Search

A theory for representation of concepts in memory is proposed which emphasizes the association of salient exemplars with the concept. Previous theories dealing with the classification processes involved in acquiring new concepts have shown that clear category boundaries do not exist. It is proposed that present theory must account for the "fuzziness" of category boundaries found in natural languages as well as how learning is nonetheless achieved. A set of exemplars, associated with the concept through learning experience, is used for comparison with a newly encountered stimulus to ascertain if that stimulus is encoded as a set of features. Thus, the exemplars associated with the concept become a collection of intersecting feature sets. In this context, the set of features common to all exemplars is considered the set of "defining" features for the concept. A concept is considered "fuzzy" to the extent that "nondefining" features in the set of represented features are more frequent than defining features. Difficulty of verifying class membership for stimuli is hypothesized to increase with fuzziness of the concept as measured by a proposed index. Tests of the theory are suggested. (Author/PJM)