Learning categories at different hierarchical levels: A comparison of category learning models

Three formal models of category learning, the rational model (Anderson, 199 0), the configural-cue model (Gluck & Bower, 1988a), and ALCOVE (Kruschke, 1992), were evaluated on their ability to account for differential learning of hierarchically structured categories. An experiment using a theoretical ly challenging category structure developed by Lassaline, Wisniewski, and M edin (1992) is reported. Subjects learned one of two different category str uctures. For one structure, diagnostic information was present along a sing le dimension (1-D). For the other structure, diagnostic information was dis tributed across four dimensions (4-D). Subjects learned these categories at a general or at a specific level of abstraction. For the 1-D structure, sp ecific-level categories were learned more rapidly than general-level catego ries. For the 4-D structure, the opposite result was observed. These result s proved highly diagnostic for evaluating the models-although ALCOVE provid ed a good account of the observed results, the rational model and the confi gural-cue model did not.