The study was designed to test several hypotheses derived from intergroup theory concerning the effects of the presence of a novel social category on the formation of intergroup attitudes. Elementary school children (N = 61; aged 6-9) were given measures of classification skill and self-esteem and assigned to 1 of 3 types of school classrooms in which teachers made: (1) functional use of "blue" and "yellow" groups assigned on the basis of a biological attribute, (2) functional use of "blue" and "yellow" groups assigned on the basis of a random drawing, or (3) no explicit groups (despite the presence of blue and yellow groups). After 4 weeks, children completed measures of intergroup attitudes and behavior. As predicted, the functional use of color groups affected children's attitudes toward group members, with children showing consistent biases favoring their own group. Children with higher levels of self-esteem showed higher levels of intergroup stereotyping.