FLEXIBILITY IN CATEGORIZATION AS A FUNCTION OF VARIABILITY OF REPRESENTATION TYPE: STUDIES IN FEATURE WEIGHTING AND CATEGORICAL BIASING

Abstract

<p>For nearly the last fifty years, most research on the acquisition and application of conceptual knowledge has focused on structural relations and abstract descriptions as the composing a concept. In this thesis I show that conceptual content is much richer than this, and includes instantiated forms of knowledge as well as abstract descriptions. This instantiated knowledge, both at the featural and procedural level, is necessary to explain error patterns in probe and biasing studies. Furthermore, this variety in content is shown to be systematically linked to different decision patterns, providing an explanation for the flexibility of categorization. Much critical learning done in concept formation, therefore, involves learning optimal feature descriptions and task-appropriate procedures. These ideas are discussed in terms of embodied cognition.</p>