What Is Cognitive Science?

Robinson, William S.

Von Eckardt, Barbara. A Bradford Book. Cambridge: The MIT Press, 1993. x + 466 pp. $45.00--How is a science--especially one not yet mature--properly to be described? This question is considered in Von Eckardt's first chapter and in a lengthy appendix that summarizes, criticizes, and amends philosophies of science offered by logical positivism, Kuhn, and Laudan. The result, which structures the book, is that a developing science can be characterized by a framework of shared commitments of its practitioners. The shared commitments of cognitive scientists, stated at the end of the first chapter and explained thereafter, concern the domain of the investigation, its basic questions, its substantive assumptions, and its methodological assumptions. In most cases, the commitments are descriptive, that is, they are agreed to by all or almost all practitioners of cognitive science. In some cases, a commitment is defended normatively, for instance, as required to render actual practice coherent. A few claims, such as the requirement for constituent structure in representations, are included in the list with a question mark.

Cognitive capacities, not behavior, are taken to be the proper domain for the explananda of cognitive, science. The basic questions concern the constitution of these capacities, how they work, and how they are related. Methodological assumptions are reviewed in the last (ninth) chapter, the most interesting issue being the relation of cognitive science to neuroscience.

The largest part of the book is, quite properly, devoted to the substantive assumptions which, put summarily, state that the mind is a computer that manipulates representations by rules that are in some sense in the machine itself (computational assumption) and that the mind does this by being a device that contains representation bearers. These representation bearers stand in a grounded representation relation to represented objects, and have significance for the subject in which they occur (representational assumption). The key issue regarding the computational assumption is whether connectionist devices count as computers, and thus whether connectionism shares a commitment that is required for being counted as cognitive science. Von Eckardt rightly includes connectionism. She can claim, however, that connectionist devices satisfy her requirement that rules be in the machine itself only by counting as rules "the transition and output functions governing the operation of each individual unit" (p. 137), despite the fact (which she explicitly recognizes) that such "rules" do not operate on the representations that figure in the cognitive capacities under discussion. Further, her case for saying that connectionist machines can "manipulate" representations (p. 138) fails to apply to the most frequently studied connectionist models.

Von Eckardt begins her discussion of representation by developing a new, very general, neo-Peircean account. There follows a lengthy treatment of vexed issues regarding mental representation, with the views of Fodor, Stich, and Burge discussed most prominently. Occasionally issues of processing of representation bearers are conflated with issues of grounding the representation relation; but on the whole this is a helpful discussion that clarifies the motivations, interrelations, strengths, and weaknesses of views in this tangled area. As in several other cases, the discussion of these issues is preceded by a review of relevant work in cognitive science itself.

This book does not discuss critics of cognitive science, for instance, Hubert Dreyfus and (although he may be surprised to find himself on the list on p. 393) John Searle, It ought, however, to be useful to would-be critics of the information processing approach, as well as to its supporters; for, by presenting relevant scientific background, thorough explanations, and articulate critiques, Von Eckardt has provided a characterization of cognitive science that clarifies the essential doctrines and issues on which productive critical discussion must focus.