Base-rate neglect in ALCOVE: A critical reevaluation

A recent hybrid model of categorization (Attention Learning Covering Map [ALCOVE]; J. K. Kruschke, 1992) has combined the most desirable properties of exemplar models with a connectionist architecture and learning rule. A critically important property of ALCOVE is its apparent ability to account for base-rate neglect, a phenomenon beyond the purview of previous exemplar models. This article reexamines ALCOVE's base-rate neglect predictions and shows that they are confined to a very limited set of circumstances. In most cases, ALCOVE is unable to produce base-rate neglect.