LINGUIST List 6.737

Fri 26 May 1995

Review: "The Elm and the Expert" Jerry Fodor (1994)

Editor for this issue: <>

Directory

  • Daniel Seely, Jerry Fodor (1994), "The Elm and the Expert"

    Message 1: Jerry Fodor (1994), "The Elm and the Expert"

     Date: Thu, 25 May 1995 10:19:35 Jerry Fodor (1994), "The Elm and the Expert"
    From: Daniel Seely <>
    Subject: Jerry Fodor (1994), "The Elm and the Expert"


    Review of Jerry Fodor (1994), "The Elm and the Expert: Mentalese and its Semantics," A Bradford Book published by The MIT Press.

    Reviewed by John R. Lee (johncogsci.edinburgh.ac.uk), University of Edinburgh.

    In this book, Fodor's project is to dig himself out of a hole that he sees himself, and many others, as having got into since cognitive science began to take shape. Being essentially a series of public lectures (the first annual series of Jean Nicod lectures), the book revels in Fodor's characteristically popularising style, replete with even more than usual of the raciness, witticisms and asides that make him so delightful or exasperating, according to taste. But it's the vigour of the arguments -- tantalising, penetrating, old or new, convincing or implausible -- that make the book such a readable canter around the main points of interest in the field, the rapid pace covering a remarkable amount of ground for the scant 130 pages the book occupies. Of the four lectures, the first three attempt to show that Fodor's preferred theory of mind is coherent, while the fourth then considers some of the implications if it's true. Two appendices deal briefly with related philosophical problems. Much in the book is, of course, not particularly new. It is intended for a relatively general audience -- if more of philosophers than of linguists -- and although those having no previous acquaintance with the literature are likely to find it hard going (if only because of the pace), a fair amount has to be in some sense introductory. The compactness of the discussion provides a very clear synoptic view of Fodor's ideas, and most usefully of the ways they have developed in recent years, but this also means that many issues are skated over rather rapidly -- as Fodor admits -- the most important omission being the arguments for and against the material Fodor tends to take as given premises, e.g. his three initial theses sketched below. Some references are given, but not many. The central difficulty that Fodor sees is that the three main theses implied by the most prevalent views on meaning and mind are apparently not altogether compatible. These theses are (1) that psychological explanation is intentional, (2) that intentional content somehow reduces to information, and (3) that psychological laws are typically implemented by computational processes. While none of these is, of course, universally accepted, they do seem to characterise a fairly "standard" view in cognitive science, and indeed one that Fodor regards as having no serious competition. The problem that arises is due to the tension between the assumption underlying (2), to the effect that content has to be described in terms of things external to the mind (which the information is "about"), and the assumption underlying (3) that the causal mechanisms in computation are sensitive only to the form (or syntax) of representations, which is wholly internal (Fodor's famous "formality criterion" of old). If intentional description can't be coherently related to the underlying mechanisms that cause behaviour, how is (1) possible? How can there be a "naturalistic" account of intentionality? In the past, Fodor has been prominent among those who held that (2) was wrong, and that nomological uses of semantics for internal representations (the same thing as intentional content, on a computational view) would have to be worked out on an "internalist" or "narrow" basis. The arguments for this are many and strong, ranging from Putnam's famous "Twin-Earth" examples, which indicate that "externalist" or "broad" content can vary between states with identical causal roles, to the Fregean observation that the psychological role of representations may differ though the broad content is the same (e.g. as when "Jocasta" has to be distinguished from "Oedipus' mother"). It's here that the title of the book comes in: Putnam also discusses cases where, although experts can tell the difference between two concepts such as "elm" and "beech", I can't; so it follows that when I think "elm" I'm in the same state as when I think "beech", even though broad content individuation would have to distinguish these, and it's the discriminative ability of the expert, not mine, that determines the truth conditions of my thought. Fodor now wants to hold, though, that broad characterisation of content, which he has come to see as essential to informational theories, is compatible with computational explanation, and in fact that narrow descriptions of content are superfluous; so these kinds of arguments must be refuted. Fodor's approach is to exploit the relationship between the notion of nomological possibility and the idea that semantics, on the informational view which he derives from Dretske, is to be characterised via counterfactuals. Broad content can be regarded as "nomologically supervenient" on computation -- i.e. computation will reliably produce behaviour that coheres with a broad interpetation of its states -- provided that the right counterfactuals are supported about how the system would behave in given circumstances. And this can happen, Fodor proposes, even if based on no "metaphysical necessity" such as would arise from identities between computational and semantic properties -- all that's needed is for these properties to co-occur in ways that are reliable and explicable, if contingent. Then it may be shown that, although Twin Earth and Fregean cases can arise in principle, they will in fact always be at most aberrations from the normal case, and we will be able to explain this by reference to the other laws of science (Twin Earths, Fodor claims, aren't in fact nomologically possible). Fodor further observes that the success of rational behaviour depends on, inter alia, agents being in "epistemic equilibrium" with respect to the facts, in the sense that having any additional relevant information would not normally cause the agent to act differently. So in normal cases, if it's important to know that "a=b", the agent does know this, and the Oedipus situation does not arise: "What happened to Oedipus was exceptional enough to write a play about". Even though the content of the internal representations of "a" and "b" may be the same, however, they may importantly differ in what Fodor (following Frege) calls "mode of presentation", interpreting the latter as syntactic distinctions between "sentences of mentalese". Hence, it's still possible to reconcile psychological laws based on broad content with explanation of aberrant cases. (This discussion, though Fodor doesn't mention the fact, has much in common with Davidson's "principle of charity", which similarly insists that to treat agents as intentional we "must count them right in most matters".) Lecture 3 presents a problem arising from considerations about reference. Semantically equivalent expressions must apply to the same things; so concepts that carry the same information should be coextensive, otherwise the theory of content can't be purely informational. This provides an excuse for a long and complex discussion of Quine's question about the indeterminacy of reference: how do we know that "rabbit" refers to rabbits rather than undetached rabbit parts? Fodor's strategy is to show that we can find the latter ontology to be inconsistent with an informant's language use if we know what inferences he accepts, and that he does not interpret terms ambiguously. That we can know this turns on the principle that conjunction (in general, connective) reduction can't apply across referential ambiguities -- so if "A" is an ambiguous name then (i) "A is a rabbit", say, and (ii) "A is a rabbit part" can both be true in the same situation; but (iii) "A is a rabbit and a rabbit part" can't be. Thus if an informant accepts the inference from (i) and (ii) to (iii) (and so agrees that (i) and (ii) can't both be true) then we thereby know that "A" is referentially unambiguous -- provided that we can identify e.g. predicate conjunction without begging Quine's question. Fodor simply derives it from sentence conjunction; but although can see that the informant relates connected sentences in various ways (and we can call this "inference", given Fodor's naturalisation of that as simply a causal relation between sentence tokens), how can we decide that a given connective is, say, conjunction? We might need a bit more convincing that Fodor is fully entitled to his assumption that the semantics of sentential connectives come "for free". In the enterprise of radical translation, it seems more likely that they will have to be worked out hand-in-hand with the development of the ontological scheme. In his fourth lecture, Fodor surveys the uses of his theory. The problem of interaction -- how the mental affects the non-mental -- seems to be solved, because we can now see that the causal powers and intentional contents of representations are directly related: the computational roles of representations are determined by just those of their properties (i.e. the syntactic ones) that also carry information about the world through their relation to the system's actual or possible causal history. The other main benefit, as compared with non-representational theories, is said to be an explanation of experimentation. Fodor suggests we are creatures that are able to have generally true thoughts, not just about the world, but also about the contents of our thoughts. Since these contents depend on truth conditions and aetiology, we are able to see how to construct situations in which (roughly speaking) we will come to have a certain thought if and only if it is true. This ability Fodor takes to point the way to integrating Quine's notion of a"naturalised epistemology" into a cognitive, as opposed to behaviourist, psychological framework. And here we return to the title of the book, because consulting an expert is somewhat like an experiment (conceived as consulting nature), in that one puts oneself knowingly into a situation where one should come to believe something just if it's true. The mistake in Putnam's original argument, Fodor thus claims, is in treating this kind of "conceptual deference" to the expert as constitutive of the semantics of terms, whereas actually it's simply epistemological. And the error of the Instrumentalists in philosophy of science was in taking their generally correct epistemology to imply a verificationist semantics, whereas according to Fodor only an externalist semantics can explain why the experimental strategy can be successful. Science depends on a kind of "cognitive management" -- on both an individual and social scale -- that can most clearly be seen to make sense given these relations between belief content and causation. Fodor concludes that we now have the makings of a "naturalist consensus that is Realist in ontology and epistemology, externalist in semantics, and computationalist in cognitive psychology", which shows how spontaneous and productive, rational behaviour can be coherently accounted for. Despite the book's combative and polemical style, and its revisionism in terms of Fodor's own earlier writing, one somehow does not feel the emerging position to be radically new; which perhaps fits in with the notion that it's a consensus. Probably the most fundamental idea in the discussion is that psychological laws, being the laws of a "special science", are ceteris paribus laws; that therefore unsystematic exceptions are not a problem; and accordingly that all psychological explanation has to do, in a certain sense, is to "get the counterfactuals right", with the emphasis on "nomic" rather than causal connections as basic. This idea underlies all the solutions offered to Putnam, Frege and Quine, and induces a striking tolerance of various consequences, including even that a word (such as "gavagai") in a foreign language might turn out in fact to be radically untranslatable. There is something very pragmatic about this approach, which appears to have moved Fodor methodologically closer to, say, Dennett. Getting the counterfactuals right is an instrumentalist imperative, and using this externalist, broad- based ascription of content as the basis of one's account of internal states seems not very far from the idea of projecting semantics inwards from the "intentional stance". The main differences thus seem more confined to the assumed metaphysical basis; but the arguments around the Realist assumptions of the Dretskean background are not considered here.