Abstract

Chapter 1 marks out the plan of the dissertation and some of the more important arguments contained in the dissertation, including arguments to the best explanation, the poverty of stimulus argument, productivity of thought argument, systematicity of thought argument, and the compositionality of thought argument. I claim that connectionist theory is a viable explanatory theory, and contender to the classical theory of cognition. The purpose of Chapter 2 is to explain the modularity thesis, and then scrutinize Fodor's arguments for perceptual modules and central system devices. Chapter 3 inspects the evidence Jerry Fodor uses to support his distinction between perceptual modules and central system devices. In particular, "Precis of The Modularity of Mind" presents the empirical evidence for modular perception while "Modules, Frames, Fridgeons, Sleeping Dogs, and the Music of the Spheres" presents the empirical evidence for central system devices. In particular, a Poverty of the stimulus argument and the Muller-Lyer illusion show how perception is computationally complex but encapsulated from the agent's beliefs and desires. The conclusion of this chapter is that one major theoretical flaw with the modularity thesis is that it cannot make sense of higher thought or decision making. I turn to an exploration of connectionism in Chapter 4. The primary purpose of Chapter 4 is to introduce connectionist architectures, and make clear the differences between a classical architecture and a connectionist architecture. Chapter 5 answers the Fodor and Pylyshyn's challenge to connectionist theorists to explain the productivity of thought, systematicity of thought, and compositionality of thought. I conclude that Paul Smolensky satisfactorily answers these arguments, but fails to answer Pinker and Prince's objection that connectionist theory is not explanatory because the input used to train connectionist models does not generalize to actual data observed in the environment. I call Pinker and Prince's argument the Empirical authenticity argument. Chapter 6 examines a connectionist model of apparent motion. The goal of this chapter is to scrutinize whether or not Dawson's connectionist architecture can solve the motion correspondence problem and do so in a way that also preserves the empirical authenticity of the input data. My claim in this chapter is that the connectionist model examined goes no distance to rebutting the empirical authenticity argument. Accordingly, one important flaw in connectionist theory is made obvious. The final chapter of the dissertation follows a theory laden observation versus theory neutral observation discussion between Paul Churchland and Jerry Fodor. The goal of this chapter is to make as clear as possible the complexity surrounding the interpretation of data. The primary conclusion of the dissertation is that connectionist must do more research on the stimulus conditions if they are to claim explanatory rights to cognitive behavior