Dissociating cognitive and affective uncertainty using a General Linear Classifier

Abstract

A number of categorization models have been proposed that consider classification performance and uncertainty in terms of prototypes, category boundaries, and exemplars. Like other models of categorization, category boundary models (e.g., GLCs) only consider cognitive uncertainty while failing to consider affective uncertainty. Using a modified GLC, measures of affective uncertainty were obtained by combining exemplar-based information (e.g., response frequency) and categorical information (e.g., categorization accuracy) in varying proportions (0/100, 25,75, 50/50, 75/25, and 100/0). We provide evidence that categorical and exemplar based representations likely inform affective uncertainty in simple categorization tasks.