CATEGORY LEARNING BY INFERENCE AND CLASSIFICATION

The nature of category formation is linked to the tasks applied to lea rn the categories. To explore this idea, we investigated how three dif ferent methods of category learning-Classification Learning, Inference Learning, and Mixed Learning (a mixture of the two)-affect the way pe ople form categories. In Classification Learning, subjects learned cat egories by predicting the class to which an individually presented exe mplar belonged given feature information about the exemplar. In Infere nce Learning, subjects learned categories by predicting a feature valu e of a stimulus given the class to which it belonged and information a bout its other features. In Mixed Learning, subjects received the Clas sification task on some trials and the Inference task on other trials. The results of two experiments and model fitting indicate that infere nce and classification, though closely related, require different stra tegies to be carried out, and that when categories are learned by infe rence or by classification, subjects acquire categories in a way that accommodates these strategies. (C) 1998 Academic Press.