LILLY-MARLENE
RUSSOW
S T I C H ON T H E F O U N D A T I O N S
OF C O G N I T I V E
PSYCHOLOGY
Most of us engage occasionally in the activity of explaining the behavior and mental states of our fellow humans. We hypothesize that Harold is worried because he believes that his job is in jeopardy, and that Mildred took the bus to work because she did not want to drive her car. The background against which we proceed is sometimes referred to as "folk psychology". Two salient features of folk psychology are its reliance on the concepts of "belief" and "desire", and its frequent reference to the content of beliefs and desires as part of its explanations. In short, folk psychology comprises intuitions about explanations in terms of causally efficacious intentional states. Philosophers who have explored the nature of cognitive psychology have often noted that much progress has been made in that discipline since psychologists abandoned the behavioristic strictures associated with Watson and Skinner, and began talking about information, the content of perceptual beliefs, and other intentionalistic concepts. This has generally been taken as evidence that cognitive psychologists are doing the right thing in importing the intentional machinery of folk psychology into their theories I or at least that psychology should make use of some of the intentional idioms associated with folk psychology, albeit perhaps in a modified form2. 1. STICH'S
POSITION
Not everyone, however, has been willing to accept the close alliance of an effective cognitive psychology with the commonplace notions of folk psychology. Stephen Stich has recently offered an extended argument to show that cognitive psychology should reject folk psychology as an adequate starting point 3. Folk psychology, he argues, often refers to the content of mental states and often characterizes them as quasi-linguistic entities which can be placed in correspondence to sentences (p. 28). Thus, folk psychology must rest on some version of a "mental sentence theory" which tells us how to identify the content of mental states. He then attempts to distinguish two sorts Synthese 70 (1987) 401-413. O 1987 by D. Reidel Publishing Company
402
LILLY-MARLENE
RUSSOW
of "mental sentence" theories, the "narrow causal account" (hereafter referred to as " N C A " ) and the "content account". After explaining the distinction, Stich argues (a) that the two are not equivalent, (b) that when we do folk psychology the former must be rejected in favor of the latter (pp. 47-9), and (c) that because folk psychology relies on the content account, a mature cognitive psychology cannot be founded on folk psychology. In other words, Stich seems to think that cognitive psychology as currently practiced, and as praised by philosophers like Fodor, is inherently schizophrenic. On the one hand, its reliance on intentional states allies it with folk psychology, and this entails, according to Stich, accepting the content account of mental sentences. On the other hand, a truly useful and well developed psychology must avoid the limitations imposed by the content account. In order to do so, Stich recommends that cognitive psychology reject the content account, which would entail a total rejection of the folk psychology which depends on it, in favor of something he calls "the syntactic theory of the mind". As its name implies, a syntactic theory does not make reference to the content of mental states (or mental sentences) at all: cognitive theories which cleave to the STM pattern treat mental states as relations to purely syntactic mental sentence tokens, and they detail the interactions among mental states in terms of the formal or syntactic properties of these tokens (p. 9).
T h e interactions among these formal structures is still causal, so it is not the case that causal accounts are inimical to the basic projects of cognitive psychology. Rather, Stich wants to claim that too much attention to the content, the semantic aspect of mental sentences, imposes damaging restrictions on the scope and methods of cognitive psychology. T h e result of all of this is that the STM follows the N C A (now discredited as a possible foundation for folk psychology) in the way it identifies states as tokens of the same sentence type. T h e difference between them is that the STM does not worry about identifying the content of those tokens and types, while the N C A was explicitly introduced as a means of identifying and analyzing the semantic element of mental states. One consequence of this is that a N C A may identify a mental sentence and an utterance as tokens of the same type, whereas the STM would identify mental sentences only as tokens of some formal language that has a syntax but no semantics.
STICH
ON
COGNITIVE
403
PSYCHOLOGY
Stich's strategy, once he has cast folk psychology in the position of ~yelying on a content account, is to argue that any version of cognitive i~sychology which preserves this essential aspect of folk psychology will be hopelessly hampered. It will not, for example, apply to those who are not sufficiently like u s - m e m b e r s of radically different cultures, small children, or people with brain damage. He therefore concludes that we would be better off avoiding talk of the content of mental sentences entirely, and restricting our psychology to a formal analysis of the sentence tokens that affects our behavior. H e labels theories which do consider the content of mental sentences important "representational theories of mind". As one might predict by now, his defense of STM is primarily an argument to show that it is better than representational theories. Accordingly, it will be undercut if it can be shown that his analysis of representational theories is flawed, that representational theories need not eschew a causal account in favor of a content account, and that, therefore, representational theories can claim many of the advantages that Stich wants to deny to them. 2.
DISTINGUISHING
THE
CONTENT
AND
NARROW
CAUSAL
ACCOUNTS
In this paper, I shall concentrate on a demonstration that Stich fails to establish a significant difference in the analyses afforded by the N C A and the content account. Although the scope of this charge may appear somewhat narrow, there are several important implications for Stich's general views on the nature of cognitive psychology. Since he fails to show that the two yield different conclusions, he cannot demonstrate that folk psychology must prefer the content account, or indeed that it must have an implicit preference for one or the other built into its basic theory. Finally, his attempt to divorce cognitive psychology from folk psychology depends on the claim that folk psychology is wedded to the content account, so this thesis also remains unproven. In the final section of this paper I will argue that Stich fails to establish the STM as the only viable foundation for cognitive psychology. In order to begin his argument, Stich must first characterize the "narrow causal account" and the " c o n t e n t account", two explanations of how we might identify sentence tokens when they occur in our heads. Both ask us to accept the hypothesis that a belief or other
404
LILLY-MARLENE
RUSSOW
propositional attitude "is a relation between a person and a sentence token in that person's brain" (p. 30). T h e distinction between them lies in the different ways each identifies the features that make a neural pattern in my brain a token of one sentence type rather than another. A causal account defines type identity in terms of "similar patterns of causal interaction" (p. 48). More specifically, a n a r r o w causal account restricts the relevant causal interactions to "those obtaining between mental states, between mental states and stimuli, and between mental states and behavior" (p. 48). In contrast to these, the content account looks to semantic properties such as truth conditions, meaning (p. 48), similarity of "doxastic n e i g h b o r h o o d " (p. 89), and other features related to content. In comparing the two approaches, Stich correctly notes that Fodor and most functionalists hold that the two approaches will almost always yield the same identification of mental sentences and beliefs (pp. 49, 107). Indeed, the assumption that content can be accounted for in terms of causal connections lies at the heart of most representational theories of mind. He then sets himself the task of showing that this is a mistake, that the two accounts diverge dramatically, and that when they do, it is the content account that matches the intuitions which inform our folk psychology. He uses two sorts of problem cases to accomplish all of this 4. T h e first group emphasizes the importance of holism in identifying our beliefs (p. 53), while the second group focuses on some puzzles regarding reference (p. 60), especially in connection with proper names, indexicals, and kind terms. In both cases, Stich argues that the N C A and the content account imply different descriptions of the content of the intentional states involved in these examples, and that the content account accords better with the intuitions of folk psychology. 3. HOLISM T h e first example Stich presents is the case of a woman, Mrs. T, who is afflicted with a progressive m e m o r y loss. She can assent to the claim " M c K i n l e y was assassinated", or utter that sentence in response to a question about what happened to McKinley, but has forgotten so many other things that she does not know whether a person who has been assassinated is alive or dead, or even whether she is alive or dead. H e claims that a content account allows us to say/that Mrs. T
STICH
ON COGNITIVE
PSYCHOLOGY
405
does not believe that McKinley was assassinated, but that the N C A entails that she does h a v e that belief. Stich attributes the conclusion that Mrs. T does believe that M c K i n l e y was assassinated to the N C A by insisting that an N C A must refer to potential as well as actual causal interactions with mental states, .stimuli, and behavior when it identifies or characterizes a mental state (p. 54). H e offers no evidence that any actual articulation of an N C A endorses this condition; he asserts it on the strength of an example. Imagine a person who is in a state which we tentatively identify as "believing that all men are mortal". According to a narrow causal account, one identifying characteristic of such a belief is its role in causing the belief that Socrates is mortal. Since no such effect would occur in a person who did not believe that Socrates was a man, Stich claims that a narrow causal theory must allow reference to potential causal interactions: we can, according to Stich, correctly identify the belief only if we are allowed to use the fact that if the subject did believe that Socrates was a man, the relevant causal interaction would obtain. H e thus concludes that "for a narrow causal theorist, the type identity of a mental state is determined by its potential causal interactions..." (p. 54, Stich's emphasis). Since we assume that if Mrs. T did have beliefs like "assassinated people are dead", and " I am not dead", those beliefs would interact with the inner token corresponding to her claim that McKinley was assassinated, Stich concludes that the N C A ' s need to admit potential causal interactions yields the conclusion that Mrs. T does have the belief in question. T h e simplest response to all of this is to charge Stich with attacking a straw man. It is certainly true that a N C A would endorse the inference from " O t t o believes that all men are mortal" to "although Otto does not believe that Socrates is a man, if he did have that belief there should be causal interactions which produce the belief that Socrates is m o r t a l " . H o w e v e r , this does not entail the further claim that such counterfactuals must play a role in determining the initial identification of the belief that Otto has. N o r does Stich offer an example of an actual N C A which endorses this sort of reliance on potential interaction. By correctly characterizing an N C A ' s analysis of potential causal interactions one would recognize, as Stich does, that certain counterfactual claims would follow f r o m an N C A about what would happen if
406
LILLY-MARLENE
RUSSOW
a subject had some beliefs she does not in fact have. H o w e v e r , they would follow only if it had first been established that the subject had certain other beliefs to begin with. Thus, if we start with the assumption that O t t o does believe that all m e n are mortal, it follows on the N C A that the t o k e n inscribed in Otto's brain would interact with other tokens (if he had them) to produce mental states like the belief that Socrates is mortal. Nothing in this commits the N C A to any claim a b o u t the role of potential interactions in establishing the antecedent assumption that Otto believes that all m e n are mortal. Thus, it seems quite reasonable to expect an N C A to insist that a fairly large n u m b e r of actual causal patterns must be present in order to establish the identity of a belief. Indeed, it must rely on such actual patterns while initially identifying a mental sentence or the belief involving that sentence, because one could generate counterfactuals of the sort Stich discusses only if one had already identified the belief state which actually obtains. Establishing type identity between belief states, or between tokens of mental sentences, must therefore precede any counterfactuals of this sort. Since Mrs. T lacks almost all of the actual causal interactions that the N C A would d e m a n d of a belief that McKinley was assassinated, the N C A supports the conclusion that she does not have that belief. In cases in which some but not all of the normal causal interactions are present, as with Mrs. T in earlier stages of her progressive loss of m e m o r y , the N C A , l i k e the content account, need not set up a standard for strict identity, but might instead m a k e its ascriptions of belief a matter of degree: the m o r e closely two states correspond with respect to their actual causal connections, the m o r e justified we are in judging t h e m to be type identical or m o r e similar to one another. In any event, there is no evidence of any radical disparity between the N C A ' s conclusions and those of a content account, although there m a y be some minimal disagreements about borderline cases. Mrs. T is not Stich's only case in which holism, or the relation of the mental sentence in question to other states, is relevant. H e also considers fanciful cases in which we are confronted with an individual who has e m b e d d e d in his brain a token of a single statement in a highly e v o l v e d future scientific theory, and m o r e m u n d a n e cases of children who can assert things like " E = m c 2'' without having the f r a m e w o r k of scientific beliefs within which that claim is properly understood. As with the case of Mrs. T, Stich thinks that an N C A
STICH
ON
COGNITIVE
PSYCHOLOGY
407
would have to say that these are clearcut examples of beliefs, while folk psychology and the content account would hold that something more confusing is going on. All of these examples involve the presupposition that the N C A must be informed by potential causal interactions, and hence is subject to the same rebuttal as was given to Stich's analysis of N C A ' s treatment of the case of Mrs. T. 4.
REFERENCE
IN MENTAL
SENTENCES
T h e second set of examples Stich uses to drive a wedge between the N C A and a content account of mental sentences deals with reference. More specifically, Stich discusses three examples which highlight different aspects of the problems surrounding reference and belief: proper names, kind terms, and indexicals. I shall begin with Stich's last criticism regarding reference and the NCA, that which addresses the role of indexicals. H e describes two presidential candidates who each believe " I will be the next president of the United States", and suggests that we have conflicting intuitions about whether they have the same belief or different beliefs. H e takes this to be a reason for rejecting the NCA: "since these tokens will be causal isomorphs of each other, the narrow causal theory cannot account for that strand in our intuition which insists that the two candidates have different beliefs" (p. 65). T h a t is, he ascribes to the N C A (1) the position that the two candidates have the same belief, and (2) the inability to account for any tension or equivocation in our judgment about the classification of their beliefs. In contrast to this, he thinks that the content account c a n do justice to these intuitions, and hence should be distinguished from, and preferred to, the NCA. Stich's reason for the first of these claims is that the two beliefs are taken to be "causally isomorphic". But surely this is, as Stich might say, "wildly implausible". T h e token of the mental sentence in one candidate's brain is causally interconnected with beliefs like "the initials of the next president will be WM", "Mrs. Mondale will like being the First Lady", and "the A F L - C I O will be pleased about the election results"; in the other candidate's brain, the set of interrelated states and mental sentences will be quite different. Thus, the N C A would lend strong support to the conclusion that the two candidates possess different beliefs. Since Stich thinks that the content account also entails that the candidate's beliefs are different (pp. 97-8), no
408
LILLY-MARLENE
RUSSOW
substantial disagreement between the two approaches has been uncovered. A second puzzle case involves two people, T o m and Dick, who have both heard a few remarks about someone called " I k e " . On the basis of these remarks, they each come to believe that Ike was a military man, that he often uttered malapropisms, and that he played golf. T h e problem is that T o m lives in the late 20th Century, and his informants on lke and his behavior were talking about Dwight Eisenhower. Dick is a Victorian Englishman who has heard stories about Reginald Angell-James, whose nickname was " I k e " . In this case, Stich argues that the content account supports our intuition that the two men have different beliefs, and even, given the appropriate context, allows us to ascribe to one the belief that Eisenhower plays golf, and to the other the befief that Angell-James plays golf (pp. 62, 95). H e then claims that "the network of beliefs about Eisenhower in which T o m ' s belief is e m b e d d e d is entirely parallel to the network of beliefs in which Dick's beliefs about Angell-James are e m b e d d e d " (p. 62); neither of them have any beliefs about " a person named Eisenhower" or " a person named AngeU-James". If this is true, it would follow that an N C A must admit that the two sets of beliefs are identical, and hence would contradict the interpretation that Stich thinks is dictated by folk psychology. T h e claim that T o m and Dick have parallel networks of beliefs, however, seems reasonable only if we substantially underestimate the range of such networks. Dick certainly must believe that his compatriots were not discussing a twentieth century American, while T o m must at least hold that possibility open. Similarly, it is reasonable to expect that Dick, knowing that Ike was in the military, would infer that Ike either fought on foot or on horseback. T o m ' s belief about Ike would not lead him to draw the same conclusions. In short, the network of beliefs in which these few fragments about Ike are embedded is vast and contains many beliefs about the way the world is, what contemporaries might be talking about, and other beliefs with respect to which the two men have little in common. We cannot, as Stich attempts to do (p. 62) wave aside considerations of holism, for it is precisely those considerations which allow a causal theory to differentiate between the two beliefs. Thus, the N C A does not lead to the conclusion that this example is a clear-cut instance of two people with the same belief. Nor need it
STICH
ON COGNITIVE
PSYCHOLOGY
409
jump onto the other horn of the dilemma and agree that T o m believes, in an unproblematic sense, that Eisenhower likes to play golf; T o m ' s belief lacks most of the causal connections that would license that claim. In fact, it can occupy the same neutral, context dependent, middle ground that leads Stich to find the content account so attractive (p. 95). T h e final case in which reference is supposed to cause a problem for accounts of mental sentences deals with kind terms. Stich begins with the claim that in England, the term " c h i c o r y " refers to what Americans call " e n d i v e " , and the term " e n d i v e " refers to what an American would call "chicory". Against this background, he spins the tale of an American named John and a Briton named Robin who both hear, and accept as true, the claim "chicory is bitter" (p. 63). If we grant Stich the assumption that the vegetable called " e n d i v e " in the U.S. ("chicory" in England) is not really bitter, it seems natural to say that John has a true belief, while Robin's belief seems prima facie false. Since the content account is sensitive to considerations of truth value, such an account would, in at least some contexts, lead to the judgment that the two men have different beliefs. T h e argument then parallels the one concerning T o m and Dick: Stich alleges that the N C A must conclude that the two beliefs are i d e n t i c a l - a conclusion he takes to conflict with the content account's analysis of this situation, with our intuitions, and with f01k psychology. This case differs in an important respect from the previous one. T o m ' s and Dick's beliefs were of the sort that inevitably have many connections to beliefs about the way the world is, what century one is living in, what "the military" is like, and so on. Since their beliefs about these matters are very different, the N C A could justify the claim that their beliefs about " I k e " were not identical. In contrast to this situation, a belief that would lead John or Robin to utter the statement "chicory is bitter" has relatively few connections with other beliefs, and the few that it has (e.g., that ice cream is sweeter than chicory) are likely to be the same for the two men. Thus, there is a sense of "same belief" in which the N C A does seem to entail that John and Robin share the same belief, and there is a principled argument to justify treating this case differently from the case of T o m and Dick. In order to argue that this example describes a situation in which the content account and the N C A yield substantial disagreements about
410
LILLY-MARLENE
RUSSOW
the identification of mental sentences, Stich must still show that the content account leads to a d r a m a t i c a l l y different analysis of the situation. Unsurprisingly, Stich thinks that it does. H e claims that the content account is m o r e flexible and context dependent. In most contexts it will lend support to the claim that Robin, but not John, believes that endive is bitter. In cases where we want to draw attention to Robin's unwillingness to produce or assent to an utterance of the f o r m " e n d i v e is bitter" (Stich asks us to imagine him as a contestant on a quiz show) we would be able to withdraw that ascription (pp. 95-6). I think that Stich has collapsed two different issues here. T h e first issue concerns the identification of the mental sentences, and the second concerns what these sentences refer to. T h e need to keep these issues separate is not limited to analyses of mental sentences. E v e n in discussions of uttered or written tokens of sentences, it is c o m m o n place to assume that two tokens of the same sentence type, such as two utterances of the sentence " m y spouse is in the next r o o m " can refer to different people, and to different states of affairs; they m a y e v e n h a v e different truth values. Factors other than the words which o c c u r in a sentence affect the reference. One would expect that if this view of reference is accepted, it would apply to mental sentences as well: two mental sentences of the same type might nonetheless refer to different things. In either case, it would be a mistake to use the divergence of reference to deny type identity. In the case of John and Robin, then, an N C A can endorse the claim that the mental sentences involved in the two beliefs are type-identical, e v e n though their beliefs are about different things. T h e question " d o they h a v e the same belief?" or " d o they believe the same thing?" is ambiguous; it sometimes asks a b o u t the mental sentence, and sometimes is c o n c e r n e d with the object of belief - whether they believe the same thing to be bitter. Since the context of the question usually reduces this ambiguity, the N C A , like the content account, will lead us to conclude that the appropriate answer to this question will depend on the context. T h e N C A and the content account do differ in one respect: the content account will lead to the conclusion that the mental sentences themselves are tokens of different types. It will do so because the sentences have different truth conditions, and truth conditions are relevant in a content account. Nonetheless, in the same context they
STICH
ON COGNITIVE'PSYCHOLOGY
411
will give the same answer to "do John and Robin believe the same thing?" Folk psychology is more concerned with the latter sort of question than it is with the esoterica of type/token distinctions, and so this difference does not establish one or the other account as the preferred foundation for folk psychology. 5.
IMPLICATIONS
FOR
COGNITIVE
PSYCHOLOGY
Stich's aim in the first part of his book was to show that the narrow causal approach to mental sentences leads to conclusions that conflict with our ordinary intuitions or folk psychology, while a content account is more congenial. In this paper, I have tried to show that the N C A and content theories coincide in their judgments about all but perhaps some borderline cases. This agreement has far-reaching consequences for the second portion of the book. Stich proceeds on the assumption that a cognitive psychology based on folk psychology must base its identification of beliefs on a content theory rather than a narrow causal account, and argues that this basis is inadequate. If, as one might suspect on the basis of the similarities between an N C A and the STM which Stich endorses, a narrow causal account avoids the difficulties posed by a content account, there may be no reason why a cognitive theory should sever its folk psychology roots. T h e problem that remains is to determine whether cognitive science should " m a k e serious use of the concepts and language of folk psychology" (p. 127), i.e., whether it should refer to states like beliefs and desires and treat those states as involving a relation to mental sentences that have a content or meaning. Stich argues that theories preserve these concepts by adopting a model or paradigm which he labels "the representational theory of mind" (RTM); R T M , then, is a model which preserves the basic assumptions and theory of folk psychology 5. If Stich's argument in the first half of the book had been successful, we would now be in a position to conclude that an R T M would rest on a content account of mental sentences. Does the failure of that argument to exclude the use of an N C A affect his thesis that his STM is superior to R T M ? A content account, according to Stich, identifies and individuates beliefs by relying on some tacit comparisons between the subject and ourselves. "In saying what someone else believes, we describe his belief by relating it to one we ourselves might h a v e " (p. 79, cf. p. 84).
412
LILLY-MARLENE
RUSSOW
T h e relation is not one of identity, but similarity; rather than judging that two beliefs have identical contents, we draw our conclusions about degrees of similarity (p. 86). Thus, content accounts, and any R T M based on a content account, will be (a) observer relative, and (b) vague and tied to context (pp. 135-36). Stich thinks that these factors render an R T M unacceptable as a foundation for a scientific enterprise. T h e charge of observer-relativity is, in Stich's mind, the most telling. It entails that the more the subject diverges from the standard reference point of a fully language speaking adult who lives in a Western E u r o p e a n culture, the less we will be able to say about his or her mental states. H e cites examples of people with radically different beliefs (pp. 98-102, 136-40), perhaps as a result of cultural difference or psychological delusions; people who, like Mrs. T, suffer from brain damage (pp. 141-43); and small children in the process of acquiring complex concepts (pp. 143-44). Where such people differ radically enough from us, we will not gain much insight or clarity by trying to find the similarities between their states and ours, so Stich concludes that "we will have to do without a cognitive developmental psychology, without a cognitive clinical psychology, and without a cognitive comparative psychology" (pp. 144-45). Stich correctly notes that an STM which relies on an analysis of the causal role of a sentence token does not require reference to our own n o r m a l belief states, and hence can "eliminate the middle m a n " (p. 158) and the limitations imposed by observer relative points. We can grant this, and note at the same time that the same advantage accrues to an N C A which refers to content as well as syntactic type. We have just seen that folk psychology, and thus RTM's, can call on N C A ' s as readily as content accounts, so the charge of restricted scope must be dropped. Stich also worries that an R T M based on a content account will be too vague and context-sensitive to support useful psychological generalizations (158-60). Causal accounts are presumed to be more precise and definitive. Although Stich's arguments are not always lucid on this point, it still seems clear that any advantage assigned to a STM in virtue of its generalizations couched in terms of causal role must also be granted to an N C A for the same reason. Th{as, Stich will be unable to show that an STM is superior to an R T M that utilizes an N C A to identify mental sentences. Such an R T M would preserve as
S T I C H ON C O G N I T I V E
PSYCHOLOGY
413
many of the concepts of folk psychology as prove useful, and will not fall prey to the confusions brought on by the content account taken in isolation. In short, a representational theory which accepts the NCA would be in a position to match the strengths of Stich's favored alternative, the Syntactic Theory of Mind, and would do so while preserving the intuition that mental states have content or meaning. Because he has not been able to show that folk psychology must reject the NCA in favor of a vague and ineffectual content account, he has not succeeded in displacing folk psychology or the Representational Theory of Mind from a mature psychology. NOTES 1 Jerry Fodor is perhaps the most outspoken defender of this view. See, for example, "Three Cheers for Propositional Attitudes" or "Methodological Solipsism Considered as a Research Strategy in Cognitive Psychology", both reprinted in RePresentations, Cambridge: MIT Press, 1981. 2 This more cautious line is taken by D.C. Dennett. See Brainstorms, Montgomery, Vt: Bradford Books, 1978. 3 Stich, Stephen: 1983, From Folk Psychology to Cognitive Science, Cambridge: MIT Press. All page references to this work will be given in parentheses in the text of the paper. 4 Stich also discusses a third set of cases which he claims presents difficulty for the NCA. These cases involve causal differences which would lead an NCA to more finely grained discriminations than those endorsed by folk psychology. I do not discuss these examples here because Stich does not give us any independent grounds for thinking that the content account will yield an analysis which differs from that of the NCA, or which is preferable from the standpoint of folk psychology. 5 Stich actually distinguishes two versions of RTM, a weak representational theory and a strong one. In Strong RTM, the generalizations of folk psychology will make essential reference to the content of mental state types (p. 129). Weak RTM is closer to STM in that the generalizations will apply to mental states in virtue of their syntax or formal properties (p. t85), but differs from STM in holding that "these syntactic objects must have content or semantic properties" (p. 186). However, since Stich thinks that the same objection applies to both versions, this distinction does not affect my argument. Department of Philosophy Purdue University West Lafayette, IN 47907 U.S.A.