Reproduction of the typicality effect by a neural network storing nonuniformly correlated patterns

Abstract

The typicality effect is one of important findings on concept hierarchies in the field of cognitive psychology. The followings are some of well-known phenomena: if asked to list up all the instances of a concept, we recall them in descending order of their typicality; if asked whether an instance belongs to a concept, we can judge faster as its typicality is higher. As the first step toward a neuronal basis of the typicality effect, this paper proposes a neural netwrok model which recalls a typical instance of a concept with a higher probability and also in a shorter period of recalling time than an atypical one.