Exemplar versus Prototype Models in Predicting Children’s Category Membership

This research compared two classical models of semantic concepts, prototype and exemplar-based models, in predicting category membership. Heit and Barsalou’s (1996) exemplar-based instantiation model was compared to a prototype predictor model based on Rosch and Mervis’ (1975) family resemblance. For three semantic categories (clothes, fruit, and vehicles), category judgements were collected for 25 possible exemplars. Four age groups (5-, 7-, 10-year-olds, and adults) performed a category membership judgement task and either a feature generation task or an exemplar generation task. The category judgement scores were correlated with the family resemblance measure based on the generated features to assess the validity of the prototype-based model. The exemplar-based model was tested by correlating category judgement scores with rated similarity of the 25 possible exemplars towards generated category exemplars. Results of the model comparison and the specific tendencies across age groups will be discussed in the light of the development of category representations.status: accepte