular movement, in other possible worlds this role could perhaps be occupied by caloric fluid. Lewis’ position seems to be by no means less plausible than Kripke’s. Even Kripke allows a certain amount of contingency in the heat example by mentioning the fact that we often refer rigidly to an entity via its contingent effects (here: feelings of heat). The following consideration is brought up by Lewis (1994:419) in favor of treating “pain” as a non-rigid designator: Suppose someone, at different times, is in the functional state of pain. Then in both cases pain is attributed to this person. But the pain role is realized by two different brain states. If “pain” would refer rigidly, only one of the two states is referred to and this state should then easily be distinguished from the other. But actually this is not the case. Only intricate experimental research can reveal a difference between both cases. This argument tacitly assumes as a premiss the possibility of MR within species (cf. I last paragraph) and it is designed for functional pain. A qualia freak would not be persuaded by it. Yet even he is not forced to take side with Kripke’s views. For suppose, contra Kripke and Lewis, “pain” is taken to be a rigid O-designator referring to felt pain and “C-fibre firing” to be a non-rigid T-term of a neurophysiological theory. Then, if certain empirical conditions are met, it seems entirely possible to interpret “pain = C-fibre firing” as a contingent statement of identity. A. Kernohan (1990) has other objections to MR in Lewis’ theory. According to Kernohan the possibility of MR can only be granted when ignoring the fact that the term “behavior” is ambiguous. On the one hand patterns of behavior, except for instinctive reactions, can be described as actions. This mode of description allows behavior terms to be applicable to a wide range of phenomena that is not restricted to certain species. Behavior of pain could thus be equally attributed to humans, apes, and even Martians etc. A necessary condition for allowing MR would be fulfilled. But action terms are intentional and because they appear as O-terms in functional definitions, those definitions seem to be circular to the extent that the same intentional content which should be eliminated by new functional specification of the definiendum still makes up part of the definiens. If however common sense intentional terms are banned from neurophysiology, then reduction of intentional psychology to neurophysiology seems to be impossible.
On David Lewis’ Philosophy of Mind
291
On the other hand this difficulty would not arise if one opts for descriptions of behavior as fine grained physical occurrences. But such a view clearly excludes MR and thereby leads to “chauvinism”. Martian pain can’t be accounted for on such a proposal and in general the folk psychological approach has to be given up. It is for similar reasons that Lewis (1994: 417) rejects views like this. As a third way out Kernohan considers, like Lewis (ibid.), the possibility of abstract physical descriptions of behavior, a view, which he does not characterize more. Such descriptions would not be chauvinistic and they could be used within psychology instead of intentional descriptions of actions, but presumably not within physical theories. Yet, equality of input and output descriptions is a necessary condition for psycho-physical identifications and for the reduction of psychological theories. I would like to point to another possible approach which might be able to avoid the mentioned incompatibilities. In the language of neurophysical theory T(P), L, just like in every language of empirical theories, observation data and experimental data must be describable. Even if Lewis doesn’t speak of the O-terms as “observation terms”, they in fact – Kernohan (op. cit.:240) also stresses this point – fulfill this role. Observed patterns of behavior clearly belong to the range of experiences which can confirm or falsify T(P). Assuming all intentional elements to be eliminated from the O-terms, it seems to be plausible to interpret them as making up partially the observation vocabulary of L. MR (at least to a certain extent) and psycho-physical identifications would be reconcilable then. (To take that route is not free of difficulties of its own. There are problems which arise for such a functionalist-materialist approach if it is assumed that these observation terms refer to secondary qualities. I will come back to this problem in section III.) It is precisely because the possibility of MR is taken seriously that Kernohan argues against the prospect of a reduction of psychology. Like other authors (e.g. Fodor 1974, Owens 1989, Seager 1991, Kim 1992 in contrast to his earlier work, e.g. Kim 1984) Kernohan is not willing to accept “disjunctive laws”. As a condition for the reduction of laws of one theory to the laws of another correspondence rules are required. These must be, if one opts for ontological reductions, of biconditional form. Yet if MR is allowed, then such biconditionals according to scheme (2) will contain disjunctive components with heterogenous predicates Mi* (P1 Z P2 Z,..., Z Pn). Conditionals of this form can’t be accepted as laws, according to Kernohan. Within a treatment of mental terms as non-rigid designators according to scheme (1) “disjunctive laws” can be avoided. The problem of reduction can then be addressed by answering two questions: (a) Is MR and its consequence, multiple reduction of terms, possible with respect to the laws of nature? Lewis
292
Erwin Rogler
(1994: 419) assumes this in certain cases. (b) Is a reduction of laws of psychology (or perhaps of a modified psychology) possible (or partially possible)? In the end, I think, both questions can only be answered empirically. Yet, many functionalists prefer instead a psychology in which M*-predicates are to be treated according to the scheme (2). Lewis therefore says (op.cit.: 421) that he doesn’t know whether he is a functionalist or not. If one also requires uniqueness of realization in relevant structures (A1,...,An) one could use conditionalized biconditionals as correspondence rules: (3)
A1 (Mi* P1), . . . , An (Mi* Pn)
On such a proposal local reductions of theories, e.g. of human psychology, may be possible (Kim 1989: 36ff). From (3) one can derive (4)
(A1 Y , ..., Y An) (Mi* P1 Z , ..., Z Pn)
So by replacing M*-terms with physical terms within a functional theory one will again arrive at “disjunctive laws”. What objections could be brought up against “disjunctive laws”? At present there is no generally accepted definition of the concept “law of nature”. What is usually done is to state more or less uncontroversial conditions which laws of nature have to fulfill. A fundamental objection against “disjunctive laws” says that they cannot satisfy confirmation and projection conditions (in the sense of Goodman). I will focus on a simple example discussed by Kim (1992: 11), but I use somewhat different predicates: the mental predicate “Mi*”, physical predicates “Pk” and “Pl”, and the behavior predicate “Oj”. Then we have (5) (6)
~x (Mi*x Ojx) psychological law ~x(Mi*x (Pkx Z Plx)) correspondence rule
From (5) and (6) we have (7) (8)
~x (Pkx Ojx) physical laws ~x (Plx Ojx)
A necessary condition that quantified conditionals of the form “~x(Fx Gx)” have to fulfill – in order to qualify as lawlike – is for Kim the principle: “observation of positive instances, Fs that are Gs, can strengthen our credence in the next F’s being G.” (Kim: ibid). This is a reinterpretation of Hempels special consequence condition: “If an observation report confirms the hypothesis H then it confirms every consequence of H.” (Hempel 1965: 31) as a “hallmark of lawlikeness” (Kim: ibid). If it turns out after discovering Pk and
On David Lewis’ Philosophy of Mind
293
Pl that all empirical evidence in favor of (5) have the form “Pkx Y Ojx”, then for (5) to fullfil Kims condition it is required that this evidence should also confirm (8) allowing projection into unconfirmed instances of the form “Plx Y Ojx”. Yet this appears to be highly implausible and for that reason we cannot consider (5) to be a law. Kim, as well as the above mentioned authors, does not generally deny the possibility of “disjunctive laws”, only if they contain heterogenous disjunctions, that is if the disjunctions are not natural kind terms. What we are looking for here is an independent criterion for natural kinds which itself does not rely on the concept of natural law, but neither Kim nor Fodor (1974) are able to present such a criterion. Kim sees considerations like these to support the conclusion that it is highly questionable to allow MR and still be faithful in the functionalists’ prospect for an universal psychology. The alternative would be that one would have to accept a multitude of local psychologies each reducible to physics respectively biology. A functionalist could reply that Kim focuses too narrowly on physical or biological similarities. P-predicates in disjunctions like (6) may indeed refer to physical or biological properties that are quite different. Nevertheless, prior experiences could support the assumption that those occupy identical causal roles, even if direct confirming instances for (8) have not yet occurred. It is functional similarities then, so to speak, that can take on the task of natural kinds. If these considerations are correct, the force of Kim’s objections are not that strong. Arguing thus presupposes the functional aspect to be of such an importance as to allowing a special type of abstract laws to be justified. The thesis that in formulations of natural laws only natural kind terms should appear seems to be heavily restrictive.1 Lewis (1983a:368) requires of the most fundamental laws of nature that their formulations should only contain terms for perfect natural kinds. Derived laws may contain less natural perhaps even gerrymandered terms. On the other hand even he excludes disjunctive events with heterogenous components (Lewis 1986: 266ff). This is in accordance with his view that mental terms of FP are to be treated as nonrigid designators. Its rationale appears to be the same in both cases. Rigid 1
D. Armstrong (1983: 147-55), for example, denies disjunctive universals but does not generally refuse to speak of “disjunctive laws”. Armstrong sees them as “oaken laws”, holding only if special conditions do not obtain. He interprets, for example, putative disjunctive laws having the form F (G Z H), if properties G and H turn out to be mutually exclusive, as a conjunction of two oaken laws: F G, if H is absent F H, if G is absent I see no principal objection to such a construal of “disjunctive laws” if MR is allowed.
294
Erwin Rogler
functional terms cannot, due to MR, refer to causally effective properties. Behavior of pain in humans is caused by neuronal processes not by “being in pain”, otherwise one would have to face overdetermination. This makes the physicalist-reductionist component of Lewis’ theory overt which with MR tends to a plurality of locally reducable psychologies and without MR to “radical physicalism”. Besides that there is the functionalist component in his folkpsychological approach by which he tries to hold on to the project of a universal psychology. His critics want to lay open the incompatibility of both components, whereas Lewis, in interpreting mental terms as non rigid designators, aims for a balance in both, so that, at least formally, “disjunctive laws” can be avoided. (Quite recently Lewis (1994: 417) has accepted MR even within worlds, species, etc. Mental terms can then no longer be defined in the style of (1); definitions in the style of (2) seem to be required. Yet even if one interprets M*-terms as being non-rigid, one needs correspondence rules with finite disjunctive predicates for local reductions. So far as they refer to physical properties which are sufficiently similar correspondence rules might not be affected by the critique of “disjunctive laws”).1
II Intentionality Ramsification of FP can be seen as an improvement of an older idea of J.J.C. Smart that mental terms should be conceived to be ontologically neutral (Smart 1971: 349ff). Accordingly, reconcilability with materialism is guaranteed and physicalist reductions seem to be at least conceptually possible. Moreover Smart (1972: 217) believes “the hard core of common-sense mentalistic idioms” to be “topic neutral”. He favors an eliminative strategy for all residual elements which are not reconcilable with materialism, especially such items as religious beliefs and sediments of spiritualistic metaphysics (immaterial soul etc.) are treated that way. Yet, will then not be much more eliminated by such a proposal? In the first place one may point to the fact that by treating mental terms as being topic neutral Smart (1959: 59ff, 1961) tries to give an answer to objections brought up by Black and Stevenson. Both argue that Smart’s identity theory is only able to remove the psycho-physical dualism of processes but not of properties. As to the removal of property dualism 1
In any case, further liberalizing MR should not trivialize the possibilities of realization in making functional theories satisfiable by arbitrary states. Certain restrictions are unavoidable. At least if the structures in local realizations are specified to a high degree functional state types should again not be identified with disjunctions of heterogeneous physical state types. Presumably such possibilities are excluded by laws of nature.
On David Lewis’ Philosophy of Mind
295
the ramsifying approach can be seen as a way because one is now in a position to abstract from all mental properties and relations. This also concerns intentional states. By replacing terms for propositional attitudes, e.g. terms for specific beliefs and desires, with predicate variables, the relation to content seems to be eliminated. B. Loar (1981: 50, 62) therefore speaks of functionalist theories of that sort as “black box theories”. Combing through the literature of the last 20 years, one is confronted with a multitude of meanings attached to the term “folk psychology”. According to the view of many philosophers (e.g. Churchland 1981: 70; Fodor 1987: ch.1, 2; Johnsen 1994: 226; Lyons 1991: 91, Stich 1996: ch. 1, 3) conscious states, intentional ones as well as qualitative ones, belong to its essential elements. The former are primarily individuated by their contents. According to this view ramsifying FP will therefore, in contrast to Smart, eliminate exactly its hard core. FP, when already ramsified, then can no longer be the object of a critique by S. Stich (1978) who questions the relevance of propositional attitudes for causal explanations. Putnam’s twin earth examples and similar thought experiments were designed to show that physical and functional duplicates can have beliefs with different truth conditions. These views are not only in conflict with the supervenience of the mental on the physical (if the physical bases are restricted to states and processes within organisms), but clearly with every materialist theory of mind. Stich therefore, and for other reasons, postulates a non-intentional science of cognition. Yet by bracketing its intentional dimension not all objections against FP lose their force.1 Its causal structure after ramsification remains largely intact and this is the focus of other eliminative arguments. According to P. Churchland (1981: 82) “the eliminative materialist is not committed to the idea that the correct account of cognition must be a naturalistic account. (...) What he does hold is that the correct account of cognition, whether functionalist or naturalistic, will bear about as much resemblance to FP as modern chemistry bears to four-spirit alchemy.” For Stich (1996: ch. 2) the causal architecture of FP differs radically from certain connectionistic networks. If these represent the most advanced branches in psychology, then FP cannot any longer be taken as reducible to a basal science; instead it must be given up. For Stich these new developments in psychology indicate a revolutionary change in paradigms. One must therefore carefully distinguish between bracketing the intentionality of FP as a result of its ramsification and those strong eliminative 1
An overwiev on the main objections against FP can be found in Stich (1996: 18ff).
296
Erwin Rogler
conclusions drawn from its alledged overall falsity by which the fictional and nonreferential character of its main concepts might be established (even if they are defined functionally). Because on those views even the Ramsey sentence of FP, RT(M), has to be questioned and interpreted as false. Lewis’ reductionism is primarily ontological and therefore in the explained sense eliminative. It differs on the one hand from Churchlands nonreductive eliminativism and on the other from a pure nomological reductionism which permits to derive the laws of the theory to be reduced from a basic theory together with lawful correspondence rules. This type of reductionism doesn’t imply an ontological reduction, even if the correspondence rules are biconditionals. It can only motivate ontological reduction by methodological considerations, e.g. the principle of parsimony, if certain conditions are fulfilled. A further point in this context might be emphazised. Unlike the intentional (and intensional) irreferentialism of Quine (1960: 218ff, 226ff) and others, Lewis’ approach in which the realm of intentionality is, due to the ramsification of FP, only bracketed, leaves it open wether propositional attitudes should be definitively abandoned; it comprises the first step of Lewis’ theory of mind. It is only when, in a second step that no other realizations than physical realizations are allowed that the Brentano-intentionality will be definitively eliminated. Yet even then some materialist variants thereof might still be possible. Functionalistically conceived ramsification of FP with its neutral concepts has not only an eliminative function according to Lewis. Quite the contrary, with its help propositional attitudes and meanings should be ascribed. This is the theoretical problem of radical interpretation which Lewis (1974) introduces in the following way: one is invited to imagine a mighty knower who has perfect knowledge of a person (Karl) as a physical and functional system. But he has no knowledge concerning Karl’s beliefs and desires and of the meanings of his assertions. These – and this seems roughly to be the task – should be determined given the physical and functional facts about Karl. For materialists all this should appear to be strongly irritating because as far as they are concerned the mighty knower already knows everything about Karl. From a strict materialist viewpoint there seems to be nothing left to be determined. An ascription of propositional attitudes can therefore only be motivated pragmatically. For general purposes the usual intentional action explanations seem to be all that is required. Neuronal explanations, on the other hand, tend often to be too complicated, yet this fact should not play an argumentative role for the mighty knower. Therefore, Lewis cannot ground his arguments alone on those pragmatical facts. Presumably Lewis tentatively opts for a moderate dualistic view in these considerations which must be in accordance
On David Lewis’ Philosophy of Mind
297
with the supervenience of the mental on the physical. What we are looking for is a scheme of interpretation for Karl “specifying the attitudes and meanings as a function of the momentary total physical state”, so Lewis (op.cit.: 119). This means that a new dimension of personality would be revealed from a dualistic viewpoint, but if the process of interpretation again leads to the identification of intentional and neuronal state types at the finish line, one will have come full circle. In what follows his last quote Lewis speaks of interpretations “by identifying certain attitudes with certain (partial) physical states”. Are these, because of their characterization as being “partial”, states that have physical as well as mental intentional components? Every concession to a dualist position here would not only be in conflict with Lewis’ materialism but would also lead to other problems within his theory. For example, the principle of explanatory adequacy of physics (Lewis 1966: 105) he endorses allows only three ontological options for non physical entities: (a) they are in no way related to the physical world; (b) they are epiphenomena; (c) they are overdetermining causes. However, on every option propositional attitudes cannot take over those causal roles FP ascribes to them. Ascriptions of intentional states and meanings have to fulfill according to Lewis (1974: 112ff, 1983a: 374ff) additional constraints so that the possibility of having different interpretations can be limited. For example, they have to comply to the basic principles of decision theory encapsulated within FP and to other norms of rationality. What Lewis wants to achieve in the end is uniqueness, but he seems to be satisfied, when reference fixing is at stake, with some moderate indeterminacy (Lewis 1984: 223). Interpreting a person in accordance with these constraints presumes that they determine the intentional activity of a person to a considerable degree. From a materialist viewpoint such determination can only be understood as determination by way of disposition’s realizers, that is, due to causally effective structures inside the organism. As a result won’t this strategy lead to an unlimited subjectivism, to quite similar difficulties as in psychologism? I will point only to one aspect of this question, which is eminent in Lewis’ theory: the role of natural properties. Perfect natural properties or elite classes are to be discovered by physics. They impose order on the objects by revealing objective similarities and differences among them thereby limiting the range of possible interpretations of theories. Common sense terms may also refer to natural properties, only to less perfect ones, but they limit the range of possible interpretations of persons in similar ways. It is by relying on his account of natural properties as the main determinants for concept formation, especially in physics, that Lewis wants to set himself apart from subjectivist and relativist positions. Such considerations don’t belong to physics, neither in its present nor in its future state, but to a
298
Erwin Rogler
realist metaphysics and epistemology. One can see in these speculations a materialist version of the old prestabilized harmony between thought and being. (How do logical principles fit into those speculations? If one conceives them in analogy to natural properties and their status in Lewis’ theory, then something not too different from Loar’s thesis (Loar 1981: ch.4) will follow. Roughly stated, this is the thesis that logical relations between propositions are mirrored by causal relations between functional or physical states.) According to Lewis (1983a: 373) as a starting point for radical interpretation only knowledge of the functional states of the person can be assumed. Then ascriptions of attitudes can be understood as (preliminary) characterizations of realizing brain states, under the assumption that these refer to contents. This is one of the main theses in Lewis (1994: 421): “A mind is an organ of REPRESENTATION”.1 He does not give a more detailed specification, be it in terms of a language of thought or in terms of nonlinguistic connectionist networks. Lewis again relies on FP which in that respect he conceives to be agnostic in contrast to its linguistic interpretation by Sellars, Stich and other authors. But then the explanatory force of FP is not that strong. Merely global states of a system will be accounted for, not the fine grained structure in its hierarchical levels of subcapacities in which information is computed. With the restitution of intentional realism in the second step the aforementioned objections by Stich (1978) will again be relevant. Lewis tries to counter them, like other philosophers, by distinguishing between wide and narrow content. Wide contents are external entities, sometimes thought to be limited to extensions (Fodor 1994: 7). Narrow contents on the other hand are “in the head”. Only these are considered to be important for psychological explanations and in agreement with the supervenience condition. But “Narrow content” is not a sharply defined concept. Within Lewis’ theory “narrow content” has several meaning components, at least the following ones: (a) Narrow content is understood holistically. Required then is a more detailed specification in order to counter objections (cf. Rey 1997: 240), and in particular to counter Fodor’s critique (Fodor 1987: 56) that a radical holistic complex of contents will undergo change in all of its elements, due to a modification at one place, especially due to new experiences. As a result there can be no repetitions, hardly any cases of similar beliefs in different persons and as a result application of laws and intentional explanations would be impossible. (b) Putnam’s twin earth cases support the assumption that the narrow content of common sense beliefs includes an indeterminate component which 1
“Mind” should be understood here as “physically reduced mind”, which is eminent already in the title of that paper.
On David Lewis’ Philosophy of Mind
299
can be more finely specified as not identical in different contexts by expert knowledge. If one understands the common sense term “water” in every day knowledge to be a cluster concept like Lewis does, then Oscar’s belief and the belief of his physical duplicate on twin earth that water falls from the clouds would have same narrow content. Only after the scientific discovery that earthly water is H2O and that twin-earthly water is a different substance XYZ, different truth conditions, i.e. different wide contents would be revealed and as a result thereof we would have it that “water” as part of the belief contents of non-experts refers non rigidly. These considerations support the view that narrow content should be conceived then as a function from contexts and sentences or predicates etc. of the lingua mentis to truth conditions or properties etc. (e.g. H2O (XYZ) falls from the clouds). (Fodor 1987: 47ff). Understood thus, truth values cannot be assigned to narrow contents, simply because a function contains free variables. Therefore Lewis rejects such a proposal; narrow content should have a semantic interpretation. (c) Lewis often considers narrow content to be knowledge de se, or to be interpretable that way, as in the example above: “Oscar self-ascribes having heard, under the name of “water”, of a liquid that falls from clouds” (Lewis 1994: 426). Now the wide content should be derivable from narrow de se ascriptions and those facts, the person is acquainted with. This means that a function is assumed that takes narrow de se contents and contexts of acquaintance as arguments and gives wide contents (propositions, properties and objects1) as its corresponding values. If Lewis considers narrow contents to be neither syntactical expressions of a lingua mentis nor to be functions or phenomenal data and if they can be judged to be true or false (op. cit.: 425) then it seems to be reasonable to think of them as sort of propositions or holistical complexes of propositions; not as classes of possible worlds but as entities which have the logical structure of sentences. But what does it mean then to say that they are in the head? A materialist should interpret this phrase literally, that is, he should aim at identifying narrow contents with neuronal states. To speak of their causal role would not be problematic for him. But is it reasonable to assume that propositions (or functions if narrow content is understood in the line of Fodor’s view) are brain states? Maybe a neurologist in the distant future will be able to tell what happens in my brain, when I think “Water falls from the clouds”. Yet this would be something rather different from the narrow content of my belief 1
According to Lewis a proposition is taken to be a set of possible worlds, that is, a set which comprises exactly those worlds as elements in which the proposition holds true; and a property is taken to be a set of possibilia having as its elements all instances in our world and also all its instances in other worlds.
300
Erwin Rogler
that water falls from the clouds. Out of mere curiosity one could make the assumption that in such a belief a certain brain state represents itself in an unclear and confused way. But nothing could be gained by that. One would have a brain state and its confused representation. The problem of their materialist identification would remain unsolved. It then seems to be unavoidable to make concessions to a dualist position. The internal character of narrow content is most likely to be interpreted as being caused by brain processes. Close to the materialist view is its interpretation as epiphenomenon, so that its causal role must be occupied by a corresponding brain process. There is widespread agreement that every acceptable concept of narrow content should fulfill at least two conditions: Narrow contents should supervene on physical states and wide contents should supervene relative to contexts on narrow contents. It would be best to have 1-1-correlations (relative to contexts), because individuation of beliefs by wide content would then correspond to psychological individuation by causal powers. Several authors have expressed doubts concerning the satisfaction of both conditions (e.g. Manfredi 1993). Be this as it may the concept of narrow content is also problematic in materialist theories of mind. It is noteworthy that naturalist orientated philosophers (e.g. J. Fodor (1994), F. Egan (1995)) have recently tried to dispense with the concept of narrow content. From a materialist viewpoint Lewis’ conception of wide content appears to be equally problematic. Wide contents, as far as propositions and properties are concerned, requires Lewis to refer to possible worlds. Yet many materialists try to do without propositions of every sort1 and express general doubts concerning modal realism. Loar (1981: 144ff) as an example, only accepts propositions provisionally. Further on in his investigations the role of propositions is taken over by sentences which fulfill certain conditions of adequacy. Lewis sees causal theories of reference to be justified in a certain sense, but first of all reference to wide contents, especially concerning theoretical terms, is dealt with in terms of the description theory of reference (Lewis 1983a: 371; 1984: 222ff). It seems to be that reference fixing is for Lewis the semantical relation of satisfaction based on interpretations relative to contexts of acquaintance, or an analogical relation when beliefs are the focus instead of sentences. “Belief” can be understood in two ways here, as a neuronal state of a person or as the narrow content of that state. In the latter sense the relation of reference fixing considered as relative to contexts of acquaintance is identical with the afore mentioned function from narrow into wide contents. Even if one takes a neuronal state of a person as a relatum here, instead of narrow content, 1
See for example Schiffer (1987: ch. 3).
On David Lewis’ Philosophy of Mind
301
referring to wide contents in terms of the description theory of reference, is then not a physical but a metaphysical relation.
III Qualia and Materialism By neutralizing mental terms, not only intentional characteristics, but also the qualitative contents of mental states are bracketed. Their elimination will be legitimized if one holds on to Smart’s materialist metaphysics, because there is no place in it for qualia of whatever possible sort, e.g. phenomenal colors. (Smart 1963: ch. 4; Smart 1989: ch. 5). Lewis’ semantical rules for ascriptions of pain are confined in similar ways to a functionalist-materialist framework. Yet in Lewis (1980: 130) he seems to point out unmistakenly that the phenomenal character of certain states belongs to them essentially. “Pain is a feeling. Surely that is uncontroversial. To have pain and to feel pain are one and the same.” However, if one concedes their qualitative characteristics as a third component of pain, then functional states of pain can be realized in two different ways, physically and phenomenally. Several questions will then arise. I will only point to some difficulties. Taking again typical human pain as the prime example, there are more irregular cases than would be possible if we had only two components of states to be accounted for. Now the relation between the phenomenal component of pain and its functional and physical components has to be taken into consideration. Accordingly, a more complex reformulation of the rules for ascribing pain appears to be required, and to do so a solution to the absent qualia problem must first be given. Can cases in which the pain role is physically realized, but not by felt pain, to be treated as a possibilty, in the sense of being in accordance with the actual laws, or in some stronger sense of possibility? Lewis doesn’t answer these eminent questions, because generally in materialist metaphysics these questions either make no real sense or they have to be first entirely transformed to do so. It is therefore advisable not to take his apparent confession regarding the phenomenal characteristics of mental states seriously. They simply don’t fit into materialism. Or does he merely speak in the “tendentious jargon” in the last quote, he attributes in his postscript to Lewis (1980) to the friend of qualia? Yet, there he opts for a clear materialist position in defending his approach against the knowledge argument developed by F.Jackson and T.Nagel. Its upshot is that there is a specific type of knowledge to be gained: phenomenal knowledge concerning qualia. Such “knowing what it’s like” (Nagel) is supposed to be neither physical nor functional information. Not to be achieved through
302
Erwin Rogler
lessons, it can only be aquired by means of qualitative experiences. “(...) if you taste Vegemite, then you will know what it is like.” (Lewis: 1980: 131). Contrary to those authors Lewis insists that “(...) phenomenal information and its special subject matter do not exist” (ib.). In his treatment of the qualia problem two points are the main focus. First, as far as possible, a convincing rejection of the knowledge argument should be spelled out. Second, Lewis must explain the erroneous, but nevertheless altogether persistent, belief in qualia and phenomenal knowledge. Lewis understands information primarily as elimination of possibilities. One argument against phenomenal information takes the following supervenience thesis as a premiss: “no difference without physical difference. That is: any two possibilities that are just alike physically are just alike simpliciter” (Lewis 1990: 507). Because, as he sees it, materialism is a contingent truth, the supervenience thesis is confined to the sphere of materialist worlds, containing the actual world and all worlds around it “that have nothing wholly alien to this world” (op.cit.: 508). The supervenience thesis is furthermore drastically limited in yet another way. The needed result can only be arrived at by assuming that from physical statements true in those worlds together with suitable definitions resp. meaning postulates of the phenomenal terms all true phenomenal statements holding in them can be logically inferred. Frequently the supervenience relation is defined for a pair of classes of properties, for instance the pair which contains the class of physical and the class of mental properties. Yet by making the above assumption phenomenal and physical properties turn out to be identical in the domain of materialist worlds. Supervenience of the former on the latter class of properties then comes down to supervenience of the class of physical properties on itself. Only in this special case will possibilities which do not differ physically then really be “just alike simpliciter”. Moreover, if a supervienience relation holds between classes of properties this fact alone does in no way exclude property dualism and therefore does not exclude genuine phenomenal information. Suppose a and b are physical duplicates. If we have n simple phenomenal properties then there are 2n logical possibilities which by phenomenal information can be reduced down to one. In this case the supervenience thesis then allows only to infer from the phenomenal properties of a those of b and vice versa.1 1
The notion of “strong supervenience” for two families of properties A and B can be defined as follows: “A strongly supervenes on B just in case, necessarily, for each x and each property F in A, if x has F, then there is a property G in B such that x has G, and necessarily if any y has G, it has F.” (Kim 1984: 165) I understand Lewis’ supervenience thesis in the above context as a version of strong supervenience, so that the second necessity operator is interpreted as
On David Lewis’ Philosophy of Mind
303
According to a second argument new qualitative experiences will not produce phenomenal knowledge because they can’t limit the range of conceivable possibilities. “I cannot present to myself in thought a range of alternative possibilities about what it might be like to taste Vegemite” (op.cit.: 512). One could object to this line of reasoning that for possibilities to be limited it is not necessarily required that they must in all cases be conceivable. This of course is heavily dependent on how the notion of knowledge is characterized beforehand. If one has a conception of “knowledge” like the one M. Schlick (1925: 18ff) once proposed, according to which knowledge can be characterized as recognition, then at least a second experience of the taste of Vegemite conjoined with a memory of the first experience of its taste is necessary to produce real knowledge of both. Even on such a conception of knowledge there seems to be no reason to deny the phenomenal character of its subject matter, because without actually having experienced the specific taste of Vegemite it would not have been acquired. One might concede to Lewis that this taste could be caused by something else, for example by the study of “lessons” concerning the physical composition of Vegemite or “lessons” concerning those neuronal processes which actually take place, when one tastes it. Yet that information would clearly be entirely different from having phenomenal knowledge concerning the taste of Vegemite. Both arguments therefore are not conclusive. The first one assumes a quite limited conception of the supervenience relation and for that reason it is question begging. The second one is not convincing. Even Lewis seems to acknowledge that the first argument can only establish the hypothesis of phenomenal knowledge to be irreconcilable with materialism. “Therefore I deny the Hypothesis. I cannot refute it outright” (op.cit.: 509). Indeed, arguing from materialist premisses leads to a startlingly simple refutation of phenomenal knowledge: because qualia do not exist, there can be no phenomenal knowledge about them. In the end this appears to be the decisive point for Lewis. On the other hand Lewis wants to understand materialism not as a dogmatic declaration of faith but as a contingent hypothesis. Then at least plausibility arguments must be given and that is what he tries to do with the above antiqualia arguments, but this requires for them to be conceptually analytical and the first as physicalist necessity (concerning the essential features of all materialistic worlds). There is then a remarkable difference between identifications of phenomenal and physical properties which are necessary analytically and contingent identity sentences containing functional-mental and physical predicates. The latter identifications reflect the ontological neutrality of functional theories and therefore allow functional psychologies to be true not only in materialistic but also in some nonmaterialistic worlds. – As to the several concepts of supervenience which are discussed lately, see also Kutschera (1992).
304
Erwin Rogler
independent from materialist premisses. Primarily for Lewis it is the success of physics which justifies materialism. Its connection to the qualia problem then comes with the principle of the causal closure of the physical world. In Lewis’ view the qualiafreak is confronted with the following problem: “To believe in the phenomenal aspect of the world, but to deny that it is epiphenomenal, is to bet against the truth of physics” (op.cit.: 513). Another argument for the epiphenomenal character of qualia doesn’t take the causal closure principle as a premise (ib.). Its discussion here, however, would take up too much space. I refer to its thorough investigation by D.Robinson (1993). Lewis does not want to establish the truth of epiphenomenalism about qualia with these arguments. What he argues for is this: if qualia are taken as real, then they can only be epiphenomena. On the other hand, because qualia are then in fact useless for the prediction and explanation of behavior, granting them the status of being real seems to be suspect. A qualia friend could reply to this line of argument by saying that the principle of the causal closure of the physical world is not as unproblematic as it seems, because within certain interpretations of the measuring process in quantum mechanics the observer is no longer included in the domain of the theory. Above all, he could make a substantial philosophical objection. Lewis takes it as established that qualia do not belong in the physical world. Assuming thus is congruent with classical forms of materialism and many dualistic theories. Yet there are other ontological conceptions which do not ban qualia from the physical world. I think of philosophers sometimes called the “neutral monists” (Russell, Schlick, Feigl, Maxwell, Lockwood, et. al.). The common core of their theories is a monistic ontology that takes qualia to be internal components of events, only to a small extent capable of human experience. The world of qualitative events is taken for them to be the same as the domain of physical theories, but from their point of view physical theories can only comprehend structural aspects of the world, in particular its causal structure. Generally neutral monists opt for a “nonmaterialist physicalism” (Maxwell 1978: 365). Their proposals are also not without internal difficulties, but such a position is nevertheless interesting because neutral monism seems to avoid those chronical problems interaction theories and materialism have with qualia, for example the problem of the so called “explanatory gap” (Levine 1983). Besides causal connections between themselves, qualia as elementary components of reality can’t be further explained. Although somewhat neglected in current discussions, neutral monism is an attractive option here for qualia freaks.1 1
A short overview on the main theses of neutral monism can be found in L.Stubenberg (1996).
On David Lewis’ Philosophy of Mind
305
Lewis’ (1980:131) principal objection to the knowledge argument relies on the ability hypothesis that acquiring and having “phenomenal information” really consists in acquiring and having certain abilities of conceiving, remembering and recognizing. Nagel’s “knowing what it’s like to be” is not a “knowing that” but a “knowing how” and since both often appear to be entangled, the illusion of phenomenal information can be explained. That illusion stems from mixing up “knowing how” and “knowing that”. Although transforming phenomenal information into dispositions does not justify a nonextistence claim concerning qualia, nevertheless they will – if their existence is still assumed – turn out to be entirely inconceivable, mystical entities completely useless for common sense and scientific concerns. One could think of the ability hypothesis as an argument in favor of materialism, yet when allowing a minimum of Cartesian intuitions it might rather be taken as a violent behaviorist reinterpretation of “knowing what it’s like”, with the addition that dispositional bases are to be understood as causally effective structures within organisms. Without doubt, the pure experience of raw feels cannot be taken as knowledge. Primarily it is a philosophical abstraction perhaps to be had only in very special situations, like in the moment of awakening or when confronted with entirely new experiences. In general though, it is not pure qualitative contents that are experienced, one usually also knows what one has experienced. Having a qualitative experience at least in common cases is a synthesis of pure phenomenal and propositional consciousness produced by activating dispositions, typically those dispositions to conceive, remember and to recognize to which Lewis refers. These must be seen as conditions for gaining phenomenal information. The ability hypothesis therefore does not respond to a confusion. Quite the contrary, it confuses conditions for gaining qualitative experiences with that which will typically be gained in case of their fulfillment. Accepting phenomenal knowledge does not by itself exclude the possibility of reducing a psychology with phenomenalistic terms to a physical science, if raw feels are not to be eliminated. The subject matter of phenomenal knowledge comprises all entities that have qualitative components. Such knowledge is therefore not limited to merely subjective states (I now have pain, I now experience red) but it also contains perceptual knowledge which objectifies qualities – especially phenomenal colors – as properties of objects (At place g there is at time t a red rose). Therefore phenomenal knowledge can even be expressed by sentences of a reistic observation language. In section I, I have pointed out that folk psychological descriptions of inputs and outputs could be conceived in the same way as observation sentences.
306
Erwin Rogler
Lewis also stresses the point that stimuli often have secondary qualities and by bringing them into functional systems qualitative descriptions have to be replaced by functional descriptions. “But the relevant feature of the stimulus will often be some secondary quality – for instance, a colour. We cannot replace the secondary quality with a specification of the stimulus in purely physical terms, on pain of going beyond what is known to folk psychology. But if we analyze the secondary quality in terms of distinctive mental states its presence its apt to invoke, we close a definitional circle. So we should take interdefinition further. Let folk psychology include folk psychophysics” (Lewis 1994: 416). In that way enlarged functional systems again stand in relations to their surroundings. Therefore the problem of input and output description arises anew. Empirical examinations of functional FP resp. functional folk psycho-physics, like results of empirical research generally, can’t be articulated without referring to qualitative components of facts. Of course this also is true for empirical tests of physical input and output descriptions. As a corrolary to the ability hypothesis we have the ontological elimination of “the phenomenal aspect of the world” (Lewis 1990: 512). Such an elimination is reconcilable with reductive color realism, which some materialists accept. Its upshot then is not phenomenal colors but specification and localization of physical colors, for instance as disjunctive properties of those microstructures colored surfaces show (Armstrong, Smart), or as properties of neuronal states (Hardin). In contrast to such views “the full-scale-error-theory” (K. Campbell 1993: 267) claims that colors simply don’t exist. Yet all these competing strategies and controversies are entirely innermaterialist. The elimination of qualia and of phenomenal knowledge manifests a rationalisticaprioristic strand in materialism. Although materialists point to the success of physics, their ontology and epistemology can’t conform with its status as an empirical science. Additionally the elimination of qualia conflicts with our intuitions so strongly that even D.M. Armstrong (1987: 12) expresses his doubts: “I confess that I myself am still not fully satisfied with the physicalist reduction. I accept the reduction; I advocate it. But the phenomenology of the affair continues to worry me.” By offering a more differentiated view of the matter Lewis (1995) tries to weaken his critique of realism about qualia, but without changing the main lines of his position. Therefore he can’t accept the “identification thesis” which assumes that we can identify qualia by way of immediate experience. What remains unchanged for Lewis is that the intrinsic nature of pain is a physical state, not a feeling. Only in a less strict sense of “knwowing what” is Lewis willing to concede knowledge of qualia. “(...) I don’t know the essences of the various qualia of my experiences, I do know what relations of acquaintance I
On David Lewis’ Philosophy of Mind
307
bear to these qualia” (op.cit.: 143/44). A short look at the grain problem (op.cit.:142) will show, how to understand this “phenomenal knowledge”. If qualia are experienced as simple, as lacking internal structure (as entities without “grain”), then they are in fact simple according to the identification thesis. In this sense H. Feigl (1967: 90) understands qualia to be “realities-in-themselves”, that is, in his words: “It is no good to claim that the phenomenal experience has a hidden (latent) structure, and that it is hence “illusory” ” (Feigl 1975: 33). Feigl clearly sees that there might be an objection to his nonmaterialist identity theory. In P. Meehl (1966: 167/8) this objection is stated with the help of the following example: assume that we are experiencing at a finite space location R a red patch and that we cannot, despite all our concentration, detect any change in one of its properties during a finite time interval t. Yet, according to our neurophysiological knowledge no corresponding unchanging brain state exists and by Leibniz’ principle therefore an identification is precluded. The grain objection seems prima facie not to affect Lewis’ materialist identity theory, because on his view it is physical science alone which can uncover, as their intrinsic nature, the hidden structure of qualia. However, in common experiences they often appear to be simple and relatively constant phenomena. Those then are – unlike scientific ones – inadequate modes of representation. They are intentional entities so to speak. Several problems emerge with that proposal. One could rightfully ask which physical states and processes will be experienced as undifferentiated qualia. Only certain brain structures, or structures in the external world or both? Assuming that there are plausible reasons in favor of one or the other option, even then the problem of the ontological status of the representing qualia remains unanswered. The answers discussed so far show that the grain problem indicates a dilemma for materialism too. One could, like G. Rey (1997: 305ff) does, treat qualia, because they are “merely” intentional entities, as fictions like the winged horse or like Mickey Mouse. That is indeed in good materialist spirit: reality should only be attributed to those aspects of our world physical theories, besides their observation sentences, might also willing to refer to, but this appears to be completely counterintuitive.1 If on the other hand qualia are instead under 1
In G. Rey’s proposal, qualia, although unreal, should have resp. fulfill all those properties and functions which are usually ascribed to them as real entities. “Intentionally characterized states could (...) enter into all manner of lawful relations with external stimuli. Psychophysical laws (...) could be taken to relate not real to mysterious phenomenal properties, but simply real to mentally represented amplitudes, indeed, amplitudes as they are represented in the output of sensory modules. (...) intentionally characterized sensory states lacking real qualia may be every bit as intense, awful, or wonderful as any such states possessing them!” (ib.) For short: qualia yes, but without paying the ontological price. Unfortunately it remains totally
308
Erwin Rogler
stood to be real, clearly an alien dualistic element will have found its way into materialism, an element which should rather be eliminated beforehand due to the ontologically neutral interpretation of mental terms.
IV Conclusion Many philosophers split the psycho-physical problem into three issues: The problem of intentionality, the problem of qualia, and the problem of selfhood or self-consciousness (e.g. Feigl 1967: 152). Within Lewis’ earlier work (Lewis 1966, 1972) such a subdivision is not eminent. There he speaks only generally of “experiences” or “mental states”. The reason for this can be located in his functional definitions of mental terms in which distinct types of mental states are merely characterized by their specific causal roles. Not until in his later and recent work does Lewis address questions pertaining to the three subproblems in several places. His theory thereby becomes more concrete and gains more plausibility. Such analyses might be placed within his overall theory below the functional level. That is, they might be conceived as philosophical hypotheses which will, as far as possible, be replaced by neuronal or perhaps microphysical models in the future. In Lewis’ analyses of intentionality and of qualia there remain difficulties concerning their integration within strict materialism. At least they suggest a more liberal version of materialism. Since, for example, materialists often rely on “experience” as if indicating some sort of truism – and if this in fact includes qualitative components – they will then have to justify their use of such a concept. Additionally, Lewis’ materialism has a strong metaphysic flavor. On the one hand he postulates a physicalist reductionism which on the other appears to be limited de facto. It contains important exceptions, for example, those exceptions that stem from his view of elite classes and their relevance for reference
unclear, how to understand his statement concerning the specific irreality of qualia. There seems to be no detectable connection with their ascribed representational function, whether adequate or not. Nobody, for example, takes a trail of an animal, whether clear or faint, as merely fictive. What does it mean then to say of qualia that they behave as if they were real? Certainly not that they differ from their real counterparts only in lacking the property of existence (or in possessing a lesser greatness of being). It is also misleading to draw parallels to common fictions. Whether there really are winged horses can be decided by certain criteria. Yet, according to those criteria, sensible qualities are alike in being part of the empirical world. To postulate their irreality is therefore a pure metaphysical thesis. On such a view the empirical confirmation basis for physics fades away and appears to be merely an ontological illusion, whereas it is this science which is taken to give us the paradigm for reliable knowledge about reality.
On David Lewis’ Philosophy of Mind
309
fixing. It is inconceivable how a physicalist reduction of these concepts might be possible. Important considerations regarding the problem of self-consciousness can be found in Lewis (1979). The central thought: self-knowledge is not propositional knowledge, rather it consists in a self-ascription of properties. This analysis is based on his modal realist views and not on that version of contingent materialism he endorses for the actual world. In the latter sense interpretations of the indexical “I” within statements of self-reassurances or statements regarding the self-ascription of properties would have to be limited to the range of physically accessible phenomena. To what extent this desideratum can be fulfilled is an open question.
References Armstrong, D.M. (1983), What is a Law of Nature ?, Cambridge ——— (1987), “Smart and the Secondary Qualities”, in: Pettit.P et al. (1987), Metaphysics and Morality, Essays in Honor of J.J.C.Smart, Oxford, 1-15 Block, N. (1978), “Troubles with Functionalism”,in: Block,N. (1980), 268-305 ——— (1980), (ed.), Readings in the Philosophy of Psy chology, Vol.1., Cambridge, Mass. ——— (1980a), “What is Functionalism?”, in: Block,N. (1980), 171-84 Borst,C.V. (1970), (ed.), The Mind/Brain Identity Theory, London Campbell,K. (1993), “David Armstrong and Realism about Colour”,in: Bacon, J. et al. (ed.) (1993) Ontology Causality and Mind, Essays in Honor of D.M.Armstrong, Cambridge, 249-68 Churchland,P. (1981), “Eliminative Materialism and the Propositional Attitudes”, in: The Journal of Philosophy 78, 67-90 Egan, F. (1995), “Computation and Content”, in: The Philosophical Review 104, 181-203 Feigl, H. (1967), The “Mental” and the “Physical”-Essay and a Postscript, Minneapolis ——— (1975), “Russell and Schlick”, in: Erkenntnis 9, 11-34 Fodor, J.A. (1974), “Special Sciences”, in: Synthese 28, 97-115 ——— (1987), Psychosemantics, Cambridge Mass. ——— (1994), The Elm and the Expert, Cambridge, Mass. Hempel, K.G. (1965), Scientific Explanation and other Essays in the Philosophy of Science, New York Johnsen, B.C. (1994), “Mental States as Mental”, in: Philosophia-Philosophical Quarterly of Israel 23, 223-45 Kernohan, A. (1990), “Lewis’ Functionalism and Reductive Materialism”, in: Philosophical Psychology 3, 235-46 Kim, J. (1984), “Concepts of Supervenience”, in: Philosophy and Phenomenological Research 44, 153-76 ——— (1989), “The Myth of Nonreductive Materialism”, in: Proceedings and Adresses of the American Phi losophical Association 63, 31-47 ——— (1992), “Multiple Realization and the Metaphysics of Reduction”, in: Philosophy and Phenomenological Research 52, 1-26 Kripke, S. (1972), Naming and Necessity, Cambridge, Mass.
310
Erwin Rogler
Kutschera, F. (1992), “Supervenience and Reduction”,in: Erkenntnis 36, 333-43 Levine, J. (1983), “Materialism and Qualia: The Explanatory Gap”, in: Pacific Philosophical Quarterly 64, 354-361 Lewis, D. (1966), “An Argument for the Identity Theory”,in: Lewis (1983), 99-107 ——— (1969), “Art, Mind and Religion”,in: The Journal of Philosophy 66, 22-7 ——— (1970), “How to Define Theoretical Terms”,in: Lewis (1983), 78-95 ——— (1972), “Psychophysical and Theoretical Identifications”, in: Block, N. (1980), 207-15 ——— (1974), “Radical Interpretations + Postscripts”,in: Lewis (1983), 108-21 ——— (1979), “Attitudes De Dicto and De Se + Postscript” in: Lewis (1983), 133-159 ——— (1980), “Mad Pain and Martian Pain + Postscript”, in: Lewis (1983), 122-32 ——— (1983), Philosophical Papers, Vol.I, New York ——— (1983a), “New Work for a Theory of Universals”,in: Australasian Journal of Philosophy 61, 343-373 ——— (1984), “Putnam’s Paradox”,in: Australasian Journal of Philosophy 62, 221-36 ——— (1986), “Events”,in: Lewis (1986), Philosophical Papers, Vol.II, New York ——— (1990), “What Experience Teaches”,in: Lycan, W.C. (ed.) (1990), Mind and Cognition, Cambridge Mass, 499-519 ——— (1994), “David Lewis: Reduction of Mind”,in: S. Guttenplan (ed.) (1994), A Companion to the Philosophy of Mind, Oxford, 412-31 ——— (1995), “Should a Materialist belief in Qualia?”, in: Australasian Journal of Philosophy 73, 140-4 Loar, B. (1981), Mind and Meaning, Cambridge Lyons, W. (1990/91), “Intentionality and Modern Philosophical Psychology I,II”, in: Philosophical Psychology 3, 247-69, 4, 83-102 Manfredi, P. A. (1993), “Two Routes to Narrow Content: Both Dead Ends”, in: Philosophical Psychology 6, 9-21 Maxwell, G.(1978), “Rigid Designators and Mind-Brain Identity”, in: Minnesota Studies in the Philosophy of Science IX, 365-403 Meehl, P. A. (1966), “The Compleat Autocereobroscopist: A Thought-Experiment on Professor Feigl’s Mind-Body Identity Thesis”, in: Feyerabend, P.K./Maxwell, G. (ed.). (1966), Mind, Matter and Method, Essays in Philosophy and Science in Honor of Herbert Feigl, Minneapolis 103-80 Owens, J. (1986), “The Failure of Lewis’ Functionalism”,in: The Philosophical Quarterly 36, 15773 ——— (1989), “Disjunctive Laws?”,in: Analysis 49, 197-202 Quine, W.V.O (1960), Word and Object, Cambridge Mass. Rey, G. (1997), Contemporary Philosophy of Mind, Cambridge Mass. Robinson, D. (1993), “Epiphenomenalism, Laws & Properties”, in: Philosophical Studies 69, 1-34 Schiffer, S. (1987), Remnants of Meaning, Cambridge, Mass Schlick, M. (1925), Allgemeine Erkenntnislehre, Frankfurt 1979, (Reprint of Berlin2, 1925) Seager, W. (1991), “Disjunctive Laws and Supervenience”,in: Analysis 51, 93-8 Smart, J.J.C. (1959), “Sensations and Brain Processes”,in: Borst (1970), 52-66 ——— (1961), “Further Remarks on Sensations and Brain Processes”, in: Borst, C.V. (1970), 93/4 ——— (1963), Philosophy and Scientific Realism, London ——— (1971), “Reports of Immediate Experiences”, in: Synthese 22, 346-59 ——— (1972), “Further Thoughts on the Identity Theory” in: Smart, J.J.C. (1987),
On David Lewis’ Philosophy of Mind
311
Essays Metaphysical and Moral, Selected Philosophical Papers, London, 215-28 ——— (1989), Our Place in the Universe, Oxford Stich, S.P. (1978), “Autonomous Psychology and the Belief-Desire Thesis”, in: The Monist 61, 573-91 ——— (1996), Deconstructing the Mind, New York Stubenberg, L. (1996), “The Place of Qualia in the World of Science”, in: Hameroff, S.T. et al. (ed.) (1996), Toward a Science of Conciousness, The First Tuscon Discussions and Debates, Cambridge, Mass., 41-49
312
Barbara Von Eckardt, Jeffrey S. Poland
BARBARA VON ECKARDT, JEFFREY S. POLAND In Defense of the Standard View Abstract In Explaining Attitudes, Lynne Rudder Baker considers two views of what it is to have a propositional attitude, the Standard View and Pragmatic Realism, and attempts to argue for Pragmatic Realism. The Standard View is, roughly, the view that “the attitudes, if there are any, are (or are constituted by, or are realized in) particular brain states” (p. 5). In contrast, Pragmatic Realism that a person has a propositional attitude if and only if there are certain counterfactuals true of that person. Baker’s case against the Standard View is a complex one. One aspect of that case is that there are no good general arguments for the Standard View. We argue that one of the three she considers, an argument from causal explanation, can be defended. The problematic premise, according to Baker, is the claim that unless beliefs are [identical to, constituted by, or realized by] brain states, they cannot causally explain behavior. Baker tackles this premise in two ways: in Chapter 4, she attempts to undermine the brain-explain thesis by taking issue with the conception of causal explanation that she believes supports it; in Chapter 5, she argues against the brain-explain thesis directly, by attempting to show that it is false, whether or not the Standard View is true. Neither of these attempts are successful, on our view. The alternative version of the causal argument we develop uses the doctrine of physicalism as a supplementary premise rather than appeal to a reductive conception of causal explanation. After presenting this alternative version, we consider how Baker might respond to it by drawing on her discussion of materialism. We conclude that Baker’s argument against materialism does not generalize to physicalism as we construe it.
Introduction In Explaining Attitudes, Lynne Rudder Baker considers two views of what it is to have a propositional attitude, the Standard View and Pragmatic Realism, and attempts to build a case for Pragmatic Realism.1 The Standard View is, roughly, the view that “the attitudes, if there are any, are (or are constituted
1
Lynne Rudder Baker, Explaining Attitudes: A Practical Approach to the Mind (Cambridge: Cambridge University Press, 1995). All chapter and page references will be to this book.
In Defense of the Standard View
313
by, or are realized in) particular brain states” (p. 5). In contrast, Pragmatic Realism denies that it is an essential feature of the attitudes that they bear some relation to brain states and holds, instead, that a person S has a propositional attitude if and only if “there are certain counterfactuals true of S, where content of the counterfactuals may be intentionally characterized” (p. 21). Baker’s case for Pragmatic Realism has two parts. The first part consists of a set of arguments designed to show that the Standard View is beset by troubles, both internal and external. The second part is a positive defense of Pragmatic Realism. The aim of this paper is to take on one component of Baker’s case against the Standard View and argue that it is not persuasive.
The Standard View It is not entirely clear from Baker’s book just what she takes the Standard View to be. Since what the precise contours of the Standard View are makes a difference for our purposes, we will have to engage in a certain amount of regimentation. There are two areas of unclarity. The core of the Standard View is clearly a view about the relation between propositional attitudes and brain states. In fact, Baker is quite explicit that it is fundamentally a view about the relation between propositional attitude and brain state tokens. What is not so clear, however, is precisely which relations are encompassed by the Standard View. All of her formulations include the relations of identity and constitution. In addition, she occasionally also mentions realization although it is not obvious from the text whether she regards realization to be distinct from identity or constitution. We will take realization to be a distinct relation. Hence, on our reconstruction of Baker, the Standard View will encompass identity, constitution, and realization. The second area of unclarity is whether the Standard View includes any other claims besides a claim about the relation between propositional attitudes and brain states. At various points in the book, Baker speaks about assumptions on which the Standard View “rests” or that certain claims are true “according to the Standard View.”1 But does she take these associated assumptions to actually be part of the Standard View or are they, on her view, simply claims typically conjoined with the Standard View? The text does not provide 1
An example of the first would be the claim that common sense psychology is “a theory ripe for replacement by scientific psychology” (p. 69); an example of the second would be that, “according to the Standard View, scientific or metaphysical theory always trumps practice” (p. 19).
314
Barbara Von Eckardt, Jeffrey S. Poland
the answer. We will interpret the Standard View narrowly to include, as constitutive of the view, only the claim about the relation between propositional attitudes and brain states. The various other assumptions and claims that Baker discusses in connection with the Standard View will be termed “companion” assumptions and treated as belonging to one context in which the Standard View (perhaps, frequently or typically) occurs. Finally, like Baker, we will formulate the view in terms of beliefs although the scope of the view is intended to be the propositional attitudes in general. As a result of these clarifications, we get the following formulation:1 The Standard View: For every belief token b, there exists a brain state token n such that b is either identical to, constituted by, or realized by n.
The Dialectic in Outline Baker’s overall case against the Standard View is a complex one. Explaining Attitudes, in fact, contains four groups of arguments against the Standard View. Even though we will only concern ourselves, in this paper, with one of those four groups, it will be useful to begin by outlining the argumentative strategy of each group and indicating, briefly, how we would respond. Such an outline should provide the reader with a glimpse of the forest before we begin taking a closer look at some of the trees. Argument Group I. Baker’s basic thrust in Chapter 1 seems to be to argue that there are no good general arguments for the Standard View. She thinks that the Standard View is “so entrenched that the literature contains few (if any) explicit arguments for it” (p. 12). Nevertheless, since she takes it that there are powerful “motivations” for the Standard View, she sets out three arguments for it which, presumably, capture these motivations. She calls these “an argument from metaphysics,” “an argument from science,” and “an argument from causal explanation.” We agree with Baker that the first two arguments do not support the Standard View. Our response to the third argument is more complicated. The third argument, which is originally presented in a fairly stripped down form can be expanded in several ways. We will argue below that there is an expanded version which is defensible. However, as we shall see, the version we defend is not the version Baker argues against. Argument Group II. Baker’s second group of arguments are contained in Chapter 2 of Explaining Attitudes. The thrust here is that all extant versions of the Standard View are problematic. In particular, Baker’s claim is that none of 1
See Hill (1997), p. 133 and Sanford (1996), p. 182 for other construals of the Standard View.
In Defense of the Standard View
315
these versions have been able to solve the problem of showing “how content may be assigned to internal physical states in such a way that beliefs can be causally explanatory” (p. 32). Baker considers three specific versions of the Standard View, one due to Lycan, one due to Fodor, and one due to Dretske and argues that none are satisfactory. Although we shall not present or argue for it in this paper, our response is to propose a fourth version which, arguably, can solve the content/causation problem without falling prey to any of the difficulties Baker raises for the other three versions. Argument Group III. According to Baker, the Standard View “rests on” or is motivated by a number of companion assumptions. Chapters 3 and 4 are designed to make these assumptions explicit and show that they are untenable. We agree that many of these companion assumptions are untenable (for example, the thesis that “reality, insofar as it is knowable, is knowable exclusively by means of science”). But we also want to suggest that, for the most part, they are not, in fact, logically connected with our conception of the Standard View. Hence, even if they are problematic, this fact in no way undermines commitment to our version of the Standard View. The one exception is the claim that unless the Standard View is correct, beliefs could not causally explain behavior. And, in fact, this claim is one we attempt to defend below against Baker’s objections. Argument Group IV. Chapter 6 contains two additional miscellaneous arguments against the Standard View. One is that the Standard View has no explanatory or epistemic import. The other is that it involves a category mistake. Once a viable version of the Standard View and a viable argument for the Standard View are in place, these final arguments can fairly easily be countered.1
An Argument for the Standard View Of the three general arguments for the Standard View that Baker presents in Chapter 1, only one is worthy of defense, the so-called “argument from causal explanation.” It goes like this: An Argument from Causal Explanation (Original Version) CE1. Unless beliefs are [identical to, constituted by, or realized by] brain states, they could not causally explain behavior. 1
Hill (1997) provides a somewhat different reading of this chapter. He locates what he takes to be Baker’s “primary argument for the view that belief-tokens are not constituted by brain states” here and argues that it is unsatisfactory.
316
Barbara Von Eckardt, Jeffrey S. Poland
CE2. Beliefs can causally explain behavior. CE3. Therefore, beliefs are [identical to, constituted by, or realized by] brain states.1 The problematic premise, according to Baker, is premise CE1 which she dubs the “brain-explain thesis.” Baker tackles this premise in two ways. First, in Chapter 4, she attempts to undermine the brain-explain thesis by taking issue with the conception of causal explanation that she believes supports it. Second, in Chapter 5, she argues against the brain-explain thesis directly, by attempting to show that it is false, whether or not the Standard View is true. Neither of these attempts are successful, on our view. We shall consider each of these arguments in turn, looking first at the Chapter 4 argument. According to Baker, the brain-explain thesis and, with it, the motivation for the Standard View “rest on a hopelessly anemic conception of causal explanation” (p. 93). This conception is one that “places a reductive, ontological requirement on properties that can be explanatory” (p. 118). It is not a fullblown account of causal explanation. Rather, the reductive conception is a second-order thesis concerning the sorts of conditions that need to be satisfied for a causal explanation to be adequate. Any specific account of causal explanation will include claims of the form: A causal explanation that F in circumstance C causes G in C is adequate only if . . . .. The reductive conception of causal explanation holds that the blank is to be filled in with necessary conditions which are, in some sense, reductive. Baker considers four proposed standards of causal-explanatory adequacy which, on her view, fall under this reductive conception. These include:
“& only if instances of the putatively explanatory properties enter into an objective relation with the instance of the property to be explained, where an objective relation is one that can be instantiated in a world without intentional states (unless the relata are intentional psychological states)” (p. 95). “& only if putatively explanatory properties supervene on the current intrinsic physical properties of their bearer” (p. 96). “& only if possession of the putatively explanatory properties affects the physical constitution of their bearers” (p. 96).
1
Baker expresses the first premise and conclusion simply in terms of beliefs being brain states. However, she makes it clear that this is intended to be short-hand for their being identical to, constituted by, or realized by brain states. See p. 12.
In Defense of the Standard View
317
“& only if the putatively explanatory properties and the explanandum property are coexemplified with properties projected by a strict law” (p. 98).1
Baker’s argumentative strategy, then, is to question the merits of the brainexplain thesis (premise CE1 of the Argument from Causal Explanation) by raising questions about a premise of the sub-argument she takes to support premise CE1. We can make this explicit by expanding the original version of the argument like this: An Argument from Causal Explanation (Expanded Version 1) A causal explanation is adequate only if some reductive ontoCE1a. logical condition ROC is satisfied. In the case of belief explanations, ROC is satisfied only if CE1b. beliefs are [identical with, constituted by, or realized by] brain states. Thus, unless beliefs are [identical to, constituted by, or realCE1. ized by] brain states, they could not causally explain behavior. Beliefs can causally explain behavior. CE2. Therefore, beliefs are [identical to, constituted by, or realized CE3. by] brain states. The real culprit, then, according to Baker, is premise CE1a and, hence, it is against premise CE1a that her counter-argument is directed. The counterargument is, basically, this. There is a huge class of nonpsychological ordinary explanations that we take to be adequate. For example, consider the following: (1) The record drop in the New York Stock Exchange on October 19, 1987 (“black Tuesday”), is causally explained by program trading. The speed of program trading overwhelmed the mechanisms that were supposed to prevent a crash. By the time that trading was topped on a particular stock, it was too late; program trading allows huge blocks of stock to be bought and sold in a few seconds. (2) Al’s application for a gun permit was turned down. The causal explanation is that Al is a convicted felon. If he had not been a convicted felon, he would have received the gun permit. (3) Two people with the same incomes and assets in 1994 as in 1993, and with no change in the tax laws, owed more federal taxes in 1994. Why? They got married in 1994, and in the United States, for purposes of assessing one’s income tax, there is something called the ‘marriage penalty’.
1
Baker associates the four proposed standards of adequacy with writings of Jaegwon Kim, Stephen Stich, Jerry Fodor, and Donald Davidson, respectively.
318
Barbara Von Eckardt, Jeffrey S. Poland (4) In 1993, Janet Reno began to be accompanied by Secret Service agents. Why? She had been nominated to be attorney general, and her nomination triggered a sequence of events that led to her having Secret Service protection. (p. 100)
According to Baker, it is reasonable to hold that “any acceptable standard of explanatory adequacy should admit explanations from this class as adequate” (p. 93). However, all four of the proposals that impose a reductive ontological condition (ROC) fail this test miserably. In particular, they fail to count as explanatory a sample of ordinary non-psychological explanations that clearly seem to Baker to be causal explanations. Hence, according to Baker, there are good inductive grounds for supposing that premise CE1a of Expanded Version I of the Argument from Causal Explanation is false. Simply ruling out the reductive conception of causal explanation without proposing an alternative is, of course, not very satisfactory. For Baker herself wants to endorse premise CE2 which claims that beliefs can causally explain behavior. So in place of the reductive conception she offers a “test” for “a common and important kind of causal explanation” (p. 121). The test is this: Control Test Let C be circumstances in which someone can produce or prevent an occurrence of a certain type of phenomenon, G, by producing or preventing an occurrence of another type of phenomenon, F, where an occurrence of F is temporally distinct from an occurrence of G and an occurrence of F does not itself entail an occurrence of G. Then: (CT) An occurrence of F in C causally explains an occurrence of G in C if: (i) If an F had not occurred in C, then a G would not have occurred in C; and (ii) given that an F did occur in C, an occurrence of G was inevitable. (p. 122) Let us call this the “pragmatic” conception of causal explanation by way of contrast with the “reductive” conception. The strength of Baker’s case against premise CE1a clearly rests on whether her sample ordinary non-psychological explanations show the inadequacy of the reductive conception of causal explanation. Although we do not have the space to consider this question in detail here, there is good reason to think that they do not. In using the fact that, in ordinary life, we would readily consider explanations like (1) - (4) to be causal explanations against premise CE1a, Baker seems to be introducing a subtle equivocation on the conditions of adequacy imposed by the reductive conception. In particular, as Antony and Levine (1997) note, she is conflating “the metaphysical requirements for a causal explanation to be true, with the epistemological requirements for a
In Defense of the Standard View
319
causal explanation to be acceptable.”1 On the first sense, the causal conception would read: CE1aM (‘M’ for ‘metaphysical’). A causal explanation is true only if some reductive ontological condition ROC is satisfied. On the second, it would read: CE1aE
(‘E’ for ‘epistemological’). A causal explanation is acceptable only if we are aware or are justified in believing that some reductive ontological condition ROC is satisfied.
Baker’s array of ordinary nonpsychological explanations count as counterexamples to premise CE1a only on the epistemological reading (CE1aE) since we are not aware, in any of the cases, that a reductive ontological condition is satisfied. However, they fail utterly to undermine the metaphysical version of CE1a (CE1aM) and that is the reading that is relevant to the brain-explain thesis. In Chapter 5, Baker attempts to develop a more direct attack against the brain-explain thesis, arguing that it is false, whether the Standard View is true or false. Clearly, the first disjunct is the critical one. Here is Baker’s argument – a conditional proof– that if the Standard View is true, then the brain explain thesis is false (p. 138): Baker’s Conditional Proof CP1. Suppose beliefs have brain states as constituents. (Conditional premise.) CP2. No intentional explanation is replaceable by physical explanations of the constituents of the intentional phenomena. CP3. Belief explanations are intentional explanations. CP4. No belief explanation is replaceable by brain-state explanations (physical explanations of brain states). (From CP1, CP2, CP3) CP5. If the “brain explain” thesis is true, then belief explanations are replaceable by brain-state explanations of the same phenomena. CP6. The “brain explain” thesis is false. (From CP4 and CP5) CP7. Thus, if beliefs have brain states as constituents, the brain explain thesis is false. (Discharging the conditional assumption) Note that as an argument for the conditional conclusion CP7, Baker’s proof has several odd logical features. To begin with, the only role that the Standard View (CP1) plays is to entail CP4, in conjunction with CP2 and CP3. If the 1
See Antony and Levine (1997), pp. 95-96.
320
Barbara Von Eckardt, Jeffrey S. Poland
argument is simplified so as to include only the last four lines (treating CP4 as a premise), then that the brain explain thesis is false (CP6) follows directly from CP4 and CP5, without any reference to the Standard View. One could still, of course, view it as an argument for the conditional conclusion of the following form: Premises (with no mention of the Standard View.) Thus, if the Standard View, then the brain explain thesis is false. but this is only because the premises themselves entail that the brain explain thesis is false. An argument that looked like this: Premises (with no mention of the antecedent of the conclusion) Thus, if the moon is made of green cheese, then the brain explain thesis is false. would be equally valid since if P entails Q then P entails (S Q). To make matters worse, one can show that there is no sound argument to the falsity of the brain explain thesis, in which the Standard View plays a role (that is, is taken as a premise), for the simple reason that the Standard View and the denial of the brain explain thesis are inconsistent.1 For all of these reasons, the best reconstruction of Baker at this point seems to be as a straight-forward (non-conditional) argument to the falsity of the brain explain thesis, viz. Baker’s Direct Argument Against the Brain-Explain Thesis DA1. No belief explanation is replaceable by a brain-state explanation. DA2. If the brain explain thesis is true, then belief explanations are replaceable by brain state explanations of the same phenomena. DA3. Thus, the brain explain thesis is false.
1
Take SV = the Standard View (“beliefs are [identical to, constituted by, or realized by] brain states”), C = Brain states can causally explain behavior, and BE = the brain explain thesis (“Unless beliefs are [identical to, constituted by, or realized by] brain states, they could not causally explain behavior.”). Note, however, that BE = (SV v -C) and, hence, the denial of BE = - (SV v -C) or (-SV & C). Thus, when SV is true, (-SV & C) will be false. To show that there is no sound argument to the denial of the brain explain thesis in which SV is a premise, we need to consider three cases: (1) where - BE follows from SV alone, (2) where it follows from the conjunction of SV and some other set of premises X, and (3) where it follows from X alone. We have already shown (1) is not possible. When SV is true, - BE is false. In case (2), SV can either be true or false. When it is false the conjunction (SV &X) will be false; hence, the argument will not be sound. When it is true, no matter what X’s contribution, - BE will be false; hence, the argument will not be valid, and, thus, not sound. In case (3), if X alone entails - BE, then since SV and - BE are inconsistent, X and SV will also be inconsistent. Hence, the conjunction of X and SV can never be true. Hence, no argument with SV and X as premises will be sound.
In Defense of the Standard View
321
Let us turn now to an evaluation of this argument. We readily accept premise DA1 if ‘replaceability’ is taken to mean the same as ‘provides the same explanation as’. Because explanation is crucially connected to understanding, it is intentional through and through. Thus even if two explanations describe the same phenomenon, if they are conceptually distinct, they will be providing different explanations. So premise DA1 is perfectly consistent with the Standard View. Premise DA2 is another matter. Why would Baker hold that the replaceability of belief explanations by brain state explanations follows from the brain explain thesis? Perhaps, again, it is because she is confusing the metaphysical conditions for a causal explanation’s being true with the epistemological conditions of a causal explanation’s being acceptable. Suppose one held that the brain-explain thesis was grounded in CE1aE, the view that a causal explanation is acceptable only if we are aware or are justified in believing that some reductive ontological condition ROC is satisfied, and, hence, that a belief explanation is acceptable only if we are aware or are justified in believing that the beliefs in question are identical to, constituted by, or realized by certain brain states. Suppose, further, that one had in hand an acceptable belief explanation of the form: belief b caused behavior m. Then, it might be plausible to think that the belief explanation could be replaced by an explanation citing the brain states in question – although even in this case, one could argue that the two explanations would not be identical explanations and, hence, that replaceability would fail. In any case, the point is that the brain explain thesis is grounded only in the metaphysical version of CE1a (that is, CE1aM) and, hence, is entirely compatible with a non-replaceability claim. Even if beliefs are identical to brain states and even if the truth of some reductive ontological condition appealing to brain states is necessary for the truth of a belief explanation, such explanations will still be different sorts of explanations from brain state explanations. Another way to put the point is this. Explanations can be factored into two components: what makes them true and what makes them explanatory, that is, what makes them count as even possible explanations. What makes an explanation explanatory very much depends on the sorts of concepts it employs; what makes it true is that its truth conditions are satisfied in the world. Insofar as the concept of a belief and the concept of a brain state differ, the explanatoriness of belief explanations and brain state explanations will differ even if they have the same truth conditions (by virtue of the fact that beliefs are brain states). But if, then, a belief explanation explains its explanandum differently than a brain state explanation does, the latter will certainly not be able to replace the former, even if the brain-explain thesis is true.
322
Barbara Von Eckardt, Jeffrey S. Poland
We have shown that Baker’s attempts to undermine the brain explain thesis and, as a result, the Argument from Causal Explanation, fail. However, as with all negative arguments, our negative argument against Baker doesn’t show that the Argument from Causal Explanation does provide a worthy defense of the Standard View. We would now like to argue the latter point, in particular, that there is a defensible alternative expanded argument from causal explanation to the conclusion that beliefs are identical to, constituted by, or realized by brain states.
An Alternative Argument for the Standard View The alternative expanded argument uses the doctrine of physicalism as a supplementary premise rather than appeal to the reductive conception of causal explanation. Take physicalism to be the view that all things, properties, events, and processes in the natural world are ultimately realized by physical things, properties, events, and processes.1 Suppose now that in addition to the doctrine of physicalism we also accept (CT), Baker’s Control Test. Consider the following putative explanation: S: Susan’s belief that logic will help her on the law boards causally explains her taking logic. One can easily argue that S is a genuine causal explanation if one accepts (CT). For it can plausibly be argued that, in Susan’s circumstances, if she had not believed that logic would help her on the law boards, she would not have taken logic; and that given that she did so believe, in the particular circumstances she was in, her taking logic was inevitable. So far so good. But now a further question suggests itself: Why are these two counterfactuals true? In particular, why is it that if she had not had the belief in the circumstances, she would not have taken logic? And why is it that, in the circumstances, her having the belief made her taking logic inevitable? These two conditionals are not true by magic. Something about the world makes them true. And if physicalism is true, that something must be some underlying process ultimately grounded in basic physical processes. But here is the key point. Since realization is a relation that proceeds from the physical to the psychological via numerous intermediate way stations, for current purposes, the story we tell about that underlying process will, at the biological level and for normal people, refer to brain states! So premise CE1 (the brain-explain thesis) of the 1
For more on the sort of physicalism we endorse, see Jeffrey Poland’s Physicalism: The Philosophical Foundations (Oxford: Oxford University Press, 1994).
In Defense of the Standard View
323
Argument from Causal Explanation is true after all. But it is true not because an occurrence can’t be causal explanatory unless it satisfies some reductive ontological requirement, but rather because physicalism demands a basis for the causal explanatoriness of an occurrence – even when this causal explanatoriness simply amounts to the occurrence satisfying certain conditionals – namely, some mechanism or process at a lower level in virtue of which the conditionals are true. In sum, for the purposes of this paper, instead of Expanded Version I of the Argument from Causal Explanation, we wish to endorse Expanded Version II, viz. An Argument from Causal Explanation (Expanded Version II) CT CE(II)1a. Suppose that there exists a class of event pairs (occurCE(II)1b. rence of F, occurrence of G) such that conditions (i) and (ii) of the Control Test are satisfied and, hence, the occurrence of F is causal explanatory of the occurrence of G, and it is the case that the occurrence of F is causal explanatory of G only because (occurrence of F, occurrence of G) satisfies the Control Test. All things, properties, events, and processes in the natural CE(II)1c. world are realized by things, properties, events, and processes ultimately grounded in the physical. (The Doctrine of Physicalism) If CE(II)1c then for each event pair in the conditionals CE(II)1d. (i) and (ii) of (CT) will be satisfied only if there exists an underlying mechanism ultimately grounded in physical processes in virtue of which (i) and (ii) are true. The psychological is realized by the physical only via the CE(II)1e. biological and, in particular, via brain states. Thus, for at least those F occurrences in that are belief CE(II)1. occurrences, unless those beliefs are [identical to, constituted by, or realized by] brain states, they will not causally explain behavior. Some beliefs do causally explain behavior. CE(II)2. Therefore, some beliefs are [identical to, constituted by, CE(II)3. or realized by] brain states. Because Baker does not take the Control Test to be necessary, in general, for a pair of events to be causally explanatorily related, the above argument has to be restricted to those event pairs for which the Control Test does play such a
324
Barbara Von Eckardt, Jeffrey S. Poland
necessary role. But it is reasonable to assume that many, if not all, of the event pairs that Baker is interested in will fall in this class. Thus, we take it that her position is undermined if the above argument is sound. To establish the brainexplain thesis and, hence, the Standard View in full generality would require showing that the sort of reasoning employed here could be extended to all causal explanations, whatever “test” they satisfy.
A Line of Response to the Argument Since Baker takes it that advocates of the Standard View endorse the brainexplain thesis because of a commitment to the reductive conception of causal explanation, she does not consider the sort of argument contained in the Expanded Version II. She does, however, have some negative things to say in Chapter 7 about what she calls “the metaphysics of the Standard View,” in particular, about the doctrine of materialism. Because materialism is a close cousin to physicalism, we can reasonably anticipate at least one of her responses to our version of the Argument from Causal Explanation by taking a look at the relevant sections in Chapter 7. The question then is: Do any of her objections against materialism constitute telling blows against physicalism? (That is, do any of these objections undermine premise CE(II)1c of the Argument from Causal Explanation (Expanded Version II)?) Baker distinguishes three kinds of materialism – what she calls “minimal materialism,” reductive materialism, which invokes the relation of strong supervenience, and non-reductive materialism, which is formulated in terms of the relation of global supervenience. The precise nature of these views need not concern us here.1 What is important for our purposes is that the argument most pertinent to our brand of physicalism is the one she directs at nonreductive materialism. Here is her argument: (p. 215) 1
Minimal materialism is the view “all that exists are physical particles and constructions out of physical particles” (p. 213). Reductive materialism holds that “all properties strongly supervene on microphysical properties,” where 'strong supervenience’ is defined as follows. Let A and B be families of properties. Then A strongly supervenes on B just in case necessarily for each x and each property F in A, if x has F, then there exists a property G in B such that x has G, and necessarily if any y has G it has F. (p. 213-214) (See Jaegwon Kim, “Supervenience and Supervenient Causation,” in Spindel Conference 1983: Supervenience, ed. Terence Horgan, Southern Journal of Philosophy 22, suppl.1984, p. 49.) Finally, non-reductive materialism is the view that “all properties globally supervene on microphysical properties,” where 'global supervenience’ is defined like this: A globally supervenes on B just in case any two worlds with the same distribution of B properties have the same distribution of A properties. (p. 214)
In Defense of the Standard View
N1. N2. N3.
325
Either materialism plays a role in inquiry or it does not. If it does, it leads to sterile projects; if it does not, it is idle. Hence, materialism is either sterile or idle.
What supports the second premise? Baker’s reasons can be summed up in the following two arguments. Argument for the Claim That Materialism Leads Us Astray (p. 216-7) L1. When materialism is imported to guide inquiry, it leads us to focus on the physical constitution of things in order to understand their nature or to understand their causal powers. L2. Thus, materialism leads us to focus on the intrinsic, non-relational, physical properties of things in order to understand them. L3. But, with respect to social and behavioral phenomena, this focus is misdirected: the nature of intentional phenomena and intentional causation cannot be understood by focusing upon intrinsic, non-relational, physical properties. L4. Thus, when materialism is put to use in inquiry, it “leads us astray” into sterile research projects.1 Argument for the Claim That Materialism Is Idle (p. 216) The only legitimate possible use of materialism in inquiry is to I1. help us determine whether something exists. But, there is no saying on what physical properties intentional I2. properties supervene. And, in any event, we don’t really need evidence regarding an I3. entity’s supervenience base, since evidence is always available at the putative entity’s own “level of description”.2 Hence, evidence concerning supervenience is either unavailable I4. or unnecessary. Hence, materialism is idle. I5. In giving these supporting arguments, Baker herself goes astray, in our opinion, by not fully appreciating the complexity of the physicalist conception. Physicalism, as we conceive of it, differs from the versions of materialism Baker considers both with respect to the claims made as well as the larger 1
See p. 216, second paragraph.
2
As she says, Our grounds for believing that some things are worth a hundred dollars or that arsenic is a poison (and for disbelieving that Rebecca is a witch) have nothing to do with any grand thesis like (GS) [non-reductive materialism]. In general, we do not appeal to overarching metaphysical theses to decide what there is. (p. 216)
326
Barbara Von Eckardt, Jeffrey S. Poland
framework of “companion assumptions”. Our contention is that the versions she considers are seriously defective and that a properly construed physicalism has the resources to both fend off Baker’s objections and to provide a viable premise in our argument for the Standard View. In what follows, we shall outline some key features of physicalism and then show why Baker’s arguments, however effective they are against the versions of materialism she considers, do not generalize to our version of physicalism.
Some Key Features of Physicalism1 Our summary here must be brief, and we shall focus on a few metaphysical and methodological points relevant to the issues at hand. With respect to metaphysics, the critical physicalist claim concerns the realization relation, rather than either identity, constitution, or supervenience. And, the claim is not that everything (natural) is realized directly by the physical. That would be an overarching metaphysical thesis with little usefulness. Rather, physicalism claims that everything (natural) is realized by things that are themselves ultimately realized by physical2 entities (i.e., by things that are “physically grounded”.) In other words, local intermediate stops are legitimate, and for purposes of assessing physicalism’s usefulness, all important. More specifically, since the realization relation may allow for many intermediate levels of analysis and understanding, high level entities are grounded in the physical base by a realization relation that may be quite theoretically complex. And, perhaps, unlike identity, constitution, and supervenience, the realization relation is open to a wide variety of theoretical elaborations depending upon the type of phenomena involved.3 For example, and of particular relevance to Baker’s arguments, in attempting to understand the realization of some high level property, one need not focus on intrinsic, physical properties. Although in some cases such properties may be an appropriate target of attention, in others a focus upon relational or intermediate level, nonphysical properties may be required. By a “realization theory” is meant a specific
1
See Poland (1994) for details.
2
Of course, the problem of identifying “the physical” entities is one that all physicalists must address and few have addressed effectively. See Poland (forthcoming) for discussion of the problem and a proposed solution.
3
Thus, Baker’s claim on p. 216 that proponents of materialism “... retain the conviction that there is a single metaphysically interesting relation (supervenience) between these familiar [psychological, social, etc.] properties and fundamental physical properties”, does not apply to our physicalism since the realization relation admits of diverse theoretical elaborations.
In Defense of the Standard View
327
theory of how some entity is realized by physical or physically-based entities;1 realization theories for high-level properties typically do not mention microphysical properties, intrinsic or relational. It follows from these points that, although there may be no way to determine what physical properties are involved in the realization of some intentional property, there may well be a way to identify the physically-based properties involved: a realization theory will enable this.2 Turning to some methodological points, first, the role of physicalism in inquiry is not as limited as Baker suggests. Physicalism does not call for a focus upon the realization by physical, or physically-based, properties (whether intrinsic or relational) as the key to all understanding of some phenomenon of interest. In particular, physicalism does not hold that “the nature” of everything is physical; rather, the claim is that everything, whatever its nature, is physically-realized. “The nature” of something puts serious constraints on whether and how it is in fact realized, but the discovery of such a nature is a matter of independent investigation. Second, the role of physicalism in determining whether some putative entity actually exists is much more complex than has usually been appreciated. Physicalism does indeed impose constraints, but the decision of what to do if the constraints are apparently not satisfied is complicated by the following considerations. There is a wide range of options from which to choose: eliminate the entity, revise the physical base, show that the constraints are in fact satisfied, claim that they are satisfied without showing it, defer a decision until better evidence is available, reject physicalism. Only the last is an antiphysicalist option. Which option is most appropriate in a given case is a matter of appropriately complex decision making in the face of relevant evidence, background theory, and the like. It is not part of the physicalist picture to claim either that there is always one option that physicalism requires or that physicalism is indefeasible. Further, in light of this, physicalism does not entail that any type of consideration (be it a type of evidence or a general principle) has absolute priority. In particular, the pressure exerted by physicalism is always potentially counter-balanced or overridden by other sorts of information, and vice versa. The quality and relative strength of any type of consideration may vary from case to case. Thus, decision making in a physicalist context is not the highly simplistic matter it is often portrayed as being.
1
A given property might have more than one realization theory associated with it if there is more than one way that physically-based entities can realize it.
2
See Von Eckardt (1993) for discussion of a variety of “theories of content determination” that exemplify the variability in the types of resources a realization theory can appeal to.
328
Barbara Von Eckardt, Jeffrey S. Poland
Finally, it is not the case that the only legitimate use of physicalism is to determine whether some putative entity exists. Physicalism has a number of useful and important roles to play in inquiry. This is, in part, because the realization relation can be theoretically elaborated in a variety of ways and at a variety of levels of analysis. And, it is in part because our scientific and practical interests often are served by reaching down in the direction of the physical basis of some phenomenon. Examples of other uses of physicalism include defining research questions (e.g., questions concerning how some phenomenon is realized) which lead to an increase in understanding of various phenomena, suggesting ways to gain greater control over some phenomenon, suggesting ways of understanding failures and breakdowns of various sorts, and suggesting ways of integrating phenomena in apparently disparate domains.
Replies to Baker’s Arguments Against Materialism Strategically, the situation is this: we will defend the Doctrine of Physicalism (viz., premise CE(II)1c of the second expanded version of the Argument from Causal Explanation) against Baker’s contention that “materialism” is either idle or sterile. We will argue that the charge is false with respect to the version of physicalism just described by showing that the supportive arguments layed out above do not apply to physicalism so conceived. With respect to the argument for the claim that materialism leads us astray, we grant that premise L3 is correct and that focusing upon physical properties is inappropriate when trying to understand the nature of social and behavioral phenomena. However, L1 and L2 are false with respect to physicalism. Physicalism does not require that we focus upon the physical realization (or constitution) of things when trying to understand their nature. Nor does physicalism preclude a relational and higher-level understanding of causes when that is appropriate for understanding causal powers of some entity (e.g., social causes.) In general, physicalism does not inevitably lead to a focus on intrinsic, non-relational, physical properties no matter what the purpose at hand. Thus, the argument fails to show that physicalism, as opposed to some versions of materialism perhaps, leads us astray when it is employed in inquiry. With respect to the argument for the claim that materialism is idle, the first important point is that, as observed in the last section, adjudicating existence questions is not the only legitimate use for physicalism. A more important use is contributing to our understanding of phenomena. Here physicalistically
In Defense of the Standard View
329
guided science plays an important function vis-a-vis the phenomena and truths of commonsense. Cognitive science helps us understand how we exercise our cognitive capacities (e.g., how we remember, how we read, how we produce language etc.). Medicine and physiology explain why we get sick and what we can do to stay healthy. In addition, lower level scientific fields frequently shed light on the phenomena studied by upper level fields. Biology studies metabolism, cell division, reproduction. But it is biochemistry that explains how metabolism, cell division, and reproduction take place. And there are other uses as well. Understanding the underlying mechanism of some higher level phenomenon often helps us to gain technological control of the phenomenon. The understanding of physical forces is the basis for engineering. The understanding of human physiology and biochemistry is the basis for psychopharmacology. Finally, physicalism can also play a heuristic role in inquiry. Knowledge of some lower level domain can sometimes contribute to advances at the higher level. For example, on the basis of theoretical developments at the atomic level, the working out of the periodic table led to the discovery of new chemical elements. Understanding how the brain works has contributed to the recent interest in connectionism as a psychological theory. Thus, if these points regarding other uses of physicalism are correct, then premise I1 is clearly false with respect to physicalism as we conceive of it. Premise I2, on the other hand, is likely true but irrelevant and potentially misleading with respect to our version of physicalism. It is true that, with respect to intentional phenomena, the microphysical objects and properties which ultimately realize such phenomena are of no interest whatsoever and are in all likelihood not identifiable. However, it is not the case that we cannot identify the “physically-based” properties and objects which are involved in realization. Thus, the intermediate fabric of a particular realization is potentially accessible and important for some, but of course not all, purposes. Finally, premise I3 is, we think, false and quite misleading regarding the methodological role of physicalism in inquiry. With respect to the issue of determining if some putative phenomenon really exists, the idea that we don’t really need evidence regarding an entity’s supervenience base, since evidence is always available at the putative entity’s own level of description seriously underestimates the complexity of decision making in this context. We can agree that there is little call for physicalistically-guided science to vindicate (or fail to vindicate) the entities of commonsense. But it does occasionally happen. We no longer explain behavior in terms of the four humors and that is, undoubtedly, because of an increased understanding of human physiology. And, deep suspicion regarding reports of paranormal phenomena are partially rooted in skepticism about the existence of underlying physical mechanisms.
330
Barbara Von Eckardt, Jeffrey S. Poland
Furthermore, in science, lower level knowledge often helps adjudicate questions about the existence of higher level entities. Neurology has provided support for the distinction between short-term and long-term memory. Mendel’s gene hypothesis was, to a large extent, vindicated by biochemistry. Priestley’s oxygen hypothesis was ultimately vindicated by molecular chemistry. These observations underscore the points made in the last section that the pressure physicalism exerts is always in a context of other considerations some of which may be at odds with physicalist constraints. Since the quality and relative strength of any type of consideration may vary from case to case, nothing has absolute epistemic privilege. And, further, in a case of conflict, there are numerous decision options that are possible. The selection of an option will depend upon the features of the case at hand. Premise I3 fails to capture this complexity. Thus, the second horn of Baker’s dilemma is not defensible: physicalism has a number of actual legitimate uses in inquiry; it does not require us to focus upon useless or inaccessible properties; it is not always outweighed by other more epistemically relevant considerations ; and it is, hence, far from idle.
Conclusion One of the arguments for the Standard View Baker considers is an argument from causal explanation. Baker attempts to undermine this argument by arguing against its first premise – the brain-explain thesis. Her arguments are twofold. First, she takes issue with the reductive conception of explanation that she believes supports the brain-explain thesis. Second, she argues against the brain-explain thesis directly, by attempting to show that it is false, whether or not the Standard View is true. In the first part of the paper, we examined both of these negative arguments and showed that neither is successful. Thus, we take Baker’s attempt to undermine the Argument from Causal Explanation to be a failure. We then proposed an expanded version (Expanded Version II) of the Argument from Causal Explanation that we believe is defensible. This argument relies, not on a reductive conception of causal explanation, but on the doctrine of physicalism, to support the critical first premise (the brain-explain thesis). In the second half of the paper, we considered how Baker might respond to our expanded version of the Argument from Causal Explanation by drawing on her discussion of “the metaphysics of the Standard View,” specifically, the doctrine of materialism. We concluded that Baker’s arguments in
In Defense of the Standard View
331
support of the critical second premise in her argument against materialism (N2) do not generalize to physicalism as we construe it. Thus, premise CE(II)1c of the second expanded Argument from Causal Explanation is not undermined by this objection, and, assuming that there are no other more telling objections, we can conclude that the argument is sound and provides good grounds for accepting the Standard View.
References Antony, L. M. and Levine, J. (1997), “Reduction with Autonomy”, in: J. Tomberlin, (ed.), Philosophical Perspectives 11: Mind, Causation, and World, pp. 83-105. Baker, L. R. (1995), Explaining Attitudes: A Practical Approach to the Mind. Cambridge: Cambridge University Press. Hill, C. S. (1997), “Critical Study”, Nous 31(1): 132-142. Poland, J. (1994), Physicalism: The Philosophical Foundations. Oxford: Clarendon Press. Poland, J. (forthcoming), “Chomsky’s Challenge to Physicalism,” in: L. Antony and H. Hornstein (eds.), Chomsky and His Critics. New York: Basil Blackwell. Sanford, D. H. (1996), “Demolishing the Standard View: A Review of Lynne Rudder Baker’s Explaining Attitudes: A Practical Approach to the Mind”, Behavior and Philosophy 24(2). Von Eckardt, B. (1993), What is Cognitive Science? Cambridge, Massachusetts: The MIT Press.
332
Thomas Baldwin
ON CONTEMPORARY PHILOSOPHY OF LANGUAGE, ONTOLOGY AND MORAL THEORY
THOMAS BALDWIN Nearly Logic All modern discussions of logical form take their point of reference from Davidson’s revival of this conception in papers such as ‘The Logical Form of Action Sentences’1. In his response to critics of this paper Davidson spelled out the ideal which guided his use there of this conception: The ideal implicit in the paper is a theory that spells out every element of logical form in every English sentence about actions. I dream of a theory that makes the transition from the ordinary idiom to canonical notation purely mechanical, and a canonical notation rich enough to capture, in its dull and explicit way, every difference and connection legitimately considered the business of a theory of meaning. The point of canonical notation so conceived is not to improve on something left vague and defective in natural language, but to help elicit in a perspicuous and general form the understanding of logical grammar we all have that constitutes (part of) our grasp of our native tongue.2
This is an ideal which Leibniz would have recognised, although its more recent point of origin is in the work of Frege and Russell, to whom we owe the conception of logical form. In the last two sentences of this passage, however, Davidson indicates an important point of disagreement with both Frege and Russell, who regarded natural language as ‘vague and defective’ in ways which rendered it unsuitable as a vehicle for proper science. Their interest in logic was guided by the aspiration to construct a ‘logically perfect language’, free 1
‘The Logical Form of Action Sentences’ in The Logic of Decision and Action ed. N. Rescher (University of Pittsburgh Pres, Pittsburgh PA: 1967) Reprinted in Davidson’s Essays on Actions and Events (Clarendon, Oxford: 1980) pp. 105-122 and my page references are to this edition.
2
‘Criticism, Comment, and Defence’ in Essays on Actions and Events (op. cit. note 1) p. 123.
333
Nearly Logic
from the imperfections of natural language and suitable for the construction of rigorous scientific theory. So for them logical form is fundamentally the logical form of thoughts, propositions or facts which natural language is inherently liable to distort. Davidson, by contrast, follows the young Wittgenstein in holding that everyday language is in perfectly good order as it is,1 so that logical form is always the logical form of our language as it is. Hence for Davidson translation into the ‘canonical notation’ of logico-semantic theory serves only the purpose of clarifying and revealing logical form so conceived. It does not belong to the project of explication through replacement by idioms congenial to the purposes of natural science which one finds in the work of Frege and Russell’s successors, Carnap and Quine.
I We shall see below that, despite this disagreement with Frege and Russell, Davidson detaches logical form from the apparent structure of natural language to a degree which threatens to reintroduce their position. But first I want to briefly situate Davidson’s ideal in relation to Wittgenstein’s later writings in which he is notoriously critical of his earlier project of logical analysis. Wittgenstein’s self-criticism rests primarily on a recognition that that project had been founded upon a mistaken presumption, namely that the semantic relationships which inform the use of natural language are founded upon simple name/object relationships which logical analysis should be able to identify. In place of that simple picture, Wittgenstein stresses the variety of semantic dimensions which inform our uses of language, the contributions of social practice to our language-games, and the complexity inherent in all name/object relations, however apparently simple. Followers of Wittgenstein might argue that it is an implication of this critique that Davidson’s ideal is misguided in its aspiration to develop a canonical notation which suffices to capture in a ‘mechanical’ way ‘every difference and connection legitimately considered the business of a theory of meaning’. For, they argue, it is precisely an implication of Wittgenstein’s later position that there are ‘differences and connections’, for example concerning the first-person and third-person uses of psychological concepts, which are inherent in the practice of our languagegames in ways which cannot be just laid out in a mechanical canonical notation.
1
Tractatus-Logico Philosophicus (Routledge Kegan Paul, London: 1922) 5.5563.
334
Thomas Baldwin
In response to this point Davidson would, I think, observe that the differences and connections he seeks to capture in his canonical notation are only those which arise from the general features of the language which a ‘theory of meaning’ encompasses, and do not include those points which arise from the understanding of specific concepts. So, just as he does not seek to capture the inference from ‘a is older than b’ to ‘b is younger than a’, he can likewise admit that his general ideal will not capture the distinctive features of particular psychological concepts such as belief. Wittgensteinians may then object that this does not meet their point, in that, say, the distinctive use of first-person psychological verbs as avowals and not descriptions is in fact quite general. But to this Davidson can respond that his ‘paratactic’ account of psychological verbs does in fact capture this point within the framework of his ideal theory1. The debate on this and related points can be continued, and I think that at some level there is an inescapable conflict between the hostility to anything merely ‘mechanical’ that is characteristic of Wittgenstein’s later writings and Davidson’s ideal of a purely mechanical canonical notation. But setting that point aside, because the application of Davidson’s ideal is restricted to clarifying general connections there is no necessary opposition between this ideal and the much more diverse explorations of the assumptions inherent in particular uses of language which are suggested by Wittgenstein’s later writings. Right at the end of the passage quoted above Davidson writes that our understanding of logical grammar constitutes only ‘(part of) our grasp of our native tongue’. Without the parenthetically implied other part Davidson would indeed appear to be vainly attempting to reanimate the ideal of the Tractatus; but the parenthetical addition leaves space for all those further dimensions of meaning whose investigation is characteristic of Wittgenstein’s later Philosophical Investigations (some of which are indeed addressed by Davidson in the context of his account of radical interpretation). Hence it is, I think, possible to reconcile Davidson’s ideal with much of the general thrust of Wittgenstein’s later writings. But Wittgenstein’s later writings also suggest the need for caution concerning the connections between grammar and logic, in particular concerning the danger of relying on oversimple assumptions about the logical theory within which the logical grammar of natural language is to be elucidated. Thus Wittgenstein remarked in connection with Moore’s paradox that ‘logic isn’t as simple as logicians think it is’
1
see ‘Moods and Performances’ in Davidson’s Inquiries into Truth and Interpretation (Clarendon, Oxford: 1984) pp. 120-1
335
Nearly Logic
(Letter to Moore, 19441). Although Davidson explicitly disavows an a priori commitment to any particular logical theory, it is notorious that in practice he ends up recommending accounts of logical form which involve translation into a canonical notation whose basic structure is that of classical first-order logic; and this then exposes him to the charge of riding roughshod over the natural language whose logical grammar he aimed to clarify. James Cargile has expressed this objection with admirable clarity: First: we might think that this sentence <‘Shem kicked Shaun’> consisted of two names and a two-place relation ...... But this is wrong! Wrong? Yes, wrong! Second: the sentence really is of a three-place relational form, with two names and an existential quantifier. An existential quantifier? Where is it?2
I shall later discuss this case in some detail, and we shall see that there is more to be said for Davidson’s prima facie odd proposal than Cargile allows. But the general anxiety here is that where a theorist approaches natural language with only a limited palette of logical techniques, the result is liable to be altogether too crude to do justice to the subtleties of natural language and thus to invite the contrast characteristic of Strawson’s early writings, between the arid elegance of the theories propounded by the ‘formal logician’ and the rich complexity of the logic of ‘ordinary speech’ which transcends the capacity of formal logic.3
II Davidson’s project of providing a systematic semantics of natural language modelled on Tarski’s work rests on the hope that this contrast is overdone. But the issues which arise here are nicely shown up by a thesis which gave rise to some of Strawson’s early doubts and which Davidson himself explicitly advocates – the thesis that ‘Someone interested in logical form ... would maintain that “All whales are mammals” is a universally quantified conditional’4. For where, we can ask in the spirit of Cargile, is the conditional (‘if ..., then ...’ or something similar) in this English sentence? Davidson’s reply involves the claim that underneath the misleading ‘surface grammar’ of language there is a ‘deep structure’ which embodies the logical form of the sen1
This letter is among Moore’s papers in Cambridge University Library. It is of course disputable whether one should regard Moore’s paradox as a logical contradiction at all.
2
J. Cargile ‘Davidson’s Notion of Logical Form’ Inquiry 13 (1970) pp. 137-8.
3
P. StrawsonIntroduction to Logical Theory (Methuen, London:1952) p. 232)
4
Essays on Actions and Events (op. cit. note 1) p. 138.
336
Thomas Baldwin
tences of the language.1 This claim, with its suspicion of surface appearances and its invocation of a hidden deeper reality, manifests an unexpectedly Platonist aspect of Davidson’s philosophy of language, though perhaps his emphasis on logical forms should have provided warning enough. For Davidson here opens up a space for hypotheses about a deep structure ‘beneath’ the surface of ordinary language which can have features, such as conditional constructions, which are simply absent from surface grammar. It is, then, to this that he appeals when he holds that complaints such as Cargile’s are beside the point – for they concern the wrong object, the appearance instead of the reality. It is this dualism of a misleading surface structure and a deep structure in which real logical form inheres which threatens to reintroduce into Davidson’s programme an analogue of Russell’s contrast between the imperfections of ordinary language and a logically perfect language (Wittgenstein’s Tractatus bears witness to the same tension)2. But Davidson does not have Russell’s motivation of constructing a language fit for science, so it is all the more pressing to ask why Davidson postulates a separate deep structure as the bearer of logical form. Its role for him cannot be just that which the conception of deep structure has for, say, Chomsky – that is, in the first instance linguistic, but then psychological and ultimately biological as an element within the individual speaker’s biologically based real I-language.3 For deep structure so conceived is a posit of theoretical natural science (human biology), and another of Davidson’s dualisms, the contrast between the explanations characteristic of strict science and those which depend upon interpretative rationality, precludes ascribing Chomsky’s conception to Davidson since Davidson’s deep structure captures logical form and is therefore an aspect of the theory of interpretative rationality. For Davidson, therefore, the legitimation of deep structures must lie in their role in elucidating the rationality of thought and talk, in particular, the validity of patterns of valid inference. But since deep structures so conceived are not posits of theoretical science, whose differences from the manifest world of ordinary life we are familiar with, it remains to be shown that there is good reason to introduce the Platonist dualism of deep form versus surface grammar, and the suspicion will always be that where contrasts do arise this is 1
Essays on Actions and Events (op. cit. note 1) p. 137.
2
Although everyday language is here said to be ‘in perfectly good logical order’ (5.5563) it also said to disguise thought (4.002) in such a way that Russell’s distinction between ‘apparent’ and ‘real’ logical form is required (4.0031). The task of philosophy is then clarify thoughts through logical analysis (4.112).
3
see N. Chomsky Foundations of Language (Praeger, London: 1986) p. 40.
Nearly Logic
337
simply a consequence of the limitations of expressive power inherent in the logical theories employed by the theorist. The theorist postulates a separate deep structure for his logical forms just because he lacks a logical theory which can adapt itself to the manifest order of natural language. The Aristotelian response to Platonist dualism has always been to maintain that forms are neither a higher, nor a deeper, domain of being, but are instead the forms of ordinary, manifest, things. When this response is applied to logical form, therefore, the Aristotelian will hold that logical form is the form of ordinary language, and that there should be no need for Davidson’s separate deep structures as bearers of logical form. It does not follow that the Aristotelian will hold that logical form should always follow surface grammar: for the concerns of logic are not those of grammar. But the Aristotelian will seek to minimize the insertion of extraneous logical structures into accounts of logical form simply in order to validate inferences whose validity does not appear to depend on them. How, then, will the Aristotelian characterise the logical form of ‘All whales are mammals’? Certainly not as a universally quantified conditional. For this treatment precisely violates the principle just enunciated. Instead, in a manner only slightly updated from Aristotle himself, he will insist on the presence here of a restricted universal quantifier, giving the statement the form ‘(∀x: A(x))(B(x))’. For reasons that are to me obscure first-order logic is standardly formulated in a way which does not allow for restricted quantifiers of this kind. Instead it is just the single-sorted, unrestricted, quantifiers such as ‘(∀x)(..x..)’ which are admitted; and of course if that is all that is available, then the best one can do to capture the logical form of ‘All whales are mammals’ is to assign it the form ‘(∀x)(A(x)H B(x))’. But this is only a reason for giving it this latter form if there is some objection in principle to a logic which allows for restricted quantifiers – if, as Davidson implies,1 it is only within a single-sorted logic that one can properly systematise and validate such inferences as: All whales are mammals Moby Dick is a whale so: Moby Dick is a mammal. Yet there is no difficulty in generalising the rules for the unrestricted quantifiers to apply to restricted quantifiers. Suitable rules for the universal quantifier are:
1
Essays on Actions and Events (op. cit. note 1) p. 138.
338
Thomas Baldwin
(∀E)* (∀x: A(x))(B(x)), A(y) /- B(y) (∀I)* If Γ, A(y) /- B(y), then Γ /- (∀x: A(x))(B(x)) as long as ‘y’ does not occur in Γ and ‘y’ is free for ‘x’ in ‘A(y)’ and ‘B(y)’ Given these rules, Davidson’s little argument is just an immediate instance of (∀E)*; and it is equally easy to see that most of the patterns of inference familiar from the syllogism (Barbara, contraposition etc.) can be readily derived. Similar rules for the restricted existential quantifier are equally easy to formulate: (∃I)* A(y), B(y) /- (∃x: A(x))(B(x)) (∃E)* If Γ, A(y), B(y) /- C, then Γ, (∃x: A(x))(B(x)) /- C as long as ‘y’ does not occur in Γ and C With these rules it is again easy to derive such familiar principles as conversion - (∃x: A(x))(B(x)) /- (∃x: B(x))(A(x)). It will be obvious too that all the usual rules for unrestricted quantifiers are instances of these rules where the restricting predicate (‘A(x)’ as formulated above) applies to everything in the domain. So one can use ‘(∃x)(B(x))’ as an abbreviation of ‘(∃x: x = x)(B(x))’ etc., and it is then easy to prove the following equivalences: (∃x: A(x))(B(x)) -//- (∃x)(A(x) & (B(x)) (∀x: A(x))(B(x)) -//- (∀x)(A(x)H (B(x)) These equivalences show that the generalised rules of inference for the restricted quantifiers do not generate theorems or derivations that were not already represented within the familiar system of first-order logic for unrestricted quantifiers. But the fact that the old system had a way of representing all the derivations and theorems of the extended system does not show that it properly represented their logical form. On the contrary, it misrepresented them by making the inferences in question depend on properties of the conditional and conjunction: inferences such as Barbara simply involve quantifiers and it should strike us as bizarre to hold that it is dependent upon the transitivity of the material conditional. Only a perverse commitment to logical minimalism could lead a theorist to think that the equivalences above show that restricted quantifiers are to be reduced to unrestricted ones, so that the resulting reduced derivations capture the logical form of the originals. This form of argument is no better than that which would lead someone to hold that the right way to capture arguments in sentential logic is to translate them all into Sheffer stroke notation and then insist on using that notation to represent their logical form. In both cases the reductive strategy introduces a wholly unnecessary ‘Platonist’ dualism between surface structure and logical form,
339
Nearly Logic
which is avoided by the ‘Aristotelian’ strategy of enriching the logical theory so that it does justice to the structure of the language. But what of semantics, which has always been central to Davidson’s conception of logical form? Again, restricted quantifiers require only a minor generalisation of the usual semantics for the unrestricted quantifiers: using the familiar conception of satisfaction by sequences and indexed variables in the object-language, the clauses in the definition of truth run as follows: s sats. (xi: A(xi))(B(xi)) iff (∀s′i s: s′ sats. A(xi))( s′ sats. B(xi)) s sats. (xi: A(xi))(B(xi)) iff (∃s′ i s: s′ sats. A(xi))( s′ sats. B(xi)) It is worth noting that the clause here for the restricted universal quantifier gives the usual Boolean reading for the case where the restriction is empty. Intuitively, the truth-condition for ‘(∀x: A(x))(B(x))’ is that everything which is A is B, and this condition is trivially satisfied where nothing at all is A. As with derivations such as Barbara earlier, the point here does not concern the material conditional, and it is therefore right that it is not made to depend on this conditional’s truth-conditions. Relativising these clauses to model-theoretic interpretations is immediate, and it is then obvious that semantic analogues of the familiar equivalences can be established: (x:i A(xi))(B(xi)) =//= (xi)(A(xi) & (B(xi)) (x:i A(xi))(B(xi)) =//= (xi)(A(xi)0 (B(xi)) With these equivalences in place it is then easy to extend the familiar proofs of soundness and completeness from first-order logic with unrestricted quantifiers to the extended logic with restricted quantifiers. So semantical considerations provide no reason for preferring a logic with unrestricted quantifiers to one which also admits restricted quantifiers. Thus, although there are further complications which I have not addressed (e.g. concerning ‘donkey’ sentences1) the ‘Aristotelian’ strategy of discerning logical form in the surface structure of ordinary language is in this case fully vindicated.
III Among the cases which are obviously problematic for a logic which employs only unrestricted quantifiers are existential statements with attributive adjectives. For ‘Some gnats are large’ is not sensibly construed as ‘(∃x)(Gnat(x) & 1
see G. Evans ‘Pronouns, Quantifiers, and Relative Clauses (I), (II)’ in his Collected Papers (Clarendon, Oxford: 1985).
340
Thomas Baldwin
Large(x))’. It is therefore worth seeing whether some progress can be made in this area once restricted quantifiers are introduced. Clearly, this can help only if further rules are introduced which block the familiar equivalences discussed just now, rules which somehow develop the role of kind terms, or sortal predicates, within restricted quantifiers. For since the semantic role of attributive adjectives is in some way dependent upon the kinds denoted by the kind terms to which they belong, when these adjectives are attached to variables bound by restricted quantifiers one needs to be able to pick up from the variable the kind specified in the restricted quantifier which binds the variable. This then requires a much more radical restructuring of quantifiers than the simple introduction of restrictive predicates discussed so far. In particular, the distinction between kind terms and predicates becomes important, with the former now playing the central role in restricted quantifiers, so that ‘Some gnats are large’ is construed along the lines of ‘(∃ Gnat g)(Large(g))’. This line of thought requires a radical rethinking of the semantic structure of quantifiers, and thus of the proof theory associated with them, within the context of a systematic theory of kind terms. Given the Aristotelian principle that form should be the form of what is manifest in language such a rethinking would not be undesirable though it would import more complexity than is often required. It is not, I think, necessary to attempt to set out such a theory here: for in order to clarify the role of attributive adjectives it suffices to hold that their logical form is such that they are to be understood as qualifying a kind term. Thus ‘George is large’ invites the question ‘Large what?’, and the answer – ‘Gnat’ – provides the completion of the initial statement – ‘George is a large gnat’. Similarly the role of the implicit restricted variable in the construal just now of ‘Some gnats are large’ (explicit in ‘Some gnats are large ones’) can be spelled out to give ‘Some gnats are large gnats’. But how is the role of attributive adjectives when all is explicit, as in ‘George is a large gnat’, to be understood? One standard view builds on a treatment of kind terms as denoting either extensions, the kinds in question (e.g. gnats), or intensions, the property characteristic of the kind, and then construes these adjectives as ‘predicate modifiers’ construed as functions from these denotations to items of the same type – i.e. from extensions to extensions, or from intensions to intensions.1 One problem with this treatment is that it fails to offer any helpful answer to the question as to what function is denoted by an adjective such as ‘large’, for this supposed role of ‘large’ is not clarified when the function is simply identified as that whose value for a kind 1
see R. Montague ‘English as a Formal Language’ in Formal Philosophy (Yale UP: New Haven: 1974) p. 211.
Nearly Logic
341
k is the large k’s. A further objection is that the bare idea of a function does not bring with it any implication that the value of a function from an extension is an extension which is a sub-set of the first extension, or the intensional analogue of this implication; so the standard adjectival inference ‘George is a large gnat; so, George is a gnat’ is not validated by this treatment and has to be regarded instead as a consequence of an extra ‘meaning-postulate’ for the adjective ‘large’.1 Admittedly there are adjectives such as ‘imitation’, ‘fake’ etc. which do not satisfy this pattern of inference, but have instead precisely the opposite implication. But from the perspective of clarifying the logical form of these sentences it would seem desirable to be able to differentiate these groups of adjectives by giving them different general semantical treatments instead of just relying on extra meaning-postulates to separate them. Can one do any better? It seems to me that if one takes seriously the idea which lies behind the role of kind terms as sortal predicates one can regard ordinary attributive adjectives such as ‘large’ as denoting conditions whose application is essentially restricted to the kind specified by the kind term which the adjective qualifies. Thus on this view ‘George is a large gnat’ is construed as saying of George that he belongs among the gnats which are large ones, or, more formally, that George {Gnat g: large(g)}. The point of the initial kind term and restricted variable here is that the identity of the members of the set is given specifically as gnats and the condition ‘is a large one’ is then specified as that for which the standards appropriate to gnats are to be applied. So although the standard adjectival inference is validated by this construal, it is not equivalent to the condition that George {x: gnat(x) & large(x)}, where being a gnat is construed only as a possible property of objects in general whose wholly unspecific identity is then supposed to specify some standard for assessing whether or not objects in general are large. The situation here is therefore on a par with the previous thesis that the restricted quantifier statement ‘(∃ Gnat g)(large(g))’ is not equivalent to the unrestricted quantifier statement ‘(∃x)(gnat(x) & large(x))’. Furthermore, even though gnats like George are insects, it does not follow that George is a large insect, for the condition ‘George {Insect i: large(i)}’ demands that George be large by the standards of insects, which will be different from those for gnats alone. The basic suggestion is therefore that the key to elucidating the semantic role of attributive adjectives is to treat the role of kind terms, or sortal predicates, in identifying objects as fundamental so that the predication of attributive adjectives of such objects is always to be understood as further specified by reference to the kind term by which the objects are identified in the first place. 1
see R. Montague op. cit. note 12 p. 212.
342
Thomas Baldwin
Not all adjectives are, of course, attributive. It might be proposed that those which are not but are straightforwardly descriptive, such as ‘two-legged’, be given the familiar conjunctive construal, according to which ‘George is a two-legged gnat’ is construed as ‘George is two-legged and George is a gnat’. But this offends against the principle that logical form should not import unnecessary logical vocabulary just in order to validate inferences which do not appear to depend on it. It is therefore preferable to give ‘George is a twolegged gnat’ the same basic logical form as ‘George is a large gnat’ which validates the adjectival inference and then to add as an extra meaning-postulate for ‘two-legged’ the fact that the condition under which something is twolegged does not depend on the kind-term by which the thing is identified – i.e. if there is a kind k such that G is a two-legged k then, for any kind k′, if G is a k′ G is a two-legged k′. In effect, therefore, the kind restrictions do not apply to such adjectives; but this is a special feature of them which their form, as adjectives, does not by itself convey.1 Adjectives for the which standard adjectival inference fails fall into two groups. In the case of some, such as ‘alleged’, it is natural to take the view that they are to be understood as sentence operators. The more interesting case is that of ‘fake’ and similar words, which set up an implication contradictory to the standard one. They are not plausibly regarded simply as sentence operators, tantamount to negation; instead, I think, in the statement ‘Alpha is a fake diamond’ the word ‘fake’ should be taken, not with ‘diamond’, but with the preceding words to form the phrase ‘is a fake’ which signifies the relation of faking which is here said to hold between Alpha and the property of being a diamond; in effect, then, from a semantic perspective fake diamonds are not a kind of diamond at all. It will then be a feature of this relation, faking, that objects which are thus related to a property lack the property though in some respects they appear to possess it. This inference is not, however, an immediate consequence of the construal of ‘fake’ (though the failure of the standard adjectival inference is).
1
There is a trade-off here between using translation into a special canonical idiom that introduces conjunction to capture the inferences that are distinctive of these non-attributive adjectives and introducing special meaning-postulates to handle them. My strategy suggests that the latter is preferable since there is no surface distinction between attributive and nonattributive cases. But the point is arguable and is perhaps a case where there is semantic indeterminacy.
343
Nearly Logic
IV The treatment proposed for adjectives offers a way of capturing the adjectival inference which respects the principle that form should be the form of things much as they appear to us and not of some separate deep structure. It is obvious that Davidson’s account of the logical form of action sentences, which is primarily designed to systematise and validate the adverbial inference, fails to satisfy this principle. Thus, according to Davidson, in order to explicate inferences such as: ‘Shem kicked Shaun hard on the knee; so, Shem kicked Shaun hard’ we have to postulate as a deep structure for the first sentence (∃x)(Kicked(Shem, Shaun, x) & Hard(x) & On the knee(x)). As well as Cargile’s objection to this, it can objected that ‘(∃x)(Kicked(Shem, Shaun, x)’ says that Shem kicked Shaun something, and the only natural further specifications of what was kicked are things like a ball, which are not the intended values. Davidson acknowledges the infelicity here, and goes on to say, concerning the intended interpretation of ‘(∃x)(Kicked(Shem, Shaun, x)’: ‘There is an event x such that x is a kicking of Shaun by Shem’ is about the best I can do, but we must remember ‘a kicking’ is not a singular term.1 This implies that Davidson’s proposed logical form should be rewritten as (∃x)(Kicking(x, Shem, Shaun) & Hard(x) & On the knee(x)) or, even better – in the light of the previous discussion, with a restricted kind quantifier, (∃ kicking k)(Did (Shem, k) & To (Shaun, k) & Hard(k) & On the knee(k)) This makes much better sense, but has the effect of implying that the logical form of any action-sentence is such that the verb of action occurs only as a gerund. As a grammatical hypothesis this would be bizarre, but (as I have already observed) Davidson’s conception of logical form is not intended as a contribution to scientific linguistics. What this suggestion does nonetheless imply is that there be a ‘mechanical’ way of obtaining an appropriate action or event sortal from any verb of action, i.e. that the familiar gerund transformation has a central semantic role. This revision of Davidson’s proposal makes it much more palatable. But the ‘Aristotelian’ critic will still complain that even as revised it implies that the 1
Essays on Actions and Events op. cit. note 1 pp. 117-8.
344
Thomas Baldwin
adverbial inference is dependent upon the logic of conjunction, when in fact the inference appears to be directly inherent in the adverbial form, and thus properly handled within a formal system of semantic categories rather than first-order logic. This point may appear congenial to the alternative predicate modifier treatment of adverbs, which construes adverbs as higher-order functions, whose arguments and values are the semantic values of verbs, whatever they are. But, as with adjectives, it turns out that on this approach the adverbial inference is no longer a matter of form at all, but dependent on special meaning-postulates concerning the functions denoted by adverbs. This is particularly hard to accept in the case of adverbs, nearly all of which do sustain the adverbial inference, except for those which can be readily identified as sentence operators, such as ‘allegedly’ and ‘possibly’. ‘Nearly’ (or ‘almost’) is the nearest one can come to adjectives such as ‘fake’, but its range of application is so much wider than that of straightforward adverbs that it is obvious that its semantics as an all-purpose qualifier is rather different. Is there then any alternative that can capture and validate the adverbial inference without importing unwanted conjunctions to do so? A central preliminary issue is whether Davidson was right to introduce reference to a domain of actions, conceived of as particulars, so that adverbs can be construed as somehow denoting properties of these actions. On the one hand, the surface grammar of the language does not include any explicit reference of this kind: on the other, it is very hard to see how to validate the adverbial inference without it. At this point, then, I think the ‘Aristotelian’ goal of locating logical form within the manifest structure of language has to be qualified if the aim of validating inference is to be met. But this does not imply that one has to embrace Davidson’s strategy of making the inferences dependent upon implicit conjunctions. Instead it is possible to adapt the account of the adjectival inference advanced in the previous section, and this will have the added bonus that it also offers a way of handling attributive adverbs, whose existence Davidson both notes, and notes that his account cannot handle.1 This model suggests that ‘Shem kicked Shaun hard on the knee’ be construed as ‘What Shem did was a hard kicking of Shaun on the knee’ – i.e. (The action a: Shem did a at t) {Kicking k: of Shaun(k): hard(k) : on the knee (k)} The use here of the definite article to form a complex singular term is required quite generally in accounts of the ‘logical grammar’ of natural language, since phrases such as ‘What Shem did’ occur throughout ordinary discourse as 1
Essays on Actions and Events op. cit. note 1 pp. 106-7.
Nearly Logic
345
singular terms. It is obvious that contextual parameters play a large part in securing unique reference for such phrases, and time in particular has a central role in individuating actions: we count them by reference to ‘times’, and there is no general pattern of inference from sentences such as ‘Tom walked slowly’ and ‘Tom walked to town’ to ‘Tom walked slowly to town’ precisely because Tom’s walks can be understood to have been done at different times. Hence (as Frege also held) there is good reason to make temporal reference explicit in such descriptive singular terms; and this also helps both to set aside Russellian objections to this treatment of descriptions and to block the notorious slingshot argument on which Davidson relies so heavily.1
1
Reference to possible worlds will also be required to cope with the use of descriptive singular terms in modal contexts as ‘non-rigid’, and it is perhaps contentious to hold that such reference is implicit in the logical grammar of natural language. But I am prepared to pay this price.
346
James E. Tomberlin
JAMES E. TOMBERLIN Logical Form, Actualism, and Ontology Actualism, as I shall understand it here, is minimally the view that there are no objects or entities of any sort that do not actually exist.1 By the lights of actualism, there are no philosophical problems whose solution calls for or requires an ontological commitment to non-actual individuals. Now I harbor a deep skepticism as regards this ontological stance. And in what follows I seek to extend and sharpen the skeptical concerns presented in my (1993, 1996, 1998a, 1998b) and Tomberlin and McGuinness (1994).2 1. A Deontic Case3 Jones, as it happens, has taken up nouvelle cuisine with its laudable emphasis on fresh and unusual ingredients. One weekend, in seclusion, he opts to prepare for himself the remarkable ragoût of wild mushrooms with veal stock and red wine concocted by Alice Waters for her renowned restaurant, Chez Panisse. For the preparation, Jones decides, why not use wild mushrooms he gathered from the nearby woods just yesterday? A splendid dish indeed, he observes upon dining. But alas, some time later, Jones, still home alone and miles from the nearest person, is rendered comatose. Several of the wild mushrooms were highly toxic and Jones, alone and physically incapable of conveying his plight, faces certain death. In this situation, I take it, (1) and (2) are true but (3) is false: (1) For any individual x, if x is a moral agent who is available and able to come to Jones’s assistance, x ought prima facie to provide Jones with aid. 1
There are grades of actualism. Alvin Plantinga (1985), for example, endorses actualism as the view that there neither are nor could have been objects that do not actually exist. But Nathan Salmon (1987) accepts actualism only by affirming the first half of Plantinga‘s characterization while explicitly rejecting the second (and modal) half. In addition, there are more technical characterizations of actualism in Menzel (1990) and Fitch (1996). As the reader may verify, my discussion here applies to all of the above. For extensive references on actualism see the bibliographies of Tomberlin and McGuinness (1994) and Tomberlin (1996).
2
The argument of Tomberlin (1993) is unique among this battery: there it is argued that actualism has no viable interpretation of quantification. I stand by that argument.
3
As set out here, the present case weaves together features of Case One and Case Two in Tomberlin and McGuinness (1994).
Logical Form, Actualism, and Ontology
347
(2) No actual moral agent is available and able to come to Jones’s assistance. (3) For any individual x, if x is a moral agent who is available and able to come to Jones’s assistance, x ought not prima facie to provide Jones with aid. If so, however, the actualist cannot treat (1) and (3) as universally quantified material conditionals. For suppose otherwise. Since no actual individual satisfies the open sentence x is a moral agent who is available and able to assist Jones, every actual individual satisfies both (1a) and (3a): (1a)
(x is a moral agent who is available and able to assist Jones)H(x ought prima facie to provide Jones with aid).
(3a)
(x is a moral agent who is available and able to assist Jones)H(x ought not prima facie to provide Jones with aid).
But then (1) and (3) are both true, after all. Denied the interpretation of (1) and (3) as universally quantified material conditionals, what alternative is open to actualism? The above difficulty naturally suggests that the actualist construe the notion of conditionality at work in (1) and (3) in such a way that a conditional of the sort in question does not come out true just because its antecedent is not in fact satisfied. This in turn suggests that the actualist try one of these interpretations: take (1) and (3) as universally quantified strict conditionals, nomic conditionals, soft (or hedged) laws, counterfactuals, or statements of conditional obligation. But against this proposal I document in Tomberlin (1996 and 1998a) that none of the above interpretations are available to the actualist. These considerations lead directly to a serious challenge for actualism: Challenge One. With objectual quantification,1 provide an account of (1) and (3) satisfying these conditions: (1) and (2) are true, (3) is false, and the quantifiers range over actual individuals only. But let’s not stop here. For there awaits another but very different problematic case for the actualist.
1
As opposed to substitutional quantification. For criticisms of substitutional quantification, see Tomberlin (1990a, 1993, and 1997).
348
James E. Tomberlin
2. Intentional Verbs Like Chisholm (1986), assume an actualism that includes these key ingredients: an ontology confined to actual individuals and attributes (some exemplified, others not); a relational account of believing; and a Russellian treatment of definite descriptions. Since a position of this sort requires that no person ever has genuine de re beliefs toward non-actual individuals, the question becomes acute as to the proper treatment of items such as (4) Ponce de Leon searched for the fountain of youth. After all, if Ponce de Leon really did search for the fountain of youth, hoped to find it and the like, he presumably can entertain beliefs of or about the object of his search. To deny that Ponce de Leon had any de re beliefs toward the fountain of youth, therefore, dictates that actualists embrace one of these alternatives: (a) deny the truth of (4); or (b) interpret (4) so that its truth does not stand Ponce de Leon in a genuine de re relation to the non-actual fountain of youth. Of these alternatives, I take it, the former is wholly implausible. And consequently we have a second test for actualism: Challenge Two. Tender a credible treatment of (4) meeting this constraint: (4) is true but its truth does not require Ponce de Leon to bear a de re relation to some nonactual individual.
3. How Not to Meet the Challenge A. The Chisholm-Kaplan Model Chisholm, quite correctly, insists on the truth of (4). Working against the background of his theoretically elegant account of believing as a relation between a believer and an attribute,1 Chisholm proposes that the intentional verb in (4) be taken as expressing a dyadic searched for relation holding between Ponce de Leon and an attribute. That is, Chisholm would have us parse (4) as follows (Chisholm (1986), pp. 56-57): (5) Ponce de Leon endeavored to find the attribute of being a unique site of a unique fountain of youth. 1
There is an extended critique of Chisholm‘s ingenious version of self-ascription for believing in Tomberlin (1990b, 1991). The very different formulations of self-ascription in Brand (1983, 1984) and Lewis (1979, 1986) are critically examined in Tomberlin (1987) and Tomberlin (1989), respectively.
Logical Form, Actualism, and Ontology
349
Because there is such an actual attribute even though it fails to be exemplified, Chisholm reasons, the truth of (4) – so construed – in no way requires Ponce de Leon to bear a genuine de re relation to the nonexistent fountain of youth. A similar actualist view is advanced independently by David Kaplan in his classic essay “How to Russell a Frege-Church” (1975). To begin with, Kaplan rightly observes that Russell’s own primary-secondary scope distinction for eliminating descriptions within intensional contexts fails in the case of (4), owing to the fact that the intentional verb there takes no sentential complement. In accord with Chisholm, Kaplan theorizes, model (4) semantically (and ontologically) as the bearing of a dyadic relation between Ponce de Leon and an attribute where the resulting paraphrase of (4) turns out much like (5) above (Kaplan (1975), p. 729). A tempting view indeed for anyone seeking to retain a relational account of items like (4) while opting for an ontology devoid of non-actual entities. And yet, there are serious difficulties with the ingenious Chisholm-Kaplan model. For openers, according to this model (4) demands that Ponce de Leon bear the searched for relation to any attribute identical with the attribute being a unique site of a unique fountain of youth. In the case of Kaplan, given his background framework of standard possible world semantics, this attribute proves to be identical with the attribute being a unique site of a unique fountain of youth or a unique odd number exactly divisible by two inasmuch as they are co-exemplified at every logically possible world. From this, taken in conjunction with the paraphrase strategy illustrated by (5), we may infer that if (4) is true so too is (6): (6) Ponce de Leon searched for the fountain of youth or the odd number exactly divisible by two. Since (4) is true whereas (6) is doubtless false, it seems that (5) is not a proper interpretation of (4), after all.1 Turn now to Chisholm. Though the above objection appears squarely to hold against Kaplan, it does not evidently apply to Chisholm. For he, unlike Kaplan, adopts an intentional account of property-identity: properties F and G are identical only if they are necessarily such that whoever conceives F conceives G (Chisholm (1986), p. 32). And insofar as I understand Chisholm on his pivotal conceptual primitive conceiving an attribute, he will deny that a 1
Here of course it is crucial that (6) not be confused with (6a) Either Ponce de Leon searched for the fountain of youth or he searched for the odd number exactly divisible by two. Though (4) trivially entails (6a) by propositional logic, it is nevertheless possible that (6a) is true when (6) is false.
350
James E. Tomberlin
person may search for the -er without searching for the "-er, if being and being " are one and the same attribute. Nonetheless, whether or not Chisholm avoids the above objection to Kaplan, it is difficult to see how his intentional account of property-identity shuns the disastrous consequence that being and being " are different properties if someone believes that they are distinct. But no matter. For issues surrounding property-identity are notoriously controversial, and the considerations raised above could perfectly well prompt friends of the Chisholm-Kaplan model to seek a finer-grained account of property-identity. And so the question becomes acute: Are there still other grounds for rejecting the Chisholm-Kaplan model? There are indeed: First, consider the schema () x searched for y. Now if the singular terms replacing ‘x’ and ‘y’ each refer to actual concrete individuals, then even under the Chisholm-Kaplan model it seems natural to suppose that the resulting instance of () ascribes a dyadic relation between these very individuals. (When Bob searched for his missing daughter last night, the object of his search really was his daughter, not an abstract property.) Suppose, however, as in (4), that whereas the singular term replacing ‘x’ picks out an actual concrete individual, the one replacing ‘y’ does not. By the Chisholm-Kaplan model, we must hold in effect that ‘the fountain of youth’ as it occurs in (4) refers to an abstract but actual attribute. (After all, (4) is parsed as (5) under this model.) Why exactly should the ‘fountain of youth’ so behave in (4)? For surely we are not to so treat this term as it occurs in either true (7) or false (8): (7) The fountain of youth does not actually exist. (8) Ponce de Leon drank from the fountain of youth. A consequence of the Chisholm-Kaplan model, then, is that a definite description like ‘the fountain of youth’ plays two quite different semantical roles: as it occurs in items such as (7) and (8) this term is genuinely non-denoting; but as this same term occurs in (4) it denotes an attribute. This alleged shift in semantical behavior of a definite description seems impossible to motivate outside of a prior commitment to an ontology without non-actual individuals. Second, let us address the lot of the Chisholm-Kaplan model as regards Property Actualism – the view that there are no genuine properties (or attributes) besides the actual ones. Let be the collection of all and only those properties that actually exist. Then for the actualist ‘the unique nonmember of
351
Logical Form, Actualism, and Ontology
that is Al’s favorite property’ does not refer to a property. Even so, I take it, statements like (9) may be true: (9) Bob wanted to study the unique non-member of vorite property.
that is Al’s fa-
Additionally, it seems plausible to theorize that (9) should be treated on a par with (4) insofar as the definite description following the intentional verb in each case fails to refer to an actual concrete individual. If so, by the ChisholmKaplan model, (9) becomes (10)
Bob wanted to study the property of being a unique non-member of that is Al’s favorite property.
But then, contra Property Actualism, the truth of (9) requires that there is such an actual property as the property of being a unique non-member of that is Al’s favorite property. And consequently, applied to cases such as (9) the Chisholm-Kaplan model is self-refuting.1 B. Adverbialism With the demise of the Chisholm-Kaplan model, how else might the actualist try to accommodate intentional verbs? According to Fitch (1996), when confronted with troublesome cases like (4) and (9), the actualist should “go adverbial”. Roughly, the proposal at stake is to parse (4), for example, as (4a) Ponce de Leon searched-for-a-unique-fountain-of-youthly. Here ‘a-unique-fountain-of-youthly’ behaves as an adverbial modifier of the now monadic predicate ‘searched for’. According to this adverbial model, then, (4) – so construed – does not ascribe a relation between Ponce de Leon and the non-actual fountain of youth. Indeed, thus understood, (4) does not ascribe a relation between Ponce de Leon and anything. Quite the contrary, unlike the Chisholm-Kaplan model, (4) is now seen to ascribe a complex but non-relational property to the single individual Ponce de Leon. As such, we are told, the truth of (4) in no manner requires an actual object to bear a de re relation to some non-actual one. Without worrying over the so far missing semantics for this adverbial treatment of (4), there seem to be telling objections to any such view. Recall schema (): () x searched for y. 1
This critique of the Chisholm-Kaplan model both corrects and extends the argument of Tomberlin (1996).
352
James E. Tomberlin
If the singular terms replacing ‘x’ and ‘y’ each refer to actual concrete objects, let us assume, then even under the adverbial view the resulting instance of () ascribes a dyadic relation between these very individuals. (When, for example, John searched for his missing wallet, he genuinely stood in the dyadic searched for relation to his wallet. Suppose, however, as in (4), that although the term replacing ‘x’ picks out an actual concrete individual, the one replacing ‘y’ does not. By the adverbial model, we are to believe that ‘the fountain of youth’ as it occurs in (4) does not behave as a singular term. (After all, (4) is parsed as (4a) and the only singular term in the latter is ‘Ponce de Leon’.) Under the present interpretation of the adverbial view, therefore, instances of schema () come in (at least) two diverse sorts: when ‘y’ is replaced by a term denoting an actual concrete individual, the instance in question ascribes a dyadic relation; and yet, if the term that replaces ‘y’ fails to pick out an actual concrete object, the instance at issue ascribes a non-relational property to a single actual individual. This alleged shift in semantical behavior of the various instances of () seems incredulous on two counts: first, outside of a prior commitment to actualism, the semantical shift at stake appears impossible to motivate or support independently; and second, any such view requires intolerably that logical form depends on matters of contingent fact – to know what sort of proposition is expressed by an instance of () one must already know whether the singular terms in question do or do not refer to actual but contingent objects. This negative verdict against the adverbial model has so far centered around the pivotal assumption that even under an adverbial treatment instances of () ascribe a dyadic relation should the terms replacing ‘x’ and ‘y’ both refer to actual objects. What happens if this assumption is rejected? Why not, that is, interpret adverbialism in such a way that every instance of () receives the treatment accorded to (4), even when the singular term replacing ‘y’ refers to an actual concrete individual? To my mind, there is a fatal objection to any such view. For even if the searched for relation never holds between actual objects, surely the same cannot be properly maintained for the loves relation – Bob genuinely stands in the latter relation to his missing daughter, Jane. If so, the adverbial model, under the current interpretation, cannot do justice to true statements like (11)
Bob searched for Jane, the missing daughter he deeply loves.
That is, in true (11) the construction ‘the missing daughter he deeply loves’ incontestably ascribes the loves relation between Bob and Jane. And consequently, (11) won’t be true unless ‘Jane’ as it occurs in ‘Bob searched for Jane’ refers to Bob’s missing daughter. But ‘Jane’ does no such thing under the
Logical Form, Actualism, and Ontology
353
present interpretation of the adverbial model, since it does not occur in (11) as a referential singular term. This is no way to preserve actualism. With a negative critique of both the Chisholm-Kaplan model and adverbialism,1 I conclude by repeating this challenge to actualism: Challenge Two. Tender a credible treatment of (4) meeting this constraint: (4) is true but its truth does not require Ponce de Leon to bear a de re relation to some non-actual individual.2,3
References Bealer, George, 1998, “Reply to Tomberlin,” Philosophical Issues 9. Brand, Myles, 1983, “Intending and Believing,” in J.E. Tomberlin, ed., Agent, Language, and the Structure of the World, Indianapolis, Hackett. Brand, Myles, 1984, Intending and Acting, Cambridge, Bradford Books, MIT Press. Chisholm, Roderick M., 1986, “Self-Profile,” in R. Bogdan, ed., Roderick M. Chisholm, Profiles, Dordrecht, D. Reidel. Devitt, Michael, 1998, “Putting Metaphysics First,” Philosophical Perspectives, 12. Fitch, G.W., 1996, “In Defense of Aristotelian Actualism,” Philosophical Perspectives, 10: 53-72. Horgan, Terence, 1998, “Actualism, Quantification, and Contextual Semantics,” Philosophical Perspectives, 12. Kaplan, David, 1975, “How to Russell a Frege-Church,” Journal of Philosophy 72: 716-729. Lewis, David, 1979, “Attitudes De Dicto and De Se” Philosophical Review 88: 513-543. ———, 1986, On the Plurality of Worlds, Oxford, Blackwell. Linsky, Bernard and Zalta, Edward N., 1994, “In Defense of the Simplest Quantified Modal Logic,” Philosophical Perspectives 8: 431-458. ——— and Zalta, Edward N., 1996, “In Defense of the Contingently Nonconcrete,” Philosophical Studies 84: 283-294. Menzel, Christopher, 1990, “Actualism and Possible Worlds,” Synthese 85: 355-389. Plantinga, Alvin, 1985, “Self-Profile,” in J.E. Tomberlin and P. van Inwagen, eds., Alvin Plantinga, Profiles, Dordrecht, D. Reidel. 1
Linsky and Zalta (1994) is an original and important version of actualism. I argue against their view in Tomberlin (1996). They reply in Linsky and Zalta (1996).
2
Bealer (1998), Devitt (1998), and Horgan (1998) are novel and fundamentally different actualistic responses to such challenges. I critically examine their fascinating proposals in work currently in preparation.
3
For rewarding discussion and/or correspondence about actualism and related matters, I am grateful to David M. Armstrong, George Bealer, John Biro, Roderick Chisholm, the late Hector-Neri Castañeda, David Cowles, Michael Devitt, Kit Fine, Greg Fitch, Gilbert Harman, Terry Horgan, David Kaplan, Bernie Kobes, Bernard Linsky, Kirk Ludwig, Bill Lycan, Chris Menzel, Al Plantinga, Greg Ray, Nathan Salmon, Bob Stalnaker, Stephen Schiffer, Ed Zalta, and my colleagues Frank McGuinness, Jeff Sicha, and Takashi Yagisawa. I do not mean to imply, of course, any agreement on their part with what I have argued here.
354
James E. Tomberlin
Salmon, Nathan, 1987, “Existence,” Philosophical Perspectives 1: 49-108. Tomberlin, James E., 1987, Critical Review of Myles Brand, Intending and Acting, Noûs 21: 45-63. ———, 1989, Critical Review of David Lewis, On the Plurality of Worlds, Noûs 23: 117-125. ———, 1990a, “Belief, Nominalism, and Quantification,” Philosophical Perspectives 4: 573-579. ———, 1990b, Critical Review of R. Bogdan, ed., Roderick M. Chisholm, Noûs 24: 332-342. ———, 1991, “Belief, Self-Ascription, and Ontology,” Philosophical Issues 1: 232-260. ———, 1993, “Singular Terms, Quantification and Ontology,” Philosophical Issues 4: 297-309. ———, 1996, “Actualism or Possibilism?,” Philosophical Studies 84: 263-281. ———, 1997, “Quantification: Objectual or Substitutional?,”Philosophical Issues 8: 155-168. ———, 1998a, “Naturalism, Actualism, and Ontology,” Philosophical Perspectives 12. ———, 1998b, “Concepts and Ontology: A Query for Bealer,” Philosophical Issues 9. ——— and McGuinness, Frank, 1994, “Troubles with Actualism,” Philosophical Perspectives 8: 459-466.
What a Noncognitivist might tell a Moral Realist
355
AREND KULENKAMPFF, FRANK SIEBELT What a Noncognitivist might tell a Moral Realist Is being motivated to act in a certain way something which entirely depends on having appropriate desires? Does being motivated to act always require a non-cognitive psychological source? Isn’t it true that we sometimes disregard, even act against, our desires while nevertheless acting motivationally, say, by being motivated simply due to our beliefs about what is morally right, valuable or required of us to do? Is a strict cognitivist account of moral motivation philosophically feasible? Quine has given us the following, somewhat simplified but sharp image that, in a nutshell, contains the non-cognitivist’s answer to those questions. “Imagine a dog idling in the foreground, a tree in the middle distance, and a turnip lying on the ground behind the tree. Either of two hypotheses, or a combination of them, may be advanced to explain the dog’s inaction with respect to the turnip: perhaps he is not aware that it is there, and perhaps he does not want a turnip.”1
Of course, the Quinian dog lacks a conception of morality, and a turnip is not an appropriate substitute for moral virtues and values. Nevertheless, the concise image illustrates basic ideas of the bifurcated non-cognitivist account which lies at the bottom of the so called “Humean theory of motivation” quite well. (For reasons why it is called “Humean”, see Hume (1985) bk.II, part III, sec.iii) To elaborate, let’s refer to those mental states an agent has to be in for her to act intentionally in one way or another, as motivational states of an agent. “Intentionally” is essential here because a lot happens due to us without being motivated at all. Laura’s breaking a Chinese vase for example by slipping accidentally in front of it, is something she did. “She broke the vase” is a true description of what happened because of her. Yet, setting Freudian presumptions of hidden destructive desires aside, she simply slipped on the carpet. Neither did she intend to or had in mind to break the precious thing, nor did
1
Quine (1981) p.54.
356
Arend Kulenkampff, Frank Siebelt
she want to slip. For acts like these to speak of a motivational state of the agent accounting for what she’s done is misleading.1 It is intentional action that requires the agent to be in a motivational state of some sort and the Humean non-cognitivist account is an answer to the question of how such motivational states should be understood. It is a basic idea of non-cognitivism that every motivational state of an agent is a composite rather than a unitary state. Non-cognitivism holds that being in a motivational state is being in a complex mental state, constituted by a pair of two distinct states, a desire-state and a belief state. The former presents the agent the world in a way she would like to have it, the latter with a view or opinions about feasible means to make the world the way she would like to have it. Thus, there is no motivation without both states, beliefs and desires actually being present in leading to an agent’s decision to act. Furthermore, what we desire according to non-cognitivism is not already included in, or is a byproduct of, what we believe. We can desire almost everything, things which we believe both to be satisfiable by some means, and things which we believe are not, things which we think are highly valuable, and things which we think are not. From the Humean perspective there is no prior constraint placed on possible desires by what we believe. There are many formulations of the Humean conception to be found in the literature, the main idea seems to be captured by the following principle (HM):2 (HM)
S constitutes a motivational state for an agent x to act in way 5 iff (a) S includes a desire to have it that1 (b) S includes a belief, that by doing 5 x can make it happen that 1, whereby (c) the belief to make it happen that1 by doing 5 and the desire to have it that1 are only contingently related in S.
There are four striking philosophical assumptions built into (HM) which are worth mentioning for our discussion: 1. The Conceptuality Assumption: (HM) is meant to be a conceptual thesis
1
Of course, when speaking of motivational states, we don’t assume that those states must always be phenomenologically transparent for an agent. There might be hidden desires that motivate an agent in a way that she is not aware of. “Has to be in” in our formulation should not be understood as implying “has to be aware of”. An agent might be in motivational state to act in certain way without being actually aware of being in such a state. See M. Smith (1987) and P.Pettit/M.Smith (1990) for further implications of that suggestion.
2
See D. Davidson (1963) for a classical exposition.
What a Noncognitivist might tell a Moral Realist
357
that covers all possible uses of what we understand as motivation. As such, (HM) is silent regarding the actual contents of desires and beliefs which de facto make up motivational states. In particular, (HM) does not distinguish between moral beliefs and other types of beliefs. Consequently, having a firstperson moral belief, say, about what the morally right act in a situation is, is likewise insufficient for being in a motivational state to act as having nonmoral beliefs about a situation. (HM) has it that belief states, if not accompanied by a suitable desire, are motivationally inert, they can’t motivate on their own. We would never by more inclined to act in one way or in another when lacking suitable desires. This is meant to also include our desires for the morally right. What happens due to us when lacking desires altogether would be as unmotivated as breaking a precious Chinese vase by accidently slipping in front of it. Being in a motivational state to act requires a belief state to be allied by a suitable desire. Only when combined in the right way can they make up a complex mental state that is – to use a phrase of Mc Naughton (1988) p.107 – motivationally “ert” or pushy for one’s actions. 2.The Distinctness Assumption: It is part of the non-cognitivist picture that beliefs and desires are regarded as distinct mental states. What accounts for their distinctness is their fulfilling different functional roles. Of course, as states having propositional content, beliefs and desires could be identical in content. What Oscar believes to be true of him, being happily married, is something which other people not so lucky strongly desire. Nevertheless, believing a proposition 1 (being in a belief-state with respect to a propositional content 1) and desiring a proposition 1 (being in a desire-state with respect to the same proposition 1) show different mental orientations with respect to their contents and those orientations can be further characterized functionally.1 As to desires, it is part of the functional role of such states, that a desire to have it that 1 can be characterized as a state whose role it is to connect belief states with those actions which bring it about that 1, if those beliefs are true. But then belief states and desire states must be distinct states. Both states show – to use a well suited term covering their functional difference – opposite “directions of fit” to the world.2 Believing that 1 is a state that aims for truth, its direction of fit is state-toworld. If an agent who is in a motivational state realizes that what is believed, i.e. that 1, is in fact false, then normally only the belief will be revised, or completely abandoned, and not the desire. Whereas desiring that 1 aims at 1
In what follows no specific theory of propositions is presupposed.
2
See M. Platts (1979) p.256, M. Smith (1987) p.51, and L. Humberstone (1992) for a wider discussion of that notion.
358
Arend Kulenkampff, Frank Siebelt
satisfaction, its direction of fit is world-to-state. Desiring that 1 where the desirer realizes that in fact the world has it that not-1 is not a failing of the state. On the contrary it is the world which has to change in order for a desire to be satisfied, which means that the desire state will normally continue to hold when it is realized that the world has it that not-1. Consequently, a single unitary mental state having the content that 1, could not, on pain of contradiction, occupy both functional roles: as a belief aimed at fitting the world so that its content that 1 represents the world truthfully, and at the same time as a desire aimed at changing the world in order to make it first happen that 1. As to their propositional contents, we cannot in full awareness view the world simultaneously both ways, still unfit for a desire that 1 to be satisfied and yet already fitted by a belief with the same content that 1 longing for a truthful representation.1 3. The Assumption of Contingency: Beliefs and desires are not only distinct mental states according to (HM). Furthermore, as distinct states they are only contingently related in S. Between being in a belief state that 1 and being in a desire state that 5, whatever content 5 and 1 may actually be, there is no necessary alignment. This is a corollary of the Humean denial of necessary connection between distinct existences applied to psychological states. (Another source as to why what we are talking about is rightfully called the “Humean picture of motivation”.) That is, in considering a belief and a desire state and propositional contents 1 and 5, whatever content 1 and 5 may actually be, there is no constraint on combinatorial possibilities. Every combination could describe a possible complex mental state a person might very well be in. According to (HM) every belief can consistently go with every desire and vice versa. (Think of Anscombe’s prominent example of a man who desires a saucer of mud. Although bizarre it is not inconsistent to assume somebody could have it, irrespective of what he believes, even if this includes his belief that to have such a desire is not praiseworthy.) That is what in our view lies behind Hume’s famous dictum that “reason [i.e. belief] can never oppose any passion in the direction of the will [i.e.desire].”2 If some belief states and desire states are nevertheless “constantly conjoined”, this will always be a contingent affair, however uniform, persistent and common their psychological alignment might appear to us.
1
This argument suggests that the Humean theory of motivation could be developed within a broad functionalist approach of characterizing mental states. See B. Loar (1981) and F. Jackson/P. Pettit (1995) for some further suggestions along this line.
2
D. Hume (1985) bk. II, Part III, sec iii.
What a Noncognitivist might tell a Moral Realist
359
4. Motivational states are explainers: There is a close connection between (HM) and the concept of action explanation. What is involved ex ante for an agent to be motivated to act in a certain way is the same as what one normally tries to find out in seeking to explain ex post actum what might have brought the agent to do what she did. Explaining an action by citing the reasons the agent had to perform it requires referring to motivational states. The connection is: for an agent, her motives are her reasons to act, and for an explainer, the goal is to single out reasons an agent had for her action, thereby making it intelligible also to others.1 A crime is commonly regarded as explained – of course, besides knowing who committed it – if its motive is revealed. Knowing the motivational state an agent was in that led to the crime is what the actionexplainer is after when trying to answer explanatory demands. The nice old lady’s murderer was her nephew. For what reason did he commit this terrible crime? His desire to pay his gambling debts and his belief that by killing the old lady he could speed up his inheritance; a cruel but nevertheless well known explanation, indeed. There may be further ideas which some non-cognitivists want to see incorporated in (HM). The four mentioned above seem to us the most widely accepted ones, therefore our picture ends here.
II Quite recently the non-cognitivist theory has been confronted with a multitude of opposing arguments. One challenge comes from philosophers who oppose the application of (HM) within metaethical theory. Today their position is known under the heading “moral realism”.2 Although not a position with sharp boundaries, its core idea centers around the assumption that moral opinions are purely cognitive in content as opposed to expressing favorable attitudes, feelings of approval or disapproval or special sentiments of those persons or social groups who have them. Moreover, being cognitive in content, the realist takes moral opinions to be beliefs whose propositional contents are true or false in virtue of special facts: moral facts. For moral realists the factuality of moral beliefs is built on the idea
1
See again D.Davidson (1963) for the link between explaining an action and making actions intelligible.
2
See, among others, J. McDowell (1998), M. Platts (1979), (1980), D.McNaughton (1988), D.O.Brink (1984), (1989) and J. Dancy (1993), J.J.Thompson/G.Harman (1996) for expositions of this view.
360
Arend Kulenkampff, Frank Siebelt
that there is moral truth in exactly the same sense in which we think there is factual truth. As one proponent puts it “...moral realism insists that there is moral reality which is independent of our moral beliefs and which determines whether or not they are true or false. It holds that moral properties are genuine properties of things or actions; they are (...) part of the furniture of the world. We may or may not be sensitive to a particular moral property, but whether or not that property is present does not depend on what we think about the matter.”1
While moral realism, as outlined in this statement, might still be compatible with (HM), the challenge comes by looking at the consequences when combined with a second assumption many moral realists also accept. Moral realists generally think that moral opinions not only motivate, but that they do so on their own. In fact, many realists uphold the view as special feature of moral reality, that its perception furnishes people with moral beliefs which eo ipso motivate to act.2 Michael Smith3 has called this “the practicality” of moral beliefs. Yet the bifurcated account of section I requires desires to be present in each motivational state but to be so only contingently and, desires are a prime example of non-cognitive states. The following argument then stresses the main point. From (1) (HM) encapsulates a conceptual truth about motivation and (2) Moral opinions express beliefs that aim at truthfully representing moral reality we are forced to also accept (3) Moral opinions have no motivational force on their own. The dilemma for moral realism is now clear. If one accepts the thesis that moral opinions weigh motivationally, are “intrinsically practical” to use Smith’s term, while also holding onto moral realism as outlined above, then (HM) must give us a distorted picture of moral deliberation. Otherwise it will, when combined with moral realism, undermine the practicality assumption, as the above reasoning shows. Contraposing the argument, if moral opinions do have motiva1
D. Mc Naughton (1988) p.7. For similar statements see also M. Platts (1979) p.243, (1980) p.69, D.O. Brink (1984) p.111, (1989) pp. 25-36.
2
There are some notable exceptions. See D.O. Brink (1989) for a realist position that is Humean to some extent as it accepts the contingency requirement (c) in (HM).
3
See M. Smith (1994) p.7.
What a Noncognitivist might tell a Moral Realist
361
tional force on their own then (HM) can’t be a conceptual truth about motivation or else moral realism would have to be abandoned. Moral realists therefore argue that(HM) must be given up when applied to moral deliberation.1 The solution many moral realists favor in light of the above argument stems from the idea that the special motivational force of moral beliefs reflects some kind of necessary alignment of moral beliefs and desires. So, abandoning (HM) means at least that its contingency assumption is being questioned. There is some kind of thick connection between believing an action to be morally required and being motivated to pursue it, they argue. Of course, if this is a true picture of moral deliberation it would prove a rather strong case against the non-cognitivist account. Let’s take a closer look at their arguments.
III As we have outlined in the last section the realist thinks it to be distinctive feature of moral reality that its perception has a special motivational authority on us. As M. Platts puts it “... to recognize the obtaining of, say some desirable moral feature in a possible state of affairs is to desire the obtaining of that state of affairs (though not just that). One cannot see the loyal, the courageous and so forth as the loyal the courageous, etc., without desiring them.” M. Platts ( 1 9 8 0 ) p.80.
J. Dancy and M. Smith2 make similar points, when characterizing moral judgements as being essentially practical, that is, in a strong sense being always motivationally relevant for the choice of ones actions. The same objection to (HM) is often stated by moral realists more indirectly by showing that in some cases we can explain an action sufficiently just by citing beliefs the agent had ex 1
Denial of (HM) by moral realists comes in weak and strong senses. In the strong sense it is the view that there simply are no Humean beliefs and desires. Moral deliberation can’t be analyzed within the bifurcated Humean scheme, not because the scheme fails for moral motivation but because it fails for every kind of motivation, be it moral or otherwise. This is the position stated most clearly in J. Dancy (1993) pp.22-26. It is also, if not that clear, argued for by M. Platts (1979) pp 255-63, (1980) pp.73-82. In a weaker sense denial of (HM) comes by distinguishing between motivating reasons and normative reasons to act. Whereas the former are seen to be captured by (HM), it is the latter that require an Anti-Humean theory of moral deliberation. On such a construal all motivated actions comply to (HM), but some desires are themselves produced by the agent’s beliefs about the moral reasons she has, where those beliefs are acquired through rational deliberation. This somewhat weaker position is argued for by M. Smith in his (1992), see especially pp.343-59, and more recently in his (1994) ch.4.
2
See also J. Dancy (1993) p. 4 and M. Smith (1994) pp 6-7.
362
Arend Kulenkampff, Frank Siebelt
ante.1 As we see it “cannot”, in Platt’s wording, is decisive here. For the noncognitivist might very well concede that moral beliefs are often motivationally relevant to one’s choice of action. He could even go along with the moral realist and accept his view that it might very well be a common feature of moral deliberation. In some sense it could even be a universal trait of moral beliefs that they weigh motivationally, so that basic moral beliefs would always be connected with a suitable desire. Yet regularity, even if universal is not what the moral realist is after. His objection is directed towards the contingency of any such connection within the non-cognitivist account, be it universal or not, among moral beliefs and desires. On the moral realist’s view, to recognize that 1 is morally required, i.e. believing that 1 is morally required, will in a strong modal sense imply a corresponding desire that 1. The idea is expressed, very clearly, in a passage by John McDowell: “Suppose, for instance, that we explain a persons’s performance of a certain action by crediting him with awareness of some fact that makes it likely (in his view) that acting in that way will be conductive to his interest. Adverting to his view of the facts may suffice, on its own, to show us the favourable light in which his action appeared to him. No doubt we credit him with an appropriate desire, perhaps for his own future happiness. But the commitment to ascribe such a desire is simply consequential on our taking him to act as he does for the reason we cite(...)”2
If this objection against (HM) is sound then to be motivated to act in a certain way doesn’t require the agent to have independent suitable desires in addition to moral beliefs. Desires do not function as “extra components” in full specifications of what is required for being motivated to act. At least in cases of moral beliefs, desire already arises alongside beliefs as their necessary companion. The appetitive then is simply a consequence of having suitable beliefs. Thus stated, the objection to (HM) says that the non-cognitivist theory fails insofar as moral motivation cannot be regarded as the product of a contingent or a merely inductive alignment between moral beliefs and desires. In a certain sense being motivated to act is the product of moral beliefs alone. Although many philosophers are sceptics in taking beliefs to be sufficient in leading to intentional action by themselves, the realist has an answer: it is that moral beliefs, unlike other beliefs, are necessarily connected to having appropriate desires in motivational states, so that having a moral belief is always sufficient (and of course necessary) for a suitable desire to arise. The stronger thesis 1
See D. Mc Naughton (1988) pp 22-3.
2
J. McDowell (1998) p 79.
What a Noncognitivist might tell a Moral Realist
363
would of course be that such a necessary alignment is not restricted to the domain of the moral. Motivation, whatever the content, does not need alongside beliefs an additionally appetitive component as an independent state. Such a thesis is even harder to argue for, yet moral realism need not rely on such strong psychological assumptions. The weaker thesis that necessary alignment of belief and desire is confined to the moral is already sufficient. We might call this weaker thesis the necessitation assumption. The necessitation assumption says there are desires, the desires for the morally worthy (our “calm desires” as Hume has called them), that are necessarily present when having moral beliefs, and this indeed goes beyond a common even universal regularity or “constant conjunction” the non-cognitivist could be willing to acknowledge. What the realist is after here, is the modal force of such an alignment, a feature Humeans have no place for in their bipartite account. For a Humean there could very well be a world not much unlike ours where people share moral beliefs with us but do not at all feel motivated to act accordingly. A world of unmotivated bystanders, so to speak, creates no problem for the Humean. The moral realist of course has difficulties here, depending on how strong he wants to have his modal connection. If we state the necessitation assumption that way, then what the moral realist objects to is not the entire Humean theory as such. A moral realist could very well live with the non-cognitivist thesis that motivation necessarily involves the presence of a suitable desire-like state besides beliefs. He could very well live with assumption (a) and (b) of (HM). The realist’s objection is primarily directed towards the contingency assumption (c); for the moral realist it is not always possible, as the Humean assumes, to separate beliefs and desires modally. The kind of necessity, involved in the necessitation assumption, is of the anti-Humean de re type between distinct existences, which we see for example at work in recent anti-Humean theories of causation.1 There are beliefs and desires which necessarily stick together and these are, the moral realist tells us, desires that arise out of our beliefs concerning moral reality. Accordingly, for the realist a world of unmotivated bystanders cannot 1
See D.M. Armstrong (1983) and E. Fales (1990) for this type of anti-Humean necessity. There are several interesting questions that arise out of this parallel. Is the moral realist’s necessitation relation a sort of law-based necessitation – perhaps based on the laws of moral reality – which hold for us humans and perhaps other creatures who have a “perception of moral reality”? Or is it rather singularist, holding between this moral belief-state and that desire-state, irrespectively of what happens with other states of the same type, when confronted with moral reality? See D.M. Armstrong (1997) pp. 202-20 and M. Tooley (1992) for a singularist account of causality. We will not try to give an answer here, but one should at least bear in mind that the moral realist’s necessitation relation might allow quite different interpretations.
364
Arend Kulenkampff, Frank Siebelt
be as close to our actual world as the Humean assumes. If there is necessary alignment of moral beliefs and desires such a world would have to be rather remote from actuality. (Perhaps, there are a big differences then concerning such a world’s moral make up.) We could also say to believe that it is morally required to make it happen that 7 necessitates having a desire that 7. Thus stated the necessitation thesis doesn’t assume hybrid states that fulfill the functional roles of both beliefs and desires, whereby the necessary connection would be part of the constitution of such states. Although some realists think this conception of a special third type – “besires” as some have dubbed them – of states to be superior to the necessitation assumption on ontological grounds, their main argument doesn’t depend on such a move.1 As with many theses of necessary connections in rebus which are not of the conceptual-logical type this necessitation relation between distinct mental states is hard to swallow. Yet medicine might be hard to swallow and can nevertheless give us much needed cure. So we would at least require the moral realist assumption to be working in areas, where we already have good manageable theories of belief and desire. If the necessitation assumption can be upheld in these accounts that would give the moral realists’ position some independent support. Unfortunately this can’t be done, as David Lewis has recently shown.2 We will take a somewhat different exposition of Lewis’ argument, which in its original form is rather complex. One of the most basic theories of belief and desire is Bayesian expected value theory, which is built around the idea that a person desires a proposition to a degree equal to that proposition’s expected value. Within Bayesian theories, belief as well as desire admit of degree. More formally, within such theories, belief is represented by a probability function P, desire by a desirability 1
Although some philosophers present the necessitation assumption that way – see J. Mc Dowell (1978), D. Mc Naughton (1988) pp. 108-10, J. Dancy (1993) pp 22-7 it is easy to see that in fact nothing specific is won by such an ontological move. Let’s think of the content1 of such as “besire”. It can’t be a single unitary content which is at the same time both desired and believed. That would be contradictory with respect to functional roles. So we have to assume that a besire is always a logically compound state (5 & ), having a belief part 5 and a desire part . The Humean now takes it that those parts are only contingently related, the belief part and the desire part can very well be modally separated. Thus every combinatorial possibility could make up a real psychological state a person might possibly be in. The realist, of course, has to insist that the possibilities really are reduced, belief plus no desire doesn’t represent a genuine possibility. But then we are back to our original presentation of the moral realist position. We now again speak about belief states and desire states and of their possible combination. Talk of a third hybrid kind of state therefore proves to be only an extravagant ontological equivalent of what the moral realist is after.
2
See D. Lewis (1988) pp.326-28.
What a Noncognitivist might tell a Moral Realist
365
function V. Both P and V are defined over a Boolean algebra Z of propositions that is closed under the usual propositional operators and contains with T (the necessary true) and F (the necessary false) two maximal elements. (The necessary false proposition F is normally declined to have any desirability at all.) If we assume that desirabilities have an upper and a lower bound we can let both functions have their values in [0,1]. The axioms or rules for P are the usual rules of normalization, nonnegativity and additivity, whereas the desirability rule tells us that the desirability of a proposition is a weighted average of the desirabilities of the cases in which it is true, where the weights are proportional to the probabilities of each case. (See R. C. Jeffrey (1983) pp. 81 ) In addition to those four basic rules we also have a learning rule, which tells us how, as a result of gaining new evidence for the truth of some proposition 1, a person’s probability and desirability functions might change from P to P1 and from V to V1. This rule, conditionalization, tells us if a change originates in a proposition 1 with P(1) > 0, we should set the resulting new probability function as P1(5) = P(5/1) and the new desirability function as V1(5) = V(1 & 5) for all propositions 5 in Z.1 (P(5/1) abbreviates the quotient P(5 & 1)/P(1); it is defined whenever P(1) is positive.) All of the above mentioned so far is common ground. The question is, how can we incorporate the moral realist’s necessitation assumption into these theories? Remember, the moral realist’s assumption was to ensure that there are beliefs, our beliefs concerning the morally good, which eo ipso motivate by necessitating a corresponding desire. As was shown by Lewis2, we could make sense of the necessitation assumption within a Bayesian framework by making just two amendments. First, we enrich our algebra Z of propositions by letting Z contain, in addition to each proposition 1 already in Z, a new proposition 1 , the proposition that it would be morally good that 1. (Formally, we may think of a “halo”function assigning to each 1 in Z a new proposition 1 .) Second, we could say that for each “haloed” proposition 1 within a person’s algebra of propositions its degree of belief k = P(1 ) corresponds to that degree m = V(1) of desirability placed on the original proposition 1. In other words, a person must desire a proposition 1 to a degree that equals his strength of belief in the corresponding “haloed” proposition 1 , that it would be morally good that 1. 1
Part of why Lewis’s original proof in his (1988) is rather complex depends on his use of Jeffrey conditionalization. This wider rule permits belief change even when one’s aquiring new information does not originate in one’s coming to believe some proposition with certainty. A simpler proof can be found in his (1996) pp. 308-9, which is a sequel to his original paper.
2
See D. Lewis (1988) pp. 326-7.
366
Arend Kulenkampff, Frank Siebelt
That is to say, we should have as an additional rule in our framework the rule DAB “desire-as-belief”: (DAB) For each 1 in Z, V(1) = P(1 ) As was proved by H.A.Costa/J.Collins/I.Levy (1995), the result of making these amendments trivializes expected value theories. From conditionalization it follows that V(1 & 5) = V5(1), hence, by DAB, this equals P5 (1 ) = P(1 /5). So V(1 & 5) = P(1 /5) provided that P(1 & 5) > 0. Moreover, they could show that V(1 & 5) = V(1 & -5), whenever P(1 & 5), P(1 & -5) > 0 and P(5/1) < or > P(5), from which they could derive that V(1) = V(-1) for all propositions 1 whose probability is greater than 0.1 This result of course trivializes expected value theory because it implies that a person must view any proposition whose probability doesn’t equal 0 with indifference. It also means that even for a proposition whose believed moral goodness is high for a person (i.e. whenever P(1 ) comes close to 1) its desirability will equal the desirability of its negation. Yet, the believed morally good can’t leave us motivationally indifferent from the moral realists’ perspective. Therefore, the result of H.A. Costa, J. Collins and I. Levy not only trivializes expected value theory, it undermines the main rationale behind the moral realists’ necessitation assumption: that what is believed to be morally good can’t leave us unmoved or indifferent. Yet, we have to accept exactly this result, if we are to incorporate the moral realists’ necessitation assumption within standard expected value theories. Similar results were proved by Lewis (1988), (1996), and for a nonquantitative version of belief and desire theory by J. Collins (1988). These cast further doubts on whether there is a way of making sense of the realist’s necessitation assumption within our most undisputed theories of belief and desire. But maybe we were running too fast and have overlooked some aspect of antiHumeanism that could present us the alleged necessitation assumption in a different light. A proposal to endorse the moral realist position from a somewhat moderate perspective can be found in M. Smith’s account of motivation, though, as we shall see next, his position is afflicted with serious difficulties of its own.
1
See H.A.Costa/J.Collins/I.Levy (1995) p. 4.
What a Noncognitivist might tell a Moral Realist
367
IV M. Smith accepts the Humean theory of motivation because he thinks noncognitivism provides us with a good, although not universally applicable, model for explaining human action. However, he strongly opposes a Humean non-cognitivist axiology and opts instead for a realist and cognitivist perspective on moral values and properties. Like other moral realists, Smith assumes that value judgements, especially those with moral contents, have truth values. He goes along with the realist view that in addition to natural facts in our world there are also facts of value to discover, facts that can be represented truly or falsely.1 Furthermore, beliefs concerning what is morally required or what is valuable give us normative reasons for actions. At the center of Smith’s theory lies what he calls “an anti-Humean theory of normative reasons”. For him noncognitivism lacks a notion of value which is conceptually independent from that of desire and therefore non-cognitivism cannot, as he sees it, give us a correct account of the nature of normative reasons. Non-cognitivism can only provide us with an analysis of motivational reasons. Their difference can be shown as follows. To explain actions motivationally or intentionally means: to explain them by citing reasons which include beliefs and desires. In a weak sense every intelligible explanation of that sort provides a justification for the action in question. In contrast, in Smith’s view explanations that refer to normative reasons provide strong justifications for the actions under consideration. Explanations of that sort he calls deliberative. A deliberative action explanation includes valuing and believing. It therefore differs from an intentional motivational explanation by containing an axiological component, that is, it refers to values.2 Nevertheless, motivational and normative reasons have something in common. Both make actions intelligible.3 This could mean two things. It could mean explaining an agent’s acts motivationally by pointing to his or her reasons to act. Actor x did 5 because x desired that 1 and because x believed that 1 could only be achieved via doing 5. In many cases this amounts to nothing else than an (elliptical) causal-psychological explanation. It might be completed by adding a general psychological premise of the sort: x always starts trying to make it happen that 5, if x is motivated to make it happen that 5. On the other hand an agent’s act is intelligible for us in the light of his or her reasons to act.
1
M. Smith (1994) p. 4.
2
M. Smith (1994) p. 132.
3
M. Smith (1994) p. 95.
368
Arend Kulenkampff, Frank Siebelt
Its performance seems to be something which had to be done from the agent’s own perspective. Accordingly, with such a weak sense of reason every “rational act” can be considered as something which had to be performed. Now, Smith wants to understand normative reasons for an act to be performed, i.e. strong reasons, to be truths expressible via propositions of the form “That x makes it happen that 5, is desirable or required”.1 It is useful to our discussion to cite a principle here which was well known within traditional philosophy. The principle, which we find for example in Locke’s writing on actions, is that in consequentially closed actions no gap could appear between what is desired and what can rationally be justified.2 For if our acts were consequentially closed the best to be done in each case would be nothing else than what is also desired. The best would then be that which, when deciding to act, we are also motivated to do. There would be no wider perspective, no higher stand-point from which our decisions to act could be deliberatively criticized, because what is desired and what is required would not differ. Similarly and this is what Locke was aiming at, there could be no divergence between motivational and normative reasons, no difference between intentional and deliberative action explanations. It appears to be a simple task to imagine acts, which come very close to the Lockean ideal of consequential closure and whose performance if desired would be desired only because of themselves. Suppose someone takes a walk and sings a song. Why did she act that way? Because she felt like it without any further intentions or instrumental reasonings. She just enjoys singing. For her in her situation it was simply the best thing to do. There was nothing in favor of the alternative not to sing. Her act then could not only be motivationally explained by citing her desire to sing. It could also be normatively justified and completely so because the act was consequentially closed. In those circumstances to sing a song was simply what had to be done. Is it possible to interpret to be valuable as to be desirable in a reductionist sense? This seems to be the main issue within Smith’s theory. If there is indeed a genuine difference between valuing and desiring then the distinction between normative and motivational reasons and a fortiori between deliberative and intentional explanations could be accounted for. Smith argues by presenting paradigmatic cases. First, there is the case of desires we had better not want to have; desires we hate to perceive in ourselves and for which we feel shame and 1
M. Smith (1994) p. 95.
2
J. Locke once noted, “if every action of ours [were] concluded within it self, and drew no Consequences after it, we should undoubtedly never err in our choice of good; we should infallibly prefer the best.” Locke (1959) bk. II, ch. xxi, sec. 60.
What a Noncognitivist might tell a Moral Realist
369
sometimes even contempt of ourselves. In judging them, in considering their pros and cons, if we deliberate in Smith’s sense, we have to admit that they do not conform to our systems of values. As a classical example we can point to the addict, who desires to get rid of his addiction and yet feels an irresistible desire for his dose from time to time. He desires what he should rather not have desired and he knows it would be better for him if he didn’t so desire. We know Ovid’s famous phrase: Video meliora proboque deteriora sequor.1 Locke cites Ovid’s dictum in discussing the paradigmatic situation of an alcoholic.2 Does the gap in those examples between what is prima facie good and therefore desired and what is valuable and good on the whole according to rational deliberation show there is an argument for Smith’s moral realism and cognitivism about values? This seems to be the case only if it proves impossible to interpret those situations in non-cognitivist terms. But we can easily give an explanation that way. We need to make only two weak assumptions which conform to common sense psychology: (A)
Desires come in degrees whereby a stronger can override a weaker one.
(B)
Desires can be in conflict, which is the case when satisfying one excludes satisfying the other.3
With those assumptions (A) and (B) we can easily describe the situation of an alcoholic in Freudian terms: the powerless addict longing for cure experiences his situation as a permanent struggle between “Lustprinzip” and “Realitätsprinzip”. Because there are two desires, two souls in his breast longing for what is de facto irreconcilable, his situation is in fact an example of (B). It is hopeless, if the desire for his dose is always stronger than his desire for cure. And in describing his struggle in these terms we assume that his desire for his dose has the feature we assumed in (A). This type of explanation making use of two weak psychological assumptions appears not to be in conflict with a noncognitivist account. Why should it follow then, given the discrepancy between what is prima facie good and what is good only after deliberation, that the 1
Ovid (1994), liber VII, 20-1.
2
“It is not want of viewing the greater good: for he sees and acknowledges it, and, in the intervals of his drinking hours, will take resolutions to pursue the greater good; but when the uneasiness to miss his accustomed delight returns, the greater acknowledged good loses its hold, and the present uneasiness determines the will to the accustomed action (...)” Locke (1959), bk.II, ch. xxi, sec.35. See Smith (1994) pp. 134-36 for similar examples.
3
Another way to handle valuing in non-cognitivist’s terms might be found by pointing to second order desires as, for example, in Lewis’ (1889) account of valuing as desiring to desire.
370
Arend Kulenkampff, Frank Siebelt
former must be categorically distinct from the latter, an object of belief and not one of desire? Our Humean description of the dilemma of an addict signals no need for such a distinction. Quite the contrary: only if insight in what is valuable is an attitude of desire can a psychological conflict arise in the sense of (A) and (B). The deliberated good or right thing is what will be desired after a process of critical balancing and weighing, which includes a long term perspective, has taken place. And such a desire can very well be weaker or stronger in degree. In the above case of an addict it will not be strong enough to outweigh the desire for the prima facie good. Cognitivism on the other side generally has difficulties to interpret situations where a decision to act is made in circumstances of diverging interests or desires. Cognitivism thereby tends to unmask itself as a strenuous effort to describe the phenomenon of conflicting desires without referring to desires. For example J. Dancy writes: “(...) that the ability of a consideration to motivate can be affected by background conditions which are not themselves motivators. In some circumstances a consideration will be an incomplete reason in favor (perhaps overwhelmed), indifferent (perhaps silenced), or even something of a reason against.”1
Dancy cites the example of someone who feels a desire to enjoy ice cream and goes on to buy herself some. If she were aware of the fact, that ice cream is bad for her teeth, she would perhaps resist her desire. This little example shows that the vocabulary of cognitivism, “consideration”, “background condition”, etc. is simply misleading. Ice-cream has many properties which don’t affect us if we think about them. Taking into “consideration” that it is also bad for teeth doesn’t leave us unmoved. Why? Because we normally feel a strong desire to secure our health. And if we resist enjoying ice cream it is because the desire to avoid expected disadvantages to our health in the long run may sometimes give us a stronger motive than expecting a little pleasure now and then. An extreme example of valueless and disgraceful desire is given by the case of a mother, who overwhelmed by a fallow desire drowns her bawling child in the bath.2 She acts intentionally, driven by a sudden affect, but in a way she doesn’t value. Her act is not justifiable whatever perspective we would chose. Does this case prove something in favor of cognitivism? As far as we see it, nothing is proven. The phenomenon of abnormal or degenerative desires doesn’t show that to be valuable is something quite different from being desirable. Of course we could use cognitivist terms and say: the mother is lacking a 1
J. Dancy (1993) p.24.
2
See M. Smith (1994) p. 134.
What a Noncognitivist might tell a Moral Realist
371
perception of what is morally right and valuable. But we could very well and psychologically more to the point, describe the case in non-cognitivist terms and say: In her case, moral sentiments and desires like sympathy and benevolence which normally let us refrain from disdainful acts were too weak. If those desires to act, which naturally arise out of motherly instincts would have been stronger, her sudden affect would not have become her reason for an abhorrent deed. It is an easy task to describe what we have said in terms that make use of the psychological principles (A) and (B) without referring to values. Second, like when we sometimes desire what we shouldn’t desire, it also happens that our acts do not conform with our systems of values because we simply lack the corresponding desires. Many emotional and physical factors – M. Stocker speaks of “depressions” – could hinder us to desire what we should desire in light of our own principles: accidie, weakness of body, illness, general apathy or despair, inability to concentrate, a feeling of uselessness or futility.1 Even a malady, compared with others a minor one, like seasickness, can lead to apathy and indifference, because all desires to satisfy ones own principles to act die down. A hindering factor of first rank is without doubt fear. It is an emotional state which can demotivate a moral person, if she is not, like most of us, a hero. It can demotivate to such an extent that she refrains from acting according to her own conception of virtue. Fear can make the voice of conscience inaudible and can be a stronger motive than compassion or benevolence. For the amoral person who recognizes the existence of moral consideration, and yet remains unmoved, moral motivation tends have a degree close to zero.2 For the non-cognitivist who interprets to be of value as in some way as to be desirable, the existence of demotivating factors pose no special problem. Desires cannot only be stronger or weaker they can also be strengthened and weakened. Cases in which various depressions can sap desires altogether, in which under the influence of disturbing factors we may lack any motivating reason to do what we think we have a rational justification for doing can be interpreted in non-cognitivist terms without further problems. Under special circumstances we are motivationally weak and are not able to wish or desire what we should wish or desire according to our own moral principles. And this again is just all too human. In our view therefore Smith has not shown that valuing can’t be interpreted as desiring. His conclusion that valuing should instead be understood as believ-
1
See M. Stocker (1979) p. 746.
2
See D.O. Brink (1989) pp.46-50 for further discussion of amoralism in connection with moral realism.
372
Arend Kulenkampff, Frank Siebelt
ing1 does not follow from his arguments. The fact that we sometimes do not desire what we should desire, poses neither special problems for the noncognitivist nor does it pose problems for externalist moral realists. Externalists deny what internalists argue for, that there is a special connection between moral beliefs and desires, what we have described in section II as the moral realists’ necessitation assumption. Externalists oppose Smith’s practicality requirement. Beliefs, whatever their content – for short: reason – can’t immediately and out of its own be practical. Non-cognitivists accept the practicality requirement, and argue in the following way: because, in Hume’s words, “morals excite passions and produce and prevent actions”2 and can be immediately practical or intrinsically motivational (provided that they are not disturbed by other affects) it can’t be the work of reason which is motivationally quite powerless. Moral opinions therefore must be expressions of desire. Moral externalism and non-cognitivisms therefore share the view that belief and desire should be understood as conceptually and modally independent mental states. And because of their modal independence, there is no strong modal truth involved in saying “that agents who judge it right to act in various ways are so motivated”3. Non-cognitivists can even accept that there is moral truth without accepting the moral realists’ independent moral facts. One can, for example, accept the golden rule, that we should not treat others in a way that we don’t want to be treated ourselves as an elementary moral truth. Yet, without being accompanied by an existing desire to act accordingly, this moral belief seems to be without a truth-maker. This belief then motivates just as little as Hutchenson’s truth “ (...) ‘an hundred stones are greater than one’ will excite a man, who has no desire of heaps, to cast them together.”4 For internalists who uphold the practicality assumption the fact that what is known to be right sometimes leaves us cold poses a special problem. The crux is Smith’s modal connection between moral judgement and the will. If reasoning and will, belief and desires, the appetitive and the cognitive are distinct existences in the sense we have outlined in section I there could be one without the other, insight in what is morally right without wishing to pursue it. Then we could more easily understand why it is indeed possible to have ineffectual knowledge without a desire to act of what is morally required.
1
M. Smith (1994) pp. 147-51.
2
D. Hume (1985) bk.III, part I, sec.i.
3
M. Smith (1994) p. 61.
4
F. Hutchenson (1971) p. 223.
What a Noncognitivist might tell a Moral Realist
373
Otherwise, if the necessitation assumption is correct, and it is in Smith’s sense a conceptual truth that agents who make moral judgments are motivated accordingly, it is hard to understand how there could be disturbing factors leaving the belief intact but destroying the desire. To emphasize our last point: At stake is therefore the very coherence of the idea “that deliberation on the basis of our values is practical in its issue to just the extent it is.”1 Indeed, we see a problem with the coherence of that idea. There is Smith’s statement that “our beliefs about normative reasons are necessarily connected with our desires to just the extent they are.”2 Imagine somebody talking about a sound logical system S saying about one of its basic transformation rules RS “RS is validity preserving to just the extent it is.” This could mean two things. Either the clause “to just the extent it is” only verbally refers to exceptions. Then the whole statement just says: RS is validity preserving, period. Or the clause does indeed refer to exceptions, then it says: except in those cases in which RS is not validity preserving. But then it was simply wrong in the first place to say that RS is a sound logical system. Basic inference-rules in a sound logical system can’t be almost validity preserving or validity preserving with some tiny little exceptions. Similarly, if our normative reasons are only ceteris paribus necessarily connected with our desires, to just the extent that they are, it seems to follows that the alignment just can’t be necessary in the first place: it must be contingent. If there is a necessary link between properties, like for example between being an electron and being negatively charged then there are simply no circumstances, no possible worlds in which a thing is an electron albeit positively charged. Smith’s thesis appears to be a reductio ad absurdum of internalist moral realism.
V One landmark of the theoretical battlefield we have not yet encountered is the moral realist’s ontology of special moral properties and facts. And for good reason, because this, in many respects, is the most difficult question in weighing the success of moral realists’ and non-cognitivists’ accounts of motivation. We don’t want to close our arguments by directly adressing the idea of furnishing our world with moral facts and moral properties. We would rather point to
1
M. Smith (1994) p.137.
2
M. Smith (1994) p.137.
374
Arend Kulenkampff, Frank Siebelt
a difficulty that arises out of that idea in the context of the realist’s construal of moral motivation. For the non-cognitivist, moral motivation arises from beliefs and desires, whereas moral motivation arises from beliefs alone for the moral realist. The necessitation assumption has been seen to be an attempt to address the question of how this might be posssible. We have already discussed some implications of that assumption, but we have not yet looked into the nature of those motivating beliefs within the realist framework. One central aspect of moral realism is the idea that there is a whole range of moral properties and facts independent of our beliefs concerning them. And, of course, as with other types of beliefs there lies the possibility of error in what we regard as a moral fact. At this point we don’t want to confront the realist with a sceptical argument “from queerness” concerning knowledge as applied to moral reality. Yet, we think there is a massive problem for the realist concerning moral properties and facts as to how he can discriminate true moral belief and hence knowledge from false beliefs or lucky guesses. Today many philosophers (who don’t see fighting Cartesian demons as the primary task in giving an account of knowledge) think of discriminating empirical knowledge from mere lucky true beliefs in terms of dependency relations. For a true belief to become knowledge there must be some kind of dependency between having exactly this belief and whatever it is in reality that gives the belief its truth-status. For a belief that 1 to count as knowledge, it must not only be true that 1, but also to have just this belief that 1 should be dependent on what makes 1 true, so to speak. You can’t know that there is a sheep standing on the field if this belief of yours is not dependent on the sheep actually standing there, say, when you haven’t noticed the sheep but a hairy dachshund in front of you without which you wouldn’t have come by your belief. In cases like this, the right dependency is missing. Whether or not this idea could be turned into a sufficiently broad account of what marks empirical knowledge from mere true belief can of course be questioned. But that some kind of true dependency hypotheses should enter as a necessary condition in any theory of empirical knowledge seems to be undisputable.1 Yet if this type of dependency is part of what turns true belief into knowledge, it is far from clear how the moral realist can make sense of this idea. Since Moore and Hare2 it is a well known peculiar feature of moral facts and properties that they are supervenient on natural facts and properties. If a person’s 1
See among others A. Goldman (1976), F.Dretske/B.Enc (1984), David Lewis (1980). See also what Lewis recently has called “the rule of reliability” in his (1996).
2
See especially G.E. Moore (1922) pp.260-62 and R.M.Hare (1952) pp.144-46. For a recent assessment of the concept of supervenience in other areas see also J. Kim (1993).
What a Noncognitivist might tell a Moral Realist
375
deed was morally valuable or good, having this property is supervenient on the natural properties of the situation in at least the following way: you can’t have another person acting in the same way as the first in a similar situation, where we have exact similarties with respect to natural properties, while only differing from the first in that this time the deed was rather not good. If something x has the moral property P and natural properties Q, R, S, ... then anything y having the same natural properties Q, R, S, ..., as the first will also have P. Contraposing the thought, if we have some x and y differing with respect to the moral property P, then x and y must also differ with respect to non-moral properties Q, R, S,... . And of course, if anything x changes with respect to the moral property P, for example by losing or aquiring it, then x must also have changed in respect to some natural properties Q, R, S, ... . Yet, if this type of supervenience holds for moral properties and, hence for moral facts, it is hard to see how the realist can apply to a dependency hypothesis when delineating moral knowledge from mere luckily true moral beliefs. Suppose Oscar, a man of moral perception, witnessing something nurse Mary did, thereby coming to believe that, what Mary did was morally good. Let’s suppose that by all standards it in fact was something we would count as morally valuable. To attribute to him knowledge of a moral fact, i.e., that what Mary did had some moral property, we have said, a dependency hypothesis should be true in these circumstances. But if moral properties are supervenient properties, it seems, that there is no work left for the genuine moral aspect of reality in backing knowledge claims. Instead of pointing to the moral properties of the situation, we can equally well point to its natural properties, and say if nurse Mary had not done what she did, then Oscar wouldn’t have believed what he actually did believe: that what Mary did hat some moral property. And of course we could describe what Mary did solely by pointing to non-moral features of her deed in those circumstances. But then the moral realists’ moral properties and facts turn out to be merely epiphenomenal bystanders in examining claims of moral knowledge, they just do not show up in assessing relevant dependency claims. We think that what happens with moral knowledge and beliefs within the realist framework will also happen in other areas of the battle between moral realists and non cognitivists. For the realists’ construal of moral motivation the consequences of the argument for the epiphenomenalist feature of moral properties and facts is rather bleak. For we could in the same way ask in assessing motivation claims, what features of a situation could have motivated a person the way she was motivated. It seems that in the same way the alleged moral properties and facts of the realist are left out as notoriously unemployed
376
Arend Kulenkampff, Frank Siebelt
epiphenomens in assessing motivation claims. But this is only a hint of an argument which needs a more detailed and seperate exposition.
References D.M. Armstrong (1983), What is a Law of Nature?, Cambridge 1983 ——— (1997), A World of States of Affairs, Cambridge 1997 D.O. Brink (1984), Moral Realism and the Sceptical Arguments from Disagreement and Queerness, in: Australasian Journal of Philosophy 62, 111-25 ——— (1989), Moral Realism and the Foundations of Ethics, Cambridge 1989 H.A.Costa/J.Collins/I.Levy (1995), Desire-as-belief implies Opinionation or Indifference, in: Analysis 55, 1-5 J. Collins (1988), Belief, Desire and Revision, in: Mind 97, 333-42 J. Dancy (1993), Moral Reasons, Oxford 1993 D. Davidson (1963), Actions, Reasons and Causes, in: Journal of Philosophy 69, 685-700 F. Dretske/B.Enc (1984), Causal Theories of Knowledge, in: Midwest Studies in Philosophy 9, 51728 A. Goldman (1976), Discrimination and Perceptual Knowledge, in: Journal of Philosophy 73, 77191 L. Humberstone (1992), Direction of Fit, in: Mind 101, 59-83 D. Hume (1985), A Treatease on Human Nature, Oxford 1985 F. Hutchenson (1971), Illustrations on the Moral Sense,in: F. Hutchenson, Collected Works, Vol. II, Hildesheim 1971 F.Jackson/P.Pettit (1995), Moral Functionalism and Moral Motivation, in: The Philosophical Quarterly 45, 451-72 R.C. Jeffrey (1983), The Logic of Decision, Second Edition, Chicago und London 1983 D. Lewis (1980), Veridical Hallucination and Prosthetic Vision, in: D. Lewis, Philosophical Papers, Vol II, Oxford 1986, 273-90 ——— (1996), Elusive Knowledge, in: Australasian Journal of Philosophy 74, 549-67 ——— (1989), Dispositional Theories of Value, in: Proceedings of the Aristotelian Society, Supp.Vol. 63, 113-37 ——— (1988), Desire as Belief, in: Mind 97, 323-32 ——— (1996), Desire as Belief II, in: Mind 105, 303-13 B. Loar (1981), Mind and Meaning, Cambridge 1981 J. Locke (1959), An Essay Concerning Human Understanding, New York 1959 J. McDowell (1998), Are Moral Requirements Hypothetical Imperatives? in: J. McDowell (1998), Mind, Value, and Reality, Cambridge 1998, 77-95 D.A. McNaughton (1988), Moral Vision, Oxford 1988 Ovid (1994), Metamorphoseon Libri Qindecim, Stuttgart 1994 M. Platts (1979), Ways of Meaning, London 1979 ——— (1980), Moral Reality and the End of Desire, in: (Ed.) M. Platts, Reference, Truth and Reality, London 1980 W.V.O.Quine (1981), On the Nature of Moral Values, in: W.V.O. Quine, Theories and Things, Cambridge Mass 1981, 54-67 M. Smith (1992), Valuing: Desiring or Believing, In: (Ed.) D.Charles/K.Lennon, Reduction, Explanation, and Realism, Oxford 1992
What a Noncognitivist might tell a Moral Realist
377
——— (1994), The Moral Problem, Oxford 1994 ——— (1987), The Humean Theory of Motivation, in: Mind 96, 36-61 ——— /P.Pettit (1990) Backgrounding Desires, in: The Philosophical Review 99, 565-92 M. Stocker (1979), Desiring the Bad: An Essay in Moral Psychology,in: Journal of Philosophy 76, 738-53 J.J.Thomson/G.Harman (1996), Moral Relativism and Moral Objectivity, Cambridge Mass 1996 M. Tooley (1992), The Nature of Causation: A Singularist Account, in: Canadian Journal of Philosophy, Supp.Vol 16, 271-322
378
Contributors
CONTRIBUTORS Thomas Baldwin Prof., Philosophy, York University, York, Great Britain, email:
[email protected] Raffaella De Rosa Prof., Rutgers University, Philosophy Department, New Brunswick NJ 08903, USA, email:
[email protected] Jay L. Garfield Prof., The University of Tasmania, Smith College and School of Philosophy, Faculty of Arts, G.P.O. Box 252-41, Hobart, Tasmania 7001, Australia, email:
[email protected],
[email protected] Alvin I. Goldman Prof., Dept. of Philosophy, University of Arizona, Tucson, AZ 85721 USA, email:
[email protected] Philip Pettit Prof., The Australien National University, Research School of Social Science, GPO Box 4, Canberra ACT 2601, Australia, email:
[email protected] Robert Gordon Prof., University of Missouri, Philosophy Dept., St. Louis MO 63121, USA, email:
[email protected] title Jane Heal Prof., St John's College, Cambridge CB2 1TP, Great Britain Christopher S. Hill, Department of Philosophy, University of Arkansas, Fayetteville, AR 72701, USA, email:
[email protected] Rebecca Kukla Prof., Department of Philosophy, Carleton University, 1125 Colonel By Ottawa, ON, Canada K1S 5B6, email:
[email protected] Arend Kulenkampff Prof., J.W. Goethe-Universtät, Institut für Philosophy, Frankfurt am Main, 60054 Frankfurt am Main, RFA. Brian P. McLaughlin Prof., Rutgers University, Philosophy Department, New Brunswick NJ 08903, USA, email:
[email protected] David Pitt Prof., Department of Philosophy Department of Philosophy and Religious Studies, Iowa State University, Ames, IA USA, email:
[email protected] Jeffrey S. Poland Prof., Department of Philosophy, University of Nebraska-Lincoln, Lincoln, NE 68588-0321, 402-472-2891, email:
[email protected] Gerhard Preyer Dr. habil. Privatdozent, J.W. Goethe-Universtät, Fachbereich: Gesellschaftswissenschaften, Frankfurt am Main, 60054 Frankfurt am Main,
Contributors
379
RFA, email:
[email protected] Louise Röska-Hardy Dr., J.W. Goethe-University, Institut für Philosophie, FB 8, 60054 Frankfurt am Main, RFA, email:
[email protected] David M. Rosenthal Prof., Program in Philosophy Coordinator, Cognitive Science Concentration Graduate School, City University of New York, 365 Fifth Avenue, New York, NY, USA email:
[email protected] Erwin Rogler erimit. Dr., J.W. Goethe-Universität, Institut für Philosophie, FB 8, 60054 Frankfurt am Main, RFA Frank Siebel Dr., Schweizerstr. 55, 60594 Frankfurt am Main, RFA. James Tomberlin, Prof., Dept. of Philosophy, Californiea State University at Northridge, Northridge, CA 91300, USA, email:
[email protected] Barbara Von Eckardt Prof., Department of Philosophy, University of Nebraska-Lincoln, Lincoln, NE 68588-0321, 402-472-2891, email: b_v@unlinfo. unl.edu
380
Imprint
IMPRINT PROTO SOCIOLOGY An International Journal of Interdisciplinary Research ISSN 1434-4319 Editor: Gerhard Preyer Johann Wolfgang Goethe-University Frankfurt am Main, Department of Social Sciences Editorial Staff: Georg Peter Editorial of the Vol. 14 1999: Georg Peter Project: Frank Siebelt, Mathias Boes Layout und Druckvorstufe: Georg Peter Editorial Office: PROTOSOCIOLOGY, Stephan-Heise-Str. 56, 60488 Frankfurt am Main, RFA, Phone: 069-769461, Email:
[email protected],
[email protected] Bank: Dresdner Bank AG, Frankfurt am Main, Kto: 44 121 168 01, BLZ: 500 800 00 Wir danken debis, concept and products für ihre freundliche Unterstützung.
Die Zeitschrift soll 1/2jährlich erscheinen. Die Anzahl der jährlich erscheinenden Hefte und Sonderhefte bleibt jedoch vorbehalten. Copyright: Die in dieser Zeitschrift veröffentlichten Beiträge sind urheberrechtlich geschützt. Alle Rechte sind vorbehalten. Übersetzungen, Nachdruck, Vervielfältigung auf fotomechanischem oder ähnlichem Weg oder im Magnettonverfahren, Wiedergabe durch Vortrag, Funk- und Fernsehsendungen sowie Speicherung in Datenverarbeitungsanlagen, auch auszugsweise, sind nur mit Genehmigung des Herausgebers möglich. Für eingereichte Beiträge wird keine Haftung übernommen. Weitere Publikationsrechte von Artikeln bleiben vorbehalten. Zweitpublikationen seitens des Autors werden eingeräumt. Bei einer Zweitpublikation ist das Heft (Nummer, Titel, Erscheinungsjahr) der PROTOSOCIOLOGY zu zitieren. Für unaufgefordert eingesandte Manuskripte wird keine Haftung übernommen. Gerichtsstand ist Frankfurt am Main. Copyright: All rights reserved. This publication may not be reproduced, stored or transmitted in any form or by any means without the prior permission in writing of the publisher. Additional publications of the articles are reserved. The authors retain the personal right to re-use their own articles. Authorization to photocopy items for internal or personal use, or the internal or personal use of specific clients is garanted by PROTOSOCIOLOGY, provided that the base fee is paid directly to VG Wort, Goethestr. 49, 80336 München RFA.The publisher accepts no responsibility for submitted manuscripts.
ProtoSociology im Internet: http://www.protosociology.de http://www.rz.uni-frankfurt.de/protosociology
381
On ProtoSociology
ON PROTOSOCIOLOGY
“Protosociology occupies an important position in the European intellectual scene, bridging philosophy, economics, sociology and related disciplines. Its volumes on rationality bring together concerns in all these topics, and present an important challenge to the cognitive sciences.” Donald Davidson, Berkeley (USA)
“Protosociology publishes original papers of great interest that deal with fundamental issues in the human and social science. No academic library is complete without it.” Nicholas Rescher, Pittsburgh (USA)
“Protosociology has been remarkably successful in publishing interesting work from different tradition and different disciplines and, as the title signals, in giving that work a new, eye-catching slant.” Philipp Pettit, Canberra, Australia
“Protosociology is a truly premier interdisciplinary journal that publishes articles and reviews on timely topics written by and for a wide range of international scholars. The recent volumes on rationality are remarkable for their breadth and depth. Protosociology would be a great addition to any library.” Roger Gibson, St. Louis (USA)
382
Volumes
PUBLISHED VOLUMES
PROTOSOCIOLOGY
An International Journal of Interdisciplinary Research Vol. 13 1999: Reasoning and Argumentation D. Mans, G. Preyer (Hrsg.) CONTENTS Ralph H. Johnson Reasoning, Argumentation and The Network Problem Leo Groarke The Fox and the Hedgehog: On Logic, Argument, and Argumentation Theory
Manfred Kienpointner Comments on Douglas Walton’s Paper Christopher W. Tindale The Authority of Testimony John Woods Peirce’s Abductive Enthusiasms
J. Anthony Blair Presumptive Reasoning/Argument: An Overlooked Class
Henry W. Johnstone, Jr. “‘Any,’ ‘Every,’ and the Philosophical Argumentum ad Hominem“
Robert C. Pinto Argument Schemes and the Evaluation of Presumptive Reasoning: some Reflections on Blair’s Account
ON CONTEMPORARY PHILOSOPHY
Douglas Walton The New Dialectic: A Method of Evaluating an Argument Used for Some Purpose in a Given Case
Hans Lenk Interdisziplinarität und Interpretation Ellery Eells Causal Decision Theory
http://www.protosociology.de
383
Volumes
PROTOSOCIOLOGY An International Journal of Interdisciplinary Research
VOL. 12, 1998 – Special Edition AFTER THE RECEIVED VIEW – Developments in the Theory of Science Gerhard Preyer, Georg Peter, Alexander Ulfig (Eds.) in memoriam Wolfgang Stegmüller Content Introduction: Developments in the Theory of Science Gerhard Preyer, Georg Peter, Alexander Ulfig LOGICAL OPERATIONALISM – SIGNIFICANCE AND MEANING Wilhelm K. Essler Truth and Knowledge. Some Considerations concerning the Task of Philosophy of Science Gerhard Preyer The Received View, Incommensurability and Comparison of Theories – Beliefs as the Basis of Theorizing Robert Schwartz Reflections on Projection Jeffrey E. Foss The Logical and Sociological Structure of Science STRUCTURALISM – MEANINGFUL MEASUREMENT – THE CONCEPTION OF PHYSICAL LAW C. Ulises Moulines Structuralism vs. Operationalism Nicholas Rescher Meaningless Numbers
384
Volumes
R. I. G. Hughes Laws of Nature, Laws of Physics, and the Representational Account of Theories James R. Brown Einstein’s Principle Theory INDUCTIVE INFERENCES – INTERPRETATION OF PROBABILITY – GAME THEORY Kevin T. Kelly, Cory Juhl Transcendental Deductions and Universal Architectures for Inductive Inferences Howard H. Harriott R.A. Fisher and the Interpretation of Probability Brian Skyrms Evolution of an Anomaly PROPERTIES – UNDERDETERMINATION – SCIENTIFIC REALISM George N. Schlesinger Degrees of Characterizations Carl A. Matheson Observational Adequacy as distinct from the Truth about Observables Thomas R. Grimes Scientific Realism and the Problem of Underdetermination Paul C. L. Tang On Paul Churchland’s Treatment of the Argument from Introspection and Scientific Realism RATIONALITY – METAPHORS – VALUES IN SCIENCE David Resnik Scientific Rationality and Epistemic Goals Aldo Montesano Rationality in Economics: A General Framework Joseph Agassi Science Real and Ideal: Popper and the Dogmatic Scientist Michael Bradie Models and Metaphors in Science David Gruender Values and the Philosophy of Science
385
Volumes
PROTOSOCIOLOGY An International Journal of Interdisciplinary Research VOL. 11, 1998 COGNITIVE SEMANTICS II – EXTERNALISM IN DEBATE Content RADICAL INTERPRETATION, ONTOLOGY AND THE STRUCTURE OF LANGUAGE Gerhard Preyer, Michael Roth On Donald Davidson’s Philosophy: An Outline Richard Manning All Facts Great and Small Barbara Fultner Of Parts and Wholes: The Molecularist Critique of Semantic Holism Louis Goble Re-Evaluating Supervaluations
David Simpson Interpretation and Skill: On Passing Theory Wulf Kellerwessel Katz on Semantics and Pragmatics EXTERNALISM AND THE INDIVIDUATION OF CONTENT Ron Wilburn Knowledge, Content, and the Wellstrings of Objectivity Anthony Brueckner Content Externalism and A Priori Knowledge Consuelo Preti The Irrelevance of Supervenience Michael Liston Externalist Determinants of Reference Arnold Silverberg Semantic Externalism. A Response to Chomsky Gerhard Preyer Interpretation and Rationality: Steps from Radical Interpretation to the Externalism of Triangulation
386
Volumes
PROTOSOCIOLOGY An International Journal of Interdisciplinary Research VOL. 10, 1997 COGNITIVE SEMANTICS I – CONCEPTIONS OF MEANING Content MEANING, TRUTH AND THE USE OF LANGUAGE Herman Cappelen, Ernie LePore Semantic Theory and Indirect Speech
Jeffrey King The Source(s) of Necessity
Kirk Ludwig The Truth about Moods
Filip Buekens The Genesis of Meaning (a Myth)
Louise Röska-Hardy Language, Use and Action REFERENCE, INDEXICALS AND SPEAKER MEANING Robert Hanna Extending Direct Reference Peter Ludlow Semantics, Tense, and Time: a Note on Tenseless Truth-Conditions for TokenReflexive Tensed Sentences
Gerhard Preyer Verstehen, Referenz, Wahrheit. Zur Philosophie Hilary Putnams Reinaldo Elugardo Descriptions, Indexicals and Speaker Meaning
RATIONAL DIALETISM OF ANTINOMIES Mark Sainsbury Can Rational Dialetheism Be Refuted By Considerations about Negation and Denial? CONTEMPORARY DEBATES IN THE PHILOSOPHY OF MIND Klaus Sachs-Hombach Philosophy of Mind: Die Simulationstheorie IN RETROSPECT Joseph Agassi Wittgenstein – The End of a Myth
Kent A. Peacock, Richard Feist The Einstein-DeSitter Controversy
387
Volumes
Forthcoming
PROTOSOCIOLOGY An International Journal of Interdisciplinary Research Vol. 15, 2001 On a Sociology of Borderlines CONTENTS Introduction: Social Process in Time of Globalization I Reconceptionalizations of the Global: Borderlines in World Society S. N. Eisenstadt, The Continual Reconstruction of Multiple Modern Civilizations and Collective Identities; Christopher Chase-Dunn, Globalization: A World-Systems Perspective; Thomas D. Hall, World-Systems, Frontiers, and Ethnogenesis: Incorporation and Resistance to State Expansion; Richard E. Lee, After History? The Last Frontier of Historical Capitalism II Defining Borderlines in World Society: The Emergence of New Memberships Gerhard Preyer, Globalization and the Evolution of Membership; Barrie Axford, Enacting Globalization: Transnational Networks and the Deterritorialization of Social Relationships in the Global System; Mathias Bös, Immigration and the Open Society: The Normative Patterns of Membership in the nation State; Uta Gerhardt, Birgitta Hohenester A Transformation of National Identity? Refugees and German Society After World War II; Marja Keränen, Citizenship, Universal-Speak, and Local-Speak III The Global and the Local: The Collapse and Reconstruction of Borderlines Christie Davies, Eugene Trivizas, The Collapse of the National Morality and National Moral Boundaries of Small Peripheral Countries: not Globalisation but the Imposition of Liberty; Walter L. Bühl, Former GDR between “Transformation” and “Social Evolution; Peter Wagner, Beyond “East” and “West”: On the European and Global Dimensions of the Fall of Communism; Ramón Grosfoguel, ‘Cultural Racism’ and ‘Borders of Exclusion’ in the Capitalist World-Economy: Colonial Caribbean Migrants in Core Zones; Francisco Entrena, Socio-Economic Restructurings of the Local Settings in the Era of Globalization. On a Sociology of Violence Konrad Thomas, Ein anderes Verständnis von Gewalt. Der gesellschaftsanalytische Beitrag des Literaturwissenschaftlers René Girard
388
PROTOSOCIOLOGY An International Journal of Interdisciplinary Research PARSONS-SPECIAL EDITION CLASSICAL SOURCES AND CURRENT USES OF TALCOTT PARSONS’S THOUGHT edited and with an introduction by Bernard Barber and Uta Gerhard This volume wishes to cut across the vast array of misunderstandings, misreadings, and misinterpretations which have marred the proper recognition of Parsons’s work for the last thirty years. If the question is asked why such recognition would be important, the answer is twofold. For one, Parsons’s work is a classic that reaches from the l930s into the l970s of the twentieth century while sociology’s now arguably uncontroversial classics - like Max Weber, or Emile Durkheim - were only able to analyze society until about the end of World War I. Therefore, Parsons’s oeuvre has a special modernity in that his work takes into account the New Deal, National Socialism, McCarthyism, and the Student Rebellion, to name but a few timely events which are part and parcel of Parsons’s sociological analysis. Furthermore, his conception of social order and anomie, especially since they seriously discuss the problematic viability of utilitarian economics, are of particular importance today. Introductionary Essay: The Parsons Agenda Bernard Barber, Uta Gerhardt PART 1: Revisiting The Structure of Social Action Mark Gould: Neoclassical Economics, Social Order, and “The Structure of Social Action” Uta Gerhardt: National Socialism and the Politics of “The Structure of Social Action” Richard Münch: The Problem of Order: Sixty Years After “The Structure of Social Action” PART 2: Understanding Parsons’s Thought Bryan S. Turner: Sacricity and Solidarity in Talcott Parsons’s Economic Sociology Edward A.Tiryakian: An Emergent French Connection: Revisiting Parsons’s Durkheim Bernard Barber: The Nature and Functions of Religion and Beyond PART III: Evaluating the Scope of Sociology Daniel Bell: The Limits of General Theory in the Social Sciences Hans Jürgen Puhle: (Title to be announced)
389
Bookpublications
Bookpublications:
LOGICAL FORM AND LANGUAGE Gerhard Preyer, Georg Peter (eds.) CONTENTS Introduction
Stephen Neale Logically Unformed
Norbert Hornstein A Grammatical Argument for A Davidsonian Semantics
Ernie Lepore, Kirk Ludwig What is Logical Form
Jason Stanley Nominal Restriction
Paul M. Pietroski Function and Concatenation
PART III: LOGICAL FORM, BELIEFS ASCRIPTION AND PROPER NAMES
Jeffrey King Two Sorts of Claim about "Logical Form"
Bernard Linsky Russell's Logical Form, LF and Truth Conditions
PART I: THE NATURE OF LOGICALFORM
Peter Ludlow LF and Natural Logic. The Syntax of Directional Entailing Environments Robert May/R.F. Fiengo Identity Statements PART II: INTENSIONALITY, EVENTS AND SEMANTIC CONTENT James Higginbotham Why is Sequence of Tens Obligatory? Richard Larson The Grammar of Intensionality Barry Schein Events and the Semantic Content of Thematic Relations
Gerhard Preyer On the Legacy of Russell's Theory of Definite Description Lenny Clapp, Robert Stainton “Obviously Propositions are Nothing.” Russell and the Logical Form of Belief Reports Robert Matthews Logical Form and the Relational Conception of Belief Marga Reimer Ordinary Proper Names
Oxford University Press 2001 For further Information see: http://www.rz.uni-frankfurt.de/protosociology
390
Bookpublications
The Contextualization of Rationality Gerhard Preyer, Georg Peter (eds.) Introduction: Problems, Concepts and Theories of Rationality RADICAL INTERPRETATION, NATURALIZED EPISTEMOLOGY AND NORMATIVITY John Heil The Propositional Attitudes
Roger F. Gibson Stich on Intentionality and Rationality
David K. Henderson Epistemic Rationality, Epistemic Motivation, and Interpretive Charity
Paul K. Moser and David Yandell Against Naturalizing Rationality
Gerhard Preyer Interpretation and Rationality: Steps from Radical Interpretation to the Externalism of Triangulation
Harvey Siegel Naturalism, Instrumental Rationality, and the Normativity of Epistemology
INTENTIONS AND THE SOCIAL ASPECT OF RATIONALITY Alfred R. Mele Rational Intentions and the Toxin Puzzle
Raimo Tuomela Rational Cooperation and Collective Goals
Peter Gärdenfors The Social Stance
Peter French Rationality and Ethics
Timo Airaksinen and Katri Kaalikoski Instrumental Rationality
Julian Nida-Rümelin The Plurality of Good Reasons and the Theory of Practical Rationality
CONCEPTS OF EXPLANATION, JUSTIFICATION AND REALITY Philip Pettit Three Aspects of Rational Explanation
Alexander Ulfig Validity, Justification and Rationality
Keith Lehrer Rationality and Trustworthiness
Nicholas Rescher Reason and Reality 266 Seiten, DM 88.-
Der Band erscheint in der Reihe:
“Perspektiven der Analytischen Philosophie” mentis Verlag, Paderborn
391
Bookpublications
PROTOSOZIOLOGIE IM KONTEXT “Lebenswelt” und “System” in Philosophie und Soziologie Gerhard Preyer, Georg Peter, Alexander Ulfig (Hrsg.) in memoriam Paul Lorenzen Einleitung: “Lebenswelt” und “System” in Philosophie und Soziologie, Gerhard Preyer, Georg Peter, Alexander Ulfig ZUM BEGRIFF DER LEBENSWELT Ernst W. Orth: Lebenswelt als unvermeidliche Illusion? Husserls Lebensweltbegriff und seine kulturpolitischen Weiterungen, Walter Biemel: Gedanken zur Genesis der Lebenswelt, Alexander Ulfig: Lebenswelt und Reflexion. Anhang: Lebenswelt als Fundament der Wissenschaft, Gerhard Preyer: Hintergrundwissen: Kritik eines Begriffs, Hubert A. Knoblauch: Soziologie als strenge Wissenschaft? Phänomenologie, kommunikative Lebenswelt und soziologische Methodologie
LEBENSWELT – BEGRÜNDUNG – WISSENSCHAFT Jürgen Mittelstraß: Das lebensweltliche Apriori, Peter Janich: Die Rationalität der Naturwissenschaften, Jürgen Mittelstraß: Rationalität und Reproduzierbarkeit, Elisabeth Ströker:Lebenswelt durch Wissenschaft, Paul Janssen: Lebenswelt, Wissen und Wissenschaft, Richard T. Murphy: E. Husserl's Phenomenology of Reason LEBENSWELT/LEBENSFORM – SPRACHE Pierre Kerszberg: Lifeworld and Language, John F.M. Hunter: The Motley Forms of Life in the Later Wittgenstein, Peter A. French: Why did Wittgenstein read Tagore to the Vienna Circle? Georg Peter: Die Nebenbeschäftigung der Symbole. Zu Wahrheit und Funktion der Metapher
SYSTEM – SOZIALSYSTEM – GESELLSCHAFT Niklas Luhmann: Die Lebenswelt nach Rücksprache mit Phänomenologen, Niklas Luhmann: Observing Re-entries, Gerhard Preyer: System-, Medien- und Evolutionstheorie. Zu Niklas Luhmanns Ansatz, Richard Münch: Autopoesis per Definition, Hans Zitko: Codierungen der Kunst: Zur Kunstsoziologie Niklas Luhmanns, James Bohman: The Completeness of Macro-Sociological Explanations: System and Lifeworld, Göran Ahrne: Outline of an Organisational Theory of Society, Anhang: Karl Otto Hondrich: Zu Göran Ahrnes Ansatz 392 Seiten, 68,- DM Verlag Königshausen & Neumann, Würzburg 1996 Digitale Neuauflage bei: Humanities-Online, Frankfurt am Main http://www.humanities-online.de
392
Bookpublications
STRUKTURELLE EVOLUTION UND DAS WELTSYSTEM Theorien, Sozialstruktur und evolutionäre Entwicklungen Gerhard Preyer (Hrsg.) Einleitung: Gerhard Preyer: Strukturelle Evolution und das Weltsystem: Theorien, Sozialstruktur und evolutionäre Entwicklungen ZENTUM UND PERIPHERIE – INSTITUTIONELLE ENTWICKLUNG – ASKRIPTIVE SOLIDARITÄT Shmuel Noah Eisenstadt: Social Division of Labor, Construction of Centers and Institutional Dynamics: A Reassessment of the Structural–Evolutionary Perspective, Bernhard Giesen, Kay Junge: Strukturelle Evolution, Gerhard Preyer: Soziale Gesetze und askriptive Solidarität. Eine Skizze zu den Grundlagen der Gesellschaftstheorie, Anhang: Die modernen Gesellschaften verstehen. Richard Münchs Entwicklungstheorie moderner Gesellschaften, Erwin Rogler, Gerhard Preyer: Relationslogische Darstellung der sozialen Gesetze, Gerhard Preyer: Die modernen Gesellschaften “verstehen”. Zu Richard Münchs Entwicklungstheorie moderner Gesellschaften, Dieter Claessens: Bemerkungen zur Entstehung der modernen Ökonomie: Das Organistionsproblem, Richard Pieper: Strukturelle Emotionen, elementare Strukturbildung und strukturelle Evolution DIE EVOLUTION POLITISCHER ORDNUNGEN Rainer C. Baum: Parsons on Evolution of Democracy, Mathias Bös: Zur Evolution nationalstaatlich verfaßter Gesellschaften, Konrad Thomas: Das Ethnische und das Staatliche, Volker Bornschier: Die westeuropäische Integration als Gesellschaftsmodell im Zentrumswettbewerb ZUR SOZIOLOGIE DES WELTSYSTEMS Immanuel Wallerstein: Evolution of the Modern World-System, Christopher ChaseDunn, Thomas D. Hall: The Historical Evolution of World-Systems: Iterations and Transformations, Albert Bergesen: Postmodernism: A World System Explanation, Richard Münch: Modernity and Irrationality: Paradoxes of Moral Modernization, Walter L. Bühl: Transformation oder strukturelle Evolution? Zum Problem der Steuerbarkeit von sozialen Systemen STATE OF THE ART: Michael Schmid: Soziologische Evolutionstheorien
Suhrkamp Verlag, Frankfurt am Main 1998, Suhrkamp Taschenbuch Wissenschaft
393
Bookpublications
LEBENSWELT – SYSTEM – GESELLSCHAFT Konstruktionsprobleme der “Theorie des kommunikativen Handelns” von Jürgen Habermas *HUKDUG3UH\HU
I DIE ENTWICKLUNGSLOGIK VON WELTBILDERN 1. Die Rationalisierung von Weltbildern 2. Strategien der Analyse von Weltbildern 3. Folgeprobleme und Kritik II GESELLSCHAFT ALS LEBENSWELT UND SYSTEM 1. Die Strukturen der Lebenswelt 2. Allgemeine Bezugsprobleme der soziologischen Evolutionstheorie 3. Die Verständigungsformen 1. Weltbilder und soziale Integration 2. Zur Durkheim-Interpretation 4. Kommunikationsmedien und generalisierte Kommunikationsweisen 5. Folgeprobleme und Kritik III DIE FORMAL-PRAGMATISCHE BEDEUTUNGSTHEORIE 1. Die sprechakttheoretische Grundlegung 2. Interpersonal geltende Bedeutungskonventionen 3. Regelbewußtsein und Handlungskompetenz 4. Die Expansion des semantischen Gehalts 5. Verständigung und die Herstellung interpersonaler Beziehungen IV DER ERWERB DES MORALISCHEN BEWUßTSEINS 1. Die sozial-kognitive Grundausstattung 2. Diskurs und moralisches Bewußtsein 3. Folgeprobleme V KONSTRUKTIONSPROBLEME UND KRITIK 1. Rekonstruktionshypothesen 2. Zu den Konstruktionsproblemen ANHANG: Max Webers Religionssoziologie als eine Typologie des Rationalismus 290 Seiten, DM 28,– , auch als CD lieferbar
digital bookpublication Humanities-Online, Frankfurt am Main http://www.humanities-online.de Email:
[email protected]
394
Bookpublications
Die globale Herausforderung Wie Deutschland an die Weltspitze zurückkehren kann Gerhard Preyer Der Begriff Globalisierung hat bereits Schlagwortcharakter bekommen und ist in aller Munde. Globale Märkte haben die Grundsituation wirtschaftlichen und politischen Handelns grundsätzlich verändert. Das heute entstehende globale Weltsystem und der Medienverbund auf dem es beruht, fordert uns alle heraus. Was bedeutet Globalisierung? Worin bestehen die globalen Herausforderungen, die zu bewältigen sind? Darauf wird eine Antwort gegeben.
NEUE MEDIEN, WISSENSCHAFT UND TECHNIK IM ZEITALTER DER GLOBALISIERUNG Was heißt Globalisierung? – Die neuen Medien: Eine Kopernikanische Wende – Der Überlebensimperativ: Technologieentwicklung – Die Zukunft der deutsche Universitäten VERÄNDERTE KONSTELLATIONEN Globale Wirtschaft und globale Ordnung – Die Globalisierung der Finanzmärkte – Auf dem Weg zur virtuellen Organisation – Das Ende des klassischen Arbeitsmarktes – Zu einer neuen Wirtschaftspolitik EUROPA IM ZEITALTER DER GLOBALISIERUNG Zur Ausgangssituation im Zentrumswettbewerb – Frankreichs Zentralismus: Grenzen der Marktwirtschaft – Das Überlebenssystem Italien – Großbritannien zwischen Tradition und Modernisierung – Deutsche Stärken im Umbruch – Neu gemischte Karten – Die Wiedergewinnung des Standorts Deutschland DIE EVOLUTION DES MITGLIEDSCHAFTSCODES Zur Soziologie der Grenzziehungen – Gesellschaft, Organisation und Interaktion – Gesellschaftsinterne Globalisierung und Entwicklungstrends – Die globalisierte Gesellschaft
290 Seiten, DM 72,–
Frankfurter Allgemeine Zeitung/Gabler Edition, Frankfurt am Main 1998
395
Bookpublications
INTEGRIERTES MANAGEMENT Was kommt nach der Lean-Production? Gerhard Preyer/Jakob Schissler Es gibt heute im Weltmaßstab kein Unternehmen, das seine Umstrukturierung nicht an den Bestandteilen des japanischen Modells der schlanken Produktion und des schlanken Managements, der Lean-Modelle, orientiert. Die Lean-Modelle sind zu einem integrierten Management fortzuentwickeln: dem über Vernetzung integrierten segmentierten Unternehmen, mit durchlässigen, ständig wechselnden Grenzen zwischen Unternehmen, Lieferanten und Kunden. Integriertes Management ist das Unternehmensmodell, das eine erhöhte Qualifizierung der Mitarbeiter und die Fähigkeit, in sich selbst steuernden Teams mitzuarbeiten, erfordert. Das bedeutet aber auch die Entwicklung eines neuen Führungsstils. Mit integriertem Management kann jedes Unternehmen sofort anfangen. Dazu wird in diesem Buch ein Weg gezeigt. Es wird eine wichtige Funktion als Impuls für weitere Entwicklungen haben. Inhalt DIE NEUEN HERAUSFORDERUNGEN 1. Vom japanischen und amerikanischen Management lernen? 2. Information, Wissen und Entscheidungen KREATIVITÄT – ORGANISATION – FÜHRUNG 1. Managementfähigkeiten 2. Innovative Organisation 3. Management und Führungsstile STRATEGIE – MARKT – KULTUR 1. Unternehmensstrategie 2. Controlling 3. Marketing 4. Zur Unternehmenskultur STICHWORT Was kommt nach der Lean Production? 164 Seiten, 32,- DM Frankfurter Allgemeine Zeitung, Verlagsbereich Wirtschaftsbücher, Blickbuch-Wirtschaft, Frankfurt am Main 1996
396
Bookpublications
ÜBER DAS KÄMPFEN Zum Phänomen des Kampfes in Sport und Gesellschaft Axel Binhack
Kämpfen ist stets als ein Beziehungsphänomen, gleichsam als “Fortsetzung der Kommunikation mit anderen Mitteln” zu verstehen. Anhand von sechs Strukturmerkmalen wird eine “Prototypik des Kampfes” entwickelt. Ihre zusammenhängende Dynamik ist gesellschafts- und kulturunabhängig in vielfältigen Konfliktformen zu erkennen. In dem vorgelegten interdisziplinären Anschnitt wird die Struktur der Kampfbeziehung von vergleichbaren Sozialphänomenen abgegrenzt und am Beispiel des Kampfsports in Form des “Zweikampfes” analysiert. 0HUNPDOHHLQHU)RUPDOVWUXNWXUGHV.DPSIHV Der duale Beziehungsaspekt: Antagonismus und Ambivalenz – Der duale Trägeraspekt: Entscheidungsorientiertheit und tendenzielle energetische Totalität – Der duale Inhaltsaspekt: Zweckgerichtetheit und riskante Offenheit 'LH6WUXNWXUPHUNPDOHGHV.DPSIHVLP9HUJOHLFKPLW$UEHLW6SLHOXQGGHQ %H]LHKXQJVSKlQRPHQHQ7DXVFK)OLUWXQG0DFKW Der “antagonistische” Unterschied – Freiheit und Zweckgerichtetheit – “Zeitenthobene” innere Unendlichkeit und “zeitbeschleunigende” Entscheidungsorientiertheit – Die formalstrukturelle “Folgenhaftigkeit” des Kampfes =XP.DPSIVSRUW Kampfsport als Radikalisierung des sportlichen Kampfes – Der Kampfsport als kulturelles Phänomen – Der Kampfdiskurs als Symbol männlichen Agierens – Ästhetische Perspektiven 'LH.DPSIVSRUWDUW.DUDWH'RDOV%HLVSLHO Historische Entwicklungslinien des Karate-Do – Der Schwertkampf und die “Leere Hand”: Zum karatespezifischen Aspekt der Waffenlosigkeit – Die Kultivierung des Kampfphänomens im Karate-Do – Der Kampf als Lehrmeister im Sinne des Zen
Campus Verlag, Campus Forschung Bd. 768, Frankfurt am Main 1998
397
Bookpublications
System der Rechte, demokratischer Rechtsstaat und Diskurstheorie des Rechts nach Jürgen Habermas hrsg. von Werner Krawietz / Gerhard Preyer Editorial: Zivilgesellschaftliche Assoziationen und spätmoderner Rechtssstaat (Werner Krawietz)
I JURISTISCHE ENTSCHEIDUNG, BEGRÜNDUNG UND RECHTFERTIGUNG DES RECHTS IN DER PERSPEKTIVE DISKURSIVER REFLEXIONSTHEORIE Enrique P. Huba: Standortbestimmung in der zeitgenössischen Rechtstheorie - Rawls, Dworkin, Habermas, Alexy und andere Mitglieder der modernen ‘Heiligen (Rede-) Familie’, Thomas McCarthy: Legitimacy and Diversity: Dialectical Reflexions on Analytical Distinctions, Gerhard Preyer: Rechtsgeltung – Argumentation – Entscheidung II ORDNUNG, RECHT UND RATIONALITÄT IM JURISTISCHEN DISKURS Karl-Heinz Ladeur: Rechtliche Ordnungsbildung unter Ungewißheitsbedingungen und intersubjektive Rationalität, Ulfrid Neumann: Zur Interpretation des forensischen Diskurses in der Rechtsphilosophie von Jürgen Habermas, Ota Weinberger: Diskursive Demokratie ohne Diskurstheorie III KONSENSUSTHEORIE VERSUS SUBJEKTPHILOSOPHIE? VORAUSSETZUNGEN UND FOLGEN DES JURISTISCHEN DISKURSES IM DEMOKRATISCHEN RECHTSTAAT Michael Pawlik: Die Verdrängung des Subjekts und ihre Folgen. Begründungsdefizite in Habermas’ “System der Rechte”, Uwe Steinhoff: Probleme der Legitimation des demokratischen Rechtsstaats, William Rehg: The Place of Consensus in Democratic Legitimation: A Recommendation
Rechtstheorie Zeitschrift für Logik, Methodenlehre, Normentheorie und Soziologie des Rechts 3 1996, Habermas-Sonderheft (erschienen 1998) Duncker und Humblot, Berlin
398
Bookpublications
INTENTION – BEDEUTUNG – KOMMUNIKATION Kognitive und handlungstheoretische Grundlagen der Sprachtheorie Gerhard Preyer, Maria Ulkan, Alexander Ulfig (Hrsg.) Einleitung: Zu kognitiven und handlungstheoretischen Grundlagen der Sprachtheorie, Gerhard Preyer, Maria Ulkan, Alexander Ulfig
I INTENTIONEN UND KOMMUNIKATIVE HANDLUNGEN Maria Ulkan: Kommunikative und illokutionäre Akte; Georg Meggle/Maria Ulkan: Grices Doppelfehler. Ein Nachtrag zum Griceschen Grundmodell; Jan Nuyts: Intentionalität und Sprachfunktionen II INTERPRETATION UND BEDEUTUNG Gerhard Preyer: Kognitive Semantik, Anhang Sprechaktsemantik: J.L. Austin, J.R. Searle, H.P. Grice, P.F. Strawson; Louise Röska-Hardy: Sprechen, Sprache, Handeln; Frank Siebelt: Zweierlei Holismus. Überlegungen zur Interpretationstheorie Donald Davidsons; Peter Rothermel: Semantische Implikaturen; Volkmar Taube: Referenz und Interpretation. Zur Theorie nichtsprachlicher Symbolisierung; Georg Peter: Zu Richtigkeit und Interpretation der Metapher: Kognitive Funktion und rekonstruktive Schemainterpretation III KLASSIFIKATION VON SPRECHAKTEN Maria Ulkan: Informations- und Aufforderungshandlungen; Dirk Hartmann: Konstruktive Sprechakttheorie; Volkmar Taube: Bildliche Sprechakte IV KOMMUNIKATIVES HANDELN UND INTERSUBJEKTIVE GÜLTIGKEIT Jürgen Habermas: Sprechakttheoretische Erläuterungen zum Begriff der kommunikativen Rationalität; Karl-Otto Apel: Illokutionäre Bedeutung und normative Gültigkeit. Die transzendentalpragmatische Begründung der uneingeschränkten kommunikativen Verständigung; Peter-Paul König: Kommunikatives und strategisches Handeln. Kritische Bemerkungen zu zwei zentralen Begriffen der “Theorie des kommunikativen Handelns” von Jürgen Habermas; Alexander Ulfig: Präsuppositionen und Hintergrundwissen. Eine Kritik am formalpragmatischen Präsuppositionsbegriff V DIALOGSTRUKTUR UND ARGUMENTATION Wilhelm Franke: Konzepte linguistischer Dialogforschung; Franz Hundsnurscher: Streitspezifische Sprechakte: Vorwerfen, Insistieren, Beschimpfen; Dieter Mans: Argumentation im Kontext, Exkurs: Zu Christoph Lumers “Praktische Argumentationstheorie” 408 Seiten, 58.- DM
Westdeutscher Verlag, Opladen 1997
399
Bookpublications
Wittgensteins Spätphilosophie Analysen und Probleme Wulf Kellerwessel, Thomas Peuker (Hrsg.)
Die Spätphilosophie Ludwig Wittgensteins entfaltet nicht nur in der gegenwärtigen Sprachphilosophie ihre Wirkung, sondern auch zunehmend über den Bereich der Sprachphilosophie hinaus. Zwar macht dieser einen Schwerpunkt des Wittgensteinschen Denkens aus, aber dennoch verdanken die Philosophie des Geistes, die Erkenntnistheorie und die Philosophie der Religion der Wittgensteinschen Sprachphilosophie wichtige Impulse. Diese Schwerpunktsetzung zugunsten der Sprachphilosophie wie auch ihre Zusammenhänge mit den genanten anderen Disziplinen der Philosophie zeichnet der vorliegende Band aus.
INHALT W. Kellerwessel, T. Peuker: Einleitung, A. Ofsti: Methodischer Solipsismus, Metasprachen(problem) und Dexis. Einige Überlegungen zum Privatsprachenproblem, T. Peuker: Die faktische Öffentlichkeit der Sprache. Zu Wittgensteins Privatsprachenargument, P. Niesen: Gemeinschaft, Normativität, Praxis: Zur Debatte über Wittgensteins Regelbegriff, A. Berndzen: Einer Regel entsprechen - einer Regel folgen. Zu einer Kontroverse in der Interpretation von Wittgensteins ‘Philosophischen Untersuchungen’, G. Preyer: Sprachbedeutung ohne Regelbefolgung, C. Stetter: Sprache und Schrift bei Wittgenstein, H.J. Schneider: Mentale Zustände als metaphorische Schöpfungen, W. Kellerwessel: Zum Begriff der Gewißheit in Wittgensteins ‘Über Gewißheit’. Ein Kommentar, M. Kroß: “Glaube Du! Es schadet nicht”. Ludwig Wittgensteins Vermischte Schriften zur Religion - Anhang.
302 Seiten, DM 86,–
Verlag Königshausen & Neumann, Würzburg
400
Bookpublications
Donald Davidsons Philosophie Von der radikalen Interpretation zum radikalen Kontextualismus Gerhard Preyer Für Bruce Aune EINLEITUNG
I RADIKALE INTERPRETATION, LOGISCHE FORM UND EREIGNISSE 1. Donald Davidsons Philosophie: Ein Überblick 2. Wahrheit, Bedeutung und radikale Interpretation 2.1 Radikale Interpretation als ein radikaler Externalismus (1) Der Zirkel zwischen Überzeugung und Bedeutung: Die Asymmetrie von RI (2) Die Strukur von Inhaltssätzen: sagen, daß ... (3) Der Grundsatz der Nachsicht (4) Die Autonomie der Bedeutung (5) Radikaler Externalismus: Der innertheoretische Schritt (6) Die Ontologie von RI 2.2 Von der Idiolekttheorie zum dritten Dogma des Empirismus (1) Ausgangs- und Übergangstheorien der Interpretation (2) Die Demontage eines Mythos (3) Die epistemischen Beschränkungen von Verstehen 3. Die logische Form von Handlungssätzen und die singuläre Kausalaussage 3.1 Logische Form und adverbiale Modifikation 3.2 Kausale Beziehungen 4. Wahrheit und Überzeugung 4.1 Zum Haupteinwand 4.2 Zur empirischen Offensichtlichkeit von Einstellungszuschreibung 4.3 Rationalität als normativer Begriff
II KÖRPERBEWEGUNGEN UND HANDLUNGEN 1. Das logische Verknüpfungsargument
401
Bookpublications
2. Primäre Gründe und die Identitätsthese: Die synkategormatische Fassung von Absichten 3. Basisakte 4. Flucht vor den Körperbewegungen 4.1 Thalbergs Handlungstheorie und der “Akkordeon-Effekt” 4.2 A.I. Goldmans Kritik an der Identitätsthese 4.3 H.L.A. Hart: Zuschreibungen 4.4 Körperbewegungen als Bestandteil von Handlungen
III EINE RADIKALE THEORIE DES HANDELNS 1. Der Begriff der Einstellungsrationalität 1.1 Die Homogenität der Interpretation 1.2 Bewertende Einstellungen 2. Handlungsbeschreibungen und Handlungsverursachung 3. Eine Handlungserklärung 3.1 Überzeugungen, Absichten, Situationen 3.2 Praktische Gedanken 3.3 Handlungsgründe 3.4 Zu Hempels Ansatz 4. Praktische Schlüsse 4.1 Zur Gültigkeit praktischen Schließens 4.2 Entscheidungen und die Ausführung von Absichten 4.3 Überzeugungen und willentliche Handlungen 5. Radikaler Kontextualismus Literatur 290 Seiten, DM 19,– Digital Publication: http://www.rz.uni-frankfurt.de/protosociology
402
Bookpublications
LANGUAGE, MIND AND EPISTEMOLOGY On Donald Davidson’s Philosophy Gerhard Preyer, Frank Siebelt, Alexander Ulfig (eds.) Introduction: On Donald Davidson's Philosophy (Gerhard Preyer, Frank Siebelt, Alexander Ulfig) PHILOSOPHY OF LANGUAGE Jerry Fodor, Ernie Lepore (New Brunswick, USA): Meaning, Holism, and the Problem of Extensionality; Olav Gjelsvik (Oslo, Norway): Davidson's Use of Truth in Accounting for Meaning; Wilhelm K. Essler (Frankfurt/Main, Germany): Was ist Wahrheit? Arend Kulenkampff (Frankfurt/Main, Germany): Eigennamen und Kennzeichnungen EPISTEMOLOGY Roger F. Gibson (St. Louis, USA): Quine and Davidson: Two Naturalized Epistemologists; Eva Picardi (Bolongna, Italy): Davidson and Quine on Observation Sentences; Ralf Naumann (Düsseldorf, Germany): Events and Externalism; Dorit Bar-On (Chapel Hill, USA): Conceptual Relativism and Translation; David K. Henderson (Memphis, USA): Conceptual Schemes after Davidson; Frank Siebelt (Frankfurt/Main, Germany): Singular Causal Sentences and two Relational Views PHILOSOPHY OF MIND AND THEORY OF ACTION Louise M. Antony (Raleigh, USA): The Inadequacy of Anomalous Monism as a Realist Theory of Mind, Louise Röska-Hardy (Darmstadt, Germany): Internalism, Externalism and Davidsons Conception of the Mental, Marcia Cavell (Berkeley, USA): Dividing the Self, Ralf Stoecker (Bielefeld, Germany): Willensschwäche - Wie ist das nur möglich? Klaus Puhl (Graz, Austria): Davidson on intentional Content and Self-Knowledge, Johannes Brandl (Salzburg, Austria): Sharing Beliefs and the Myth of the Subjective, Kirk Ludwig (Gainesville, USA): First Person Knowledge and Authority, Gerhard Preyer (Frankfurt/Main, Germany): Rationalität: Absichten, Primärgründe und praktisches Denken. Donald Davidson: Dialectic and Dialogue 445 Seiten Kluwer Academic Publishers, Synthese Library, Dordrecht
403
Bookpublications
Reality and Humean Supervenience Essays on the Philosophy of David Lewis Gerhard Preyer, Frank Siebelt (eds.)
Preface Gerhard Preyer, Frank Siebelt Reality and Humean Supervenience - Some Reflections on David Lewis’ Philosophy
MODAL REALISM Phillip Bricker: Island Universes and the Analysis of Modality, John Bigelow: Time Travel Fictions, Peter Forrest: Counting the Cost of Modal Realism, Paul Teller: Lewis's Defence of Counterpart Theory, Harold W. Noonan: The Case for Perdurance,
PHYSICALISM, CAUSATION AND CONDITIONALS Daniel Bonevac: Naturalism for the Faint of Heart, D. M. Armstrong: Going throug the Open Door again: Counterfactual vs. Singularists Theories of Causation, Jonathan Bennett: On Forward and Backward Counterfactual Conditionals
REDUCTION OF MIND Terence Horgan: Multiple Reference, Multiple Realization and the Reduction of Mind, Michael Tye: Knowing what it is like: The Ability Hypothesis and the Knowledge Argument
Rowman and Littlefield Publishers, Lanham, USA
404
Subscription
PROTOSOCIOLOGY ISSN 0940-4147 Subscription I subscribe PROTOSOCIOLOGY 30,- DM (15,- Euro) each issue, abroad: $ 30 (including courtage); payable by cheque. No price fixing for Double-Volumes.
Date, signature ................... Separate order I order PROTOSOCIOLOGY volume Nr. ....... Single issue: 35,- DM (17,50 Euro), abroad: $ 35 (including courtage); payable by cheque. All prices plus postage and packing
Date, signature ............
Your address: Address ....................
Name ....................
Send the order to: PROTOSOCIOLOGY ISSN 0940-4147 Editorial office: Stephan-Heise-Str. 56, D-60488 Frankfurt am Main 90, E-Mail:
[email protected], Tel. 069/769461 Can be ordered directly from:
Bockenheimer Bücherwarte Bockenheimer Landstr. 127 60325 Frankfurt am Main, RFA Tel.: 069-771088
PROTOSOCIOLOGY participates in Poiesis Online Service, Philosopical Dokumentation Center, Bowling Green State University, Bowling Green, OH 43403-0189, http://www. bgsu.edu/pdc/
Editor: Gerhard Preyer, Johann Wolfgang Goethe-Universität Frankfurt am Main, Department of Social Sciences. Editorial staff: Georg Peter. http://www.protosociology.de
Digital Publications – Special Offer
405
Digital Publications – Special Offer: Digital versions of ProtoSociology. Vol. 11 is a free download. Visit our homepage under “Special Service”. Vol 14 – Folk Psychology, Mental Concepts and the Ascription of Attitudes Vol 13 – Reasoning and Argumentation Vol 12 – After the Received View – Developments in the Theory of Science Vol.11 – Cognitive Semantics II – Externalism in Debate (free download!) Vol 10 – Cognitive Semantics I – Conceptions of Meaning Vol 8/9 – Rationality II &III (double volume) Single digital version: 7,50 Euro or $ 15.-
Special offer: Three digital Journals for the price of two: 15.- Euro or $ 30 Payable by cheque. E-Mail posting!
Date, signature ............ Your address: Name ....................
Address ....................
Send the order to: PROTOSOCIOLOGY, editorial office: Stephan-Heise-Str. 56, D-60488 Frankfurt am Main 90, Tel. 069/769461 Order via E-Mail:
[email protected] or
[email protected]
Editor: Gerhard Preyer, Johann Wolfgang Goethe-Universität Frankfurt am Main FB 3: Department of Social Sciences. Editorial staff: Georg Peter. Project: Frank Siebelt, Mathias Bös
http://www.rz.uni-frankfurt.de/protosociology
ProtoSociology An International Journal of Interdisciplinary Research Volume14 - 2000 Folk Psychology, Mental Concepts and the Ascritpion of Attitudes: On Contemporay Philosophy of Mind P Editors of the Vol. 14: Gerhard Preyer P Editorial of the Vol. 14: Georg Peter P Layout and Technical Conception: Georg Peter P Editorial Office: ProtoSociology, Stephan-Heise-Str. 56, 60488 Frankfurt am Main, RFA, Phone: 069-769461, E-Mail:
[email protected] P Homepage: http://www.protosociology.de