Classifying Partial Exemplars: Seeing Less and Learning More

Categories underlie a variety of functions beyond just classification, including inference and explanation. To classify, people need to distinguish between categories, but other functions rely on within-category information (things true of a particular category, independent of others). Despite the need for both types of knowledge, recent work shows that classification does not lead to learning an important type of within-category information, prototypical nondiagnostic information. However, most classification studies are conducted under narrow conditions that do not cover many basic ways that people learn categories. In 2 experiments, the authors compared standard classification learning with a slightly different task where items appeared with occluded features (as many objects appear); they hypothesized that this change might lead to broader attention and learning of within-category, prototypical nondiagnostic information. The results support this prediction, offering evidence that classification can lead to learning within-category information. They discuss the possibility that other classification results may depend on specifics of the standard paradigm.