The theoretical rationale underlying the present investigations begins with the assumption that perceptual categorization of racial groups is a prerequisite for subsequent attitude development. It is predicted that if increased perceptual similarity of other groups does indeed initiate attitude acquisition, then it follows that a decrease in similarity should conversely make it more difficult for the child to maintain negative attitudes. For the 1st study, 192 nursery school and kindergarten Ss were used. Half of the Ss were black and half white. In order to test the prediction, two-choice discrimination learning tasks were used to assess similarity. The tasks employed schematic facial drawings as discriminanda. Stimuli were presented by means of a Kendler-type apparatus with two apertures. The findings are generally in accordance with the view that racial labels may increase the perceptual similarity of another race prior to the time that children enter the first grade. The second study sought to obtain data with regard to the question of what happens to attitudes when perceptual differentiation techniques are introduced. Ss were 96 black and white children from the second and sixth grade. The two tests of major interest to this study are the Projective Prejudice Test and a Social Distance Index. On both instruments, Ss were told that testers wanted to know what kinds of things make children want to be friendly with other children. Findings include: (1) Younger children expressed more prejudice, and (2) Both distinctive labeling and stimulus predifferentiation training elicited lower prejudice scores than did a no-label control condition. (CK)