RELATIONS BETWEEN PROTOTYPE, EXEMPLAR, AND DECISION BOUND MODELS OF CATEGORIZATION

Categorization models make assumptions about: (1) the representation of the stimulus and the contrasting categories; (2) the exact information that must be collected before a response can be made; and (3) response selection-how the subject selects a response after all the relevant information has been collected. Equivalence relations are derived between prototype, exemplar, and decision bound models with respect to these three assumptions. An exemplar model that assumes a deterministic response selection process is proposed that includes the generalized context model (GCM) as a special case. The general linear classifier of general recognition theory (GRT) is shown to contain the probabilistic weighted prototype model as a special case. The set of all percepts that are equivocal with respect to category membership define the equivocality contour. The GCM is shown to predict the same equivocality contour as the GRT optimal classifier under certain restrictive distributional conditions. Several more general exemplar models are proposed that can mimic the equivocality contour of the GRT optimal classifier under less restrictive conditions. © 1993 Academic Press, Inc.