Try a bit of thinking outside the vat

This is a volume in the series Philosophers and their Critics and it is a measure of Noam Chomsky's importance that he is the only figure in the series so far who is not a professional philosopher. Chomsky is here because he made linguistics a more vital and central discipline by drawing philosophical conclusions from his theories of language structure and of language learning, reviving a version of the innateness hypothesis, against the prevailing currents of behaviourism and empiricism.

The collection contains ten papers on these and other philosophically important aspects of Chomsky's work, along with his responses. There are four or five discernible themes and the book makes a useful contribution by exploring them in this form. As in Chomsky's own writing, his views on the topics are typically scattered in footnotes or emerge during the course of a discussion of something else, and are seldom the sole topic of one of his papers or books.

The first four papers, by William Lycan, Jeffrey Poland, Galen Strawson and Frances Egan, explore physicalism and the mental. Chomsky's own view seems to be that: (a) the opposition between the mental and the physical is largely (for science) meaningless - the physical simply describes what we understand through the naturalistic methods of the hard sciences; (b) (some) mental phenomena will eventually be explainable in these terms - the mental is not a different realm from the physical; (c) some mental phenomena (free will, consciousness) may always resist this kind of explanation, at least for creatures constituted as we are, but in this respect they are no different from some phenomena we think of as physical.

Identity theorists claimed that mental states were just a different way of describing neurophysiological states, which ties mental states to particular realisations. Lycan defends functionalism - the claim that mental states are individuated via their causal role, and that the same state can be realised within different hardware - as a solution to this difficulty, but does not persuade Chomsky that this is the correct path. Chomsky's reply suggests that he thinks it premature to impose a priori requirements of abstraction on investigations of cognitive processes. Rather, a satisfactory naturalistic account would not pose an "identity question" requiring some extra solution over and above an understanding of the mechanisms embodying these processes within the organism.

Poland's paper draws out some of the consequences of Chomsky's views on the physical, recommending "methodological physicalism", an endorsement of Chomsky's views on the to-be-hoped-for eventual unification of cognitive theories with physical mechanisms. Strawson describes "real materialism", pointing out, among many other interesting things in a long and detailed essay, just how entrenched dualist thinking is in our tradition, and how in reality our knowledge of the so-called mental is more direct and understandable than our knowledge of the so-called physical, although the emphasis in philosophical discussion is almost always in the other direction.

Papers by Egan and Georges Rey discuss the notion of mental representation as it appears in Chomsky. The normal construal of "representation" is as a relational noun: representations are of something. In Chomsky, as in much of the computational theory of mind literature, it is not so clear that representation is being used in this way. It often seems like a synonym for what computer scientists call a data structure: a representation that may or may not be of something external to the machine but where this external reference is not important - rather, the syntax and associated computational properties are what define the structure.

Chomsky has argued that the content of such representations does not contribute to their explanatory power, a position that seems inconsistent with the desire to say that whatever computational operations are being performed in using language, seeing, reasoning and so on, they are more than the mere shufflings of symbols.

Egan points out that without referring to content, we cannot say that a cognitive event involves a misrepresentation: a mistake. Chomsky's view is, I think, that this is just not part of the theory, but an intentional external interpretation on our part: just as for Jerry Fodor's methodological solipsism, the theory itself cannot distinguish brains in people from brains in vats.

Peter Ludlow and Paul Horwich discuss the role of semantics in Chomsky's approach to language. Chomsky has frequently made comments about referential semantics that have been interpreted as critical of the enterprise known as formal semantics - in essence, the attempt to treat natural languages as if they were formal ones more complicated than, but still akin to, first-order predicate calculus.

For example, he has pointed out that this approach assumes an oversimplified relation between words and their referents, ignoring the fact that whatever relation there is between "London" and the corresponding city will be different from that between a relational word such as "flaw" or "hole" and whatever they are taken to refer to. This is all true but pushing against an open door: people such as Donald Davidson and Richard Montague were well aware that except for a very few cases the truth-as-meaning enterprise "leaves the whole matter of what individual words mean exactly where it was". Chomsky's rejoinder is that it is not much of an advance in a theory that claims to be explaining how people can use language to refer to things in the world by taking as primitive a notion of reference that does not distinguish between the ways in which terms such as "flaw", "London" and "snow" relate to those things.

In the remaining papers, Paul Pietrowski produces an analysis of the inference patterns exemplified by "if John boiled the water, then the water boiled" using Davidsonian events and some hidden verbs that he tries to justify on minimalist syntactic grounds. Ruth Millikan attempts to produce a definition of public or e-language, a concept that Chomsky has claimed to serve no useful purpose, not corresponding to anything in reality. Chomsky is not convinced by the attempt. Alison Gopnik defends her "theory theory" account of learning - that children are little scientists - as an alternative to the innateness hypothesis. Chomsky points out several crucial differences between scientific theory formation and language learning, and he concludes that the theory theory does not provide a plausible alternative.

I found this a stimulating if uneven collection of essays. There is much to be learnt and occasionally some amusement to be gained from Chomsky's patient responses to what he clearly regards as gross misunderstandings of his position.

Stephen Pulman is professor of general linguistics, University of Oxford.