Math modeling, neuropsychology, and category learning:

raising the issue of the implications of findings in neuropsychological subjects for category learning [Knowlton, B. (1999) What can neuropsychology tell us about category learning? Trends Cognit. Sci. 3, 123–124]1. Knowlton and Squire2,3 reported various dissociations between categorization and recognition in normal and amnesic subjects to support their view that multiple memory systems mediate these tasks. The multiple-system view is more complex than a single-system approach is to modeling categorization and recognition. On grounds of parsimony, a question of scientific interest is whether a single-system model can suffice to account for these data. In support of this possibility, we demonstrated that an extremely simple version of a single-system exemplar model accounted in quantitative detail for the bulk of the dissociation findings reported by Knowlton and Squire, as long as plausible assumptions were made regarding parameter differences in memory ability between groups4. At the same time, we acknowledged some remaining challenges for the exemplar model, and concluded our article by stating, ‘Whether or not a full account of Knowlton and Squire’s complete set of dissociation findings is available within the exemplar approach remains a subject for future research, but we feel that we have made an important start’ (Ref. 4, p. 255). We also outlined various directions in which the exemplar model could be extended to meet these challenges. In the first part of her commentary1, Knowlton reiterates these acknowledged limitations of the simple exemplar model and argues against the possible lines of extension that we suggested. Her first point pertains to the dissociation exhibited by the severe amnesic E.P., who despite showing zero ability to discriminate between old and new exemplars in a recognition task, displayed normal performance in a classification task5. We had pointed out that stimulus conditions were dramatically different across the particular recognition and classification tasks in question. In the test phase of the recognition task, presentations of the target item were rare and widely spaced between distractor items. By contrast, in the classification test, members of the target category were frequently presented. We noted that if E.P. had even a small residual N o s o f s k y a n d Z a k i – R e s p o n s e Update Comment