Systematicity, Structured Representations and Cognitive Architecture: A reply to Fodor and Pylyshyn

TL;DR: The connectionist backlash is under way. as discussed by the authors argues that the Fodor and Pylyshyn critique is based on a mixture of philosophical confusion and empirical shortsightedness, and that connectionism constitutes a distinctive but inadequate cognitive model, or if it constitutes an adequate cognitive model it must do so by specifying an implementation of distinctively classical processing strategies and data structures.

Abstract: The connectionist backlash is under way. In a recent edition of Cognition, J. Fodor and Z. Pylyshyn launch a powerful and provocative critique aimed at the very foundations of the connectionist programme. In effect, they offer the friend of connectionism an apparently fatal dilemma. Either connectionism constitutes a distinctive but inadequate cognitive model, or, if it constitutes an adequate cognitive model, it must do so by specifying an implementation of distinctively classical processing strategies and datastructures. (A similar dilemma is urged on connectionists by Pinker and Prince in the same issue.) In this Reply I argue that the Fodor and Pylyshyn critique is based on a mixture of philosophical confusion and empirical shortsightedness. The structure of the Reply is as follows. I begin (Section 1) by contesting, on purely philosophical grounds, a certain picture of the nature of thought ascription. Once thought-ascription is seen for the holistic enterprise it is, the facts concerning the systematicity of thought (the mainstay of Fodor’s and Pylyshyn’s argument) are revealed as conceptual inevitabilities. The power of the Fodor and Pylyshyn critique depends, by contrast, on seeing those facts as a contingent empirical regularity to be explained by a certain model of in-the-head processing. I go on (Section 2) to consider a variety of more specific worries concerning the need for structured, in-the-head representations.