A model of perceptual classification in children and adults

Abstract

The developmental trend from overall-similarity to dimensional-identity classifications is explained by a quantitative model. I begin with the assumption that objects are represented in terms of constituent dimensions and that the representation of objects changes little with development. Given this assumption, the model has three major parts. First, the similarity between objects is a function of the combination of the constituent dimensional differences. I propose developmental change in the likelihood that dimensions are differentially weighted in the calculation of similarity. Second, the perceived similarities between objects are valued for the purpose of constructing classifications. I propose that similarities are valued more dichotomously with age, such that identity becomes increasingly special. Third, the valued similarities are used to choose the best classification of those possible. The model provides good qualitative fits to the extant data. Three experiments examining classifications in 2- to 8-year-olds and in adults support specific new claims of the model. The data and the model provide new insights about development, classification, and similarity.