Abstract

The dissertation is a detailed exploration of the philosophical implications of a new research program in cognitive science known as Connectionism or Parallel Distributed Processing . Connectionism is a model of cognitive processing inspired by the network-like architecture of neuronal arrays in the brain. It is a dramatic departure from past accounts of thought, which were typically based upon a more conventional computer metaphor that viewed human cognition as a form of symbol manipulation. It is important to note that my project is not a defense of connectionism. Rather, it should be regarded as an extended answer to the following question: If connectionism provides us with a plausible new theory of the mind, in what ways could this modify current philosophical views or provide us with important insight in areas typically of concern to philosophers? ;The first chapter provides a brief overview of the dissertation. The second chapter is devoted to introducing connectionism and explaining how it differs from earlier approaches to cognitive modeling. The third chapter examines the ways in which the success of connectionism would bear upon various philosophical accounts of mental representation. In the fourth chapter I consider a recent argument by Paul Thagard that connectionist models undermine functionalism. It is my contention that Thagard's charge is mistaken and stems from a failure to distinguish two different notions of program and a failure to distinguish two senses of hardware. The focus of the fourth chapter is on the link between connectionism and eliminative materialism. I argue that folk psychology presupposes propositional attitudes are represented in a discrete and modular fashion. Since many connectionist models of memory and information processing encode information in a way that is incompatible with this assumption, they endorse a picture of the mind that undermines common-sense psychology. Finally, in Chapter 6, I assess the frequently heard claim that connectionist language learning devices undermine Chomsky's arguments for nativism. Contrary to these claims, I argue that connectionist language models must possess biases antecedent to learning, but that those biases need not be the sort of linguistic universals endorsed by Chomsky.