Abstract

Fodor and Pylyshyn's critique of connectionism has posed a challenge to connectionists: Explain such cognitive regularities as systematicity and productivity without postulating a LOT architecture. Some connectionists took the challenge seriously. They developed some models that purport to show that they can explain the regularities without becoming classical. The key to their claim is that their models can and do provide non-concatenatively realized syntactically complex representations that can also be processed in a structure sensitive way. Surprisingly, Fodor and McLaughlin seem to agree: such models are non-classical. I develop an abstract notion of syntax and of what it is to be structure sensitive that classicism is essentially committed to. Consequently, I argue that implementation of this minimal abstract notion is enough to explain the cognitive regularities that had prompted to postulate a LOT in the first place. Any model that satisfies a formal system in this abstract sense ought to be seen as a classical LOT model. I show that concatenative or non-concatenative realization of complex representations is an implementation level issue and as such is irrelevant for an essentialist defense of the LOTH. ;I also criticize Stich's Syntactic theory of Mind, and show that the STM has nothing to do with syntax when the notion of syntax is properly understood. First, I argue that if Stich is right in his claim that content-based psychologies have the disadvantages he enumerates, then STM-style theories have exactly the parallel problems; so it is false that the STM framework is scientifically superior to content-based psychologies as Stich claims. Secondly, I argue that the STM as a version of a narrow causal individuation scheme cannot type-individuate abstract syntactic objects onto which the brain states, according to Stich, are supposed to be mapped. Thirdly, I show that the STM-theorist, at any rate, is committed to intentional vocabulary at some stage of theory construction. ;I argue that since a NCA of typing Mentalese expressions is bankrupt, the only individuation scheme is semantic and a molecularist one at that: I criticize Fodor's purely denotational semantics. I also discuss how this provides the right perspective in answering one troublesome question for intentional psychology: if psychological processes are computational, how can psychological laws be intentional?