In Experiment I, first and fourth graders were shown pictures of faces. Shapes of ears (relevant features) and of eyes and moustaches (irrelevant features) varied orthogonally. The task was to identify the faces according to the shape of the ears. In transfer, old or new shapes of eyes, ears, and moustaches varied independently over trials. Loss of discrimination was expected with new eyes and moustaches, and a greater loss with eyes, designated critical features (CF), than with moustaches, noncritical features (NCF). In Experiment II, first, fourth, and sixth graders were given the same task as in Experiment 1. Half of each group was given 24 trials beyond criterion in acquisition. In both experiments, errors in identifying old ears increased significantly with age and with new CF and NCF. With additional training introducing one irrelevant feature increased errors more when the second one was also changed. No differential CF and NCF effects were found. The findings were consistent with a prototype model.A controversy has arisen around the question of whether individuals acquire concepts by abstracting only distinctive relevant features, as suggested by. Gibson (1969) or through the formulation of a schema or prototype which includes irrelevant aspects as well. Evidence for Gibson's distinctive-feature hypothesis (1969) was given in a study by Pick (1965). She found that when children were trained to make discriminations based on the transformations of letter-like forms, transfer was better with new forms and the same transformations than with old forms and new transformations. However, in a replication by Caldwell and Hall (1970), children performed better with the same letterlike forms and new transformations, thus providing some support for a prototype model.The present study was designed to further determine the extent to which irrelevant aspects of the stimulus form part of concept learning in children. The study is concerned with systematic investigation of substitution of different values of an irrelevant dimension during transfer on concept loss. Pictures of faces were presented and children were required to identify them by the shape of the ear. Shapes of eyes and moustaches also varied but were irrelevant to identification. During transfer, shapes of eyes and ears were either the same asThe authors thank D.