Computational Models of Higher Cognition

Models in cognitive science span all three of Marr’s (1982) levels (see Holyoak & Morrison, Chapter 1). Normative systems (see Chater & Oaksford, Chapter 2), including the Bayesian framework (see Griffi ths et al., Chapter 3), address the level of computational theory; neural models (see Morrison & Knowlton, Chapter 6) are specifi ed at the implementation level; and informationprocessing models, or process models (perhaps the most common type of model in cognitive science), are specifi ed at the level of representation and algorithm. Process models of higher cognition come in three basic varieties: traditional symbolic models, traditional connectionist models, and symbolicconnectionist models. Th is chapter reviews the basic representational and processing assumptions embodied in each of these approaches and considers the strengths and limitations of each.