Human Category Learning: Toward a Broader Explanatory Account

Abstract This chapter describes and integrates a set of theoretical and empirical contributions to the psychology of human category learning. A synthesis of computational and behavioral approaches is brought to bear to advance a similarity-based account that rejects the stimulus generalization underpinnings of the reference point view (such as exemplar models) and extends its explanatory scope beyond the circumscribed traditional artificial classification learning (TACL) paradigm. The framework advanced in this chapter emphasizes the role of generative (as opposed to discriminative) category learning and takes its focus in the form of the DIVA model based on a neural network architecture called the divergent autoencoder. Directions and advances that extend beyond TACL include: modes of category learning, item presentation conditions, types of categories, and measures of category representation.