Single and multiple systems in categorization and category learning

Abstract

Learning to classify the world into categories is fundamental to human cognition. Some categorizations seem to be made explicitly based on rules whereas other categorizations seem to be made implicitly based on similarity. Several theories posit either that multiple learning systems are involved in categorization or that categorization is carried out by a single learning system. The multiple-system approach assumes that people learn new categories via an explicit verbal system and an implicit procedural system. The single-system approach assumes that categories are learned by a single cognitive system that relies on stimulus similarity and selective attention. In this Review, we first provide an overview of the primary theories and models in the field of categorization and highlight the assumptions and operating characteristics of each. We then discuss evidence from cognitive psychology, cognitive neuroscience, computational modelling and comparative psychology to determine which approach is best supported. We conclude that the debate between a multiple-system theory and a single-system approach has not yet been resolved and suggest avenues for future research to create a robust theory that accounts for category learning beyond the laboratory and beyond the confines of the classification learning paradigm. Classifying the world into categories is fundamental to human cognition. In this Review, Minda et al. highlight the assumptions and operating characteristics of theories positing multiple versus single category learning systems and detail evidence for each approach.