Computation and Content

Abstract

The dominant program in cognitive psychology since the demise of behaviorism in the 1960s has been computationalism. Computational theories treat human cognitive processes as a species of information processing, and the systems that implement such processing as symbol-manipulating systems. Describing a device as a symbol manipulator implies that it is possible to treat some of its internal states as representations of properties or objects in a particular domain. Computational theories of vision, for example, posit internal states that can be interpreted as representing the depth of the distal scene. There has been considerable disagreement about the nature and function of representational contents assigned to the states posited by computational theories. It is widely thought that such theories respect what Jerry Fodor (1980) has called the condition, which requires that computational processes have access only to the formal (that is, nonsemantic) properties of the representations over which they are defined. It is by respecting the formality condition that computationalism promises to answer one of the most pressing problems in the philosophy of mind-how can representational mental states be causally efficacious in the production of behavior? Representational mental states, according to computationalism, have their causal roles in virtue of (roughly) their structural properties.' But this advantage comes at a price. The formal character of computational description appears to leave no real work for the semantic properties of the mental states it characterizes. Thus, computationalism has been thought by some to support a form of eliminativism, the thesis that denies that intentionally characterized states play a genuinely explanatory role in psychology (see, for example, Stich 1983). If the content of computational states is indeed explanatorily idle, then the relation