Children and adults learn family-resemblance categories analytically

3 experiments examined the modes of processing used by children and adults in learning family-resemblance categories. The materials were cartoon faces (Experiments 1 and 2) and bugs (Experiment 3) divided into categories that possessed no single defining attributes, but rather several characteristic attributes that were each partially predictive of category membership. The categories were structured so that a holistic mode of processing in which the individual did not selectively weight any given attributes could have led to success. Nevertheless, preschoolers (Experiments 2 and 3), first and third graders (Experiment 1), and adult college students (all experiments) all exhibited primarily analytic modes of learning that consisted of single- and dual-attribute approaches. Although the proportion of analytic learners among the preschoolers was lower than among the adults in Experiment 3, in no case were holistic modes of learning evident. The results are discussed in terms of their implications for young children's apparent relative success in learning natural categories. It is suggested that children's success in learning real-world categories may be based, in part, on an interaction between a basically analytic processing style and natural category structures that provide many partially informative attributes.