For Pietroski the relevant worry is over Gricean "non-natural" meaning he has not left Fodor unscathed, even on the assumption that Fodor need have nothing to say about my other cases. I would defend Fodor and his philosophy against Dewitt's charge that they consider mental representation "an idle puzzle". I also believe his worry that a lapse into circularity in the absence of an adequate definition of representation has occured is overstated.
1.1. Pietroski (1993) agrees that Fodor's (1990) theory of mental representation proves inapplicable to the cited work from cognitive science. However, Pietroski asserts that Fodor's intent as well as his commitments require only that he provide a theory of representation for states like beliefs. Pietroski adds that a different, (possibly) compatible theory of representation will help one to understand the work in cognitive science to which I allude. Hence, Pietroski concludes that by pointing out the inapplicability of Fodor's theory of representation to vast areas of cognitive science, my target article (Wallis 1992) is misdirected. For Pietroski, the relevant worry is over Gricean "non-natural" meaning.
1.2. I worry that philosophers of cognitive science pay too little attention to actual, Nobel prize winning work in cognitive science. Equally worrisome is that, on Fodor's account, the outputs of at least some of these "other" processes are Fodorian beliefs. Indeed, properties like barnhood are exactly the sort of "inferred representations" to which Fodor claims to address himself (Fodor 1991, pp.255-57). In my target article, I surveyed various contemporary theories of these "other" processes to see (1) whether any result in Fodor's brand of mental representation, and (2) just how much explanation one can do without Fodor's mental representation. I take it that it should embarrass Fodor that his account does not assign representational content to the outputs of a process like object recognition given the best current theory of that process. Pietroski, however, does not share my worries, so I put aside this disagreement and turn to the barn example, the one case that Pietroski finds interesting.
1.3. Pietroski suggests (section 5) that Fodor remains unscathed by my barn case because, "suppose agents come to believe (because they look closely) that the building in question is only an inch deep. This would break the large-red-(etc.)/B-token connection, at least for the object at hand, without breaking the barn/B-token connection." I find two problems here. First, if the only way to break the facsimile to Sb connection is by having another belief (that barns have a depth greater than one inch, that the object is too thin to be a barn, and/or etc.), then Fodor becomes a content holist. Second, if one could break the facsimile to Sb connection without breaking the barn to Sb connection merely by gaining a new belief about a particular instance, then one could break the barn to Sb connection without breaking the facsimile to Sb connection by falsely believing a barn (perhaps because of the testimony of a friend) to be a facsimile. This last problem is part of the reason Fodor wants "to do it in terms of nomic relations among properties rather than causal relations among individuals" (Fodor 1990, p.102).
1.4. Defining representation in terms of lawlike relations between properties raises its own potential problems for Fodor that I do not mention in my article. For a state to have nomic relations between the property of barnhood and the property of Sb-hood, all manner of barns must cause Sb. So red barns, green barns, barns shaped like giant question marks, barns shaped like balls of yarn, etc., must all cause Sbs. Hence, for Fodor, one cannot recognize red barns as barns unless one can recognize any instance of a barn as a barn. Fodor might add a ceteris paribus clause to the law in an attempt to rule out barns that neither he nor I recognize as barns. I have commented upon the deficiencies of that strategy elsewhere (Wallis, forthcoming).
1.5. Regardless of the problems (or lack thereof) in appealing to properties, the appeal to properties does not help with the first problem: holism. Pietroski, I think, agrees with me that for Sb to count as a barn representation for Fodor, it must be the case that (A) one can somehow rule out every kind of facsimile and (B) the mechanism whereby one rules out facsimiles does not necessitate one's having any other contentful states (i.e. mental representations). The former, (A), is required by Fodor's definition. I hold that (B) is necessary for Fodor to avoid content holism with regard to the mental. Indeed, one can make the following substitutions into Fodor's definition of content holism: The language of thought would be content holistic if "no expression in [the] language [of thought] can have [a content] unless... at least one other [expression] does" (Fodor & Lepore 1992, 257-260). As a result, having a belief that "Yonder lies a barn" cannot require my having other beliefs, such as the belief that "Barns have a depth greater than one inch."
1.6. Pietroski might offer Fodor's defence against the sort of holism accusations I raise here and in my target article. Pietroski might note that Fodor could respond by saying: The functional role (the beliefs and the inferences one makes from them) does not determine the content identity of Sb; content identity gets determined by the fact that the system's tokening of Sbs satisfies the definition for representation with respect to barns. One gets the bad version of holism only when one identifies having a content with having some particular functional role. Fodor might add that he makes appeals to functional role, but that on his account functional role only mediates the system's satisfaction of the representational definition with respect to barns; the functional role does not determine the content. Many different functional relationships might well mediate the satisfaction of the definition. So Fodor's account would be holistic, but not in the bad way, because it does not identify content with some specific functional role (Fodor, 1991, pp.268-71, 301-3)
1.7. Facsimiles are problems for Fodor precisely because ruling them out (even to the extent to which humans are capable of doing do) seems to involve a great deal of knowledge. It is no accident, in my opinion, that Pietroski finds it relevant that "one cannot KNOW (pace Wallis) that (all) large red buildings having extended triangular bodies are barns" (section 4). Fodor/Pietroski have admitted that they need everything -- the system's entire cognitive apparatus -- in order to determine mental representational content. In fact, they need the system and the super-theory by which one could detect all facsimiles. They need lots and lots of beliefs, most (all?) of which must be true. False beliefs, or too few beliefs, will let facsimiles slip by. Indeed, Fodor/Pietroski cannot accept the following scenario involving too little knowledge: Knowing only that barns are red-extended-triangle-on- top-of-extended-rectangle buildings in which farmers keep livestock, I walk out to a good facsimile, and look inside. I see nothing, saying, "Wow, this barn is sure empty." I note that the walls are weak and thin, saying, "Wow, this barn is sure flimsy." I walk away. Nor can they allow the following scenario involving false belief: I know incredible amounts about the king. I think about him often in my dealings with him. I even know that the king has an evil look-alike. However, I falsely believe that the look-alike, not the king, has a small birthmark -- their only difference. Fodor must reject this story, because one can break the king to 'king' connection by revealing the birthmark without breaking the look-alike to 'king' connection. Hence, according to Fodor's account I have never thought of the king as the king. Rather, I think always of the king as the look-alike.
1.8. Are the above scenarios a problem for Fodor? Consider the following dark confessions of the Fodorian soul:
"Now, consider the following (slight) difference between me and my almost-Twin; he thinks "You never find water in Pepsi bottles" [note deleted] and I don't think this. So then there are likely to be environments... in which he would have 'water' thoughts where I would not... This is, of course, a real problem for a naturalistic semantics; maybe it's the ultimate problem for a naturalistic semantics" (Fodor 1991, p.302).
How, then, might Fodor deal with this problem?
"I don't want to consider how an informational theory might hope to cope with this problem; suffice it to say that you get some room to wiggle if you think of the denotation in terms of the nomic relations among properties (instead of covariances among their instantiations) since you might then argue that the Pepsi guy and the non-Pepsi guy are both subsumed by a water 'water' law despite the postulated differences between them. (Remember these are ceteris paribus laws." (Fodor 1991, pp.302-303).
In appealing to functional roles, one inherits the problem (from functional role semantics) of determining the relevant functional roles for assigning content. So, appealing to functional roles does not necessarily solve the disjunction problem, as one may have false beliefs or incomplete sets of beliefs. One therefore cannot determine the relevant functional roles and hence solve the disjunction problem simply by somehow resorting to nomic relations among properties (forget the problems there) and ceteris paribus clauses (forget the problems there). I conclude that Pietroski has not seen Fodor unscathed, even on the assumption that Fodor need have nothing to say about my other cases. I now turn to DeWitt's commentary.
2.1. I agree with DeWitt's (1993) claim (section 3.5) that representation "has important implications for foundational issues in cognitive science... [a]nd these foundational issues are potentially more worrisome than they are generally given credit for." However, I would defend Fodor and philosophy against the charge that they consider mental representation "an idle puzzle" (DeWitt, section 3.5).
2.2. DeWitt first raises concerns regarding the causal efficacy of mental representations given that the physical properties of a system's state (and not the representational properties) seem to cause a system's state transitions. I note that the causal status of mental representations is an instance of the problem of the causal efficacy of mental properties, a problem faced by dualists (the problem of interaction between mental and physical substances) and reductive physicalists alike. Moreover, this problem has not gone unnoticed in philosophy. A brief list of recent attempts by philosophers to address this problem would include Cummins (1989), Dretske (1988), Fodor (1990), Skillen (1984), Sosa (1984), and myself (Wallis, forthcoming).
2.3. DeWitt suggests that Fodor's commitments give him particular problems as regards the causal efficacy of mental representations; he concludes that Fodor must find a naturalized theory of mental representation to save his career project from collapse. It is true that Fodor needs a naturalized theory of mental representation in order to make his overall position work. However, a great many philosophers and cognitive scientists would find their career projects damaged by a the failure to develop a naturalized theory of mental representation. Though the causal efficacy of mental representations is a problem, I find no reason to single out Fodor.
2.4. Turning to DeWitt's allegation of vacuous circularity with regard to explanations of cognition in terms of mental representation, I think that even allowing for his necessary brevity he overstates the case. DeWitt claims (3.4) that lacking a naturalized account "mental representations are, roughly, those things that cognitive agents have. In turn, the usual definition of cognition, and hence of cognitive agents, makes essential use of the notion of representation." However, merely lacking a definition (of mental representation) does not commit one to vacuously circular explanations. In "Le Malade Imaginaire," Moliere points to the seeming vacuous circularity of the medical knowledge of his time. Moliere's doctoral candidate explains the ability of opium to put one to sleep by appealing solely to its "virtus dormitiva," the latin phrase for "sleep-causing power." Representation is not used in this manner in cognitive science. One does not explain object recognition solely by saying that the system has mental representations. Nor does one respond to queries about the nature of mental representation by saying that mental representation consists in the power to recognize objects. One may hold the commitment that cognitive systems have representations without holding either that only cognitive systems represent, or that being had by a cognitive system is sufficient for being a representation. One can in fact maintain the separateness of being a representation from being had by a cognitive system, even if one wants, as Fodor does, to draw a distinction between mental and nonmental systems in terms of the types of representational properties had by their respective states. DeWitt correctly warns against lapsing into circularity in the absence of an adequate definition of representation, but I think his worry that such a lapse has already occurred is overstated.
Cummins, R. (1989) Meaning and Mental Representation. Cambridge, MA: MIT Press.
DeWitt, R. (1993) Representation and the Foundations of Cognitive Science. PSYCOLOQUY 4(11) fodor-representation.3
Dretske, F. (1988) Explaining Behavior. Cambridge, MA: MIT Press.
Fodor, J. (1990) A Theory of Content and Other Essays. Cambridge, MA: MIT Press.
Fodor, J. (1991) Replies. In: B. Loewer & G. Rey (eds.) Meaning in Mind: Fodor and his Critics. Cambridge, MA: Blackwell.
Fodor J. and Lepore, E. (1992) Holism: A Shopper's Guide. Cambridge, MA: Blackwell.
Pietroski, P. (1993) Fodor Unscathed. PSYCOLOQUY 4(10) fodor-representation.2.
Skillen, A. (1984) Mind and Matter: A Problem that Refuses Dissolution in Mind 93 (372) 514-526.
Sosa, E. (1984) Mind-body Interaction and Supervenient Causation in Midwest Studies in Philosophy, vol 9. Minneapolis, MN: University of Minnesota Press.
Wallis, C. (forthcoming) Dretske, Representation and the Mental/Non-mental Distinction. The Journal of Experimental and Theoretic Artificial Intelligence.
Wallis, C. (1992) Asymmetric Dependence and Mental Representation. PSYCOLOQUY 3(70) fodor-representation.1.