Conservative Rationalism I

Abstract

Reductive behaviourism, as we have seen, collapses to behavioural holism; we shall now turn to see that Fodor’s computational mentalism, together with the referential theory of meaning and nomic theory of reference, collapses to reductive behaviourism, so that Fodor, willy nilly, ends up with the holists. In Section 1.2, I described Fodor’s account of the mind in terms of a hypothetical supervenience chain that links organisms with their environment in this sense: identical organisms instantiate the same computational mental system; and identical mental systems carry the same semantic contents; and identical semantic contents determine the same extensions and truth-conditions. I then rehearsed the familiar Twin-Earth arguments to the conclusion that no theory of mind structured by the supervenience chain can be correct, since the physical and social environment itself contributes to the semantic identity of mental states. Thus one could vary the environment whilst holding a computational system the same; and this shows that the sameness of computational states is not sufficient for the sameness of extensions or truth-conditions; which in turn shows that mental states cannot be identified with the computational states of a physical system. In short, the Twin-Earth arguments uphold the principle of extension-meaning supervenience, but reject the supervenience of the mind on a computational neural system, and in general on any physical system whatever. Fodor’s (1987) response was that the supervenience of mind on brain is compatible with the supervenience of extension on meaning, relative to a context. A context is a relevantly local universe of discourse, or domain of interpretation, with respect to which an organism’s mental symbols have their meanings fixed by certain nomic relations of reference; and the requirement of relevant localness ensures that a domain including both Earth and Twin Earth is not a context. In contrast, Fodor’s (1994) response was that the Twin-Earth scenario is not nomologically possible, and therefore cannot tell against the nomic theory reference and referential theory of mental content.