The proposition that individuals engage in intergroup discrimination to increase or maintain positive social identity and a high level of self-esteem has received some empirical support. An attempt was made to extend prior findings by evaluating whether intergroup allocation behaviour consistent with subjects' social values would lead to higher self-esteem than inconsistent allocation behaviour. More specifically, it was predicted that competitive subjects' self-esteem will be higher following discriminatory choices than fair choices and prosocial subjects' self-esteem will be higher following fair choices than discriminatory choices. It was also predicted that after subjects were constrained to make discriminatory choices, competitors' self-esteem would be higher than prosocials' self-esteem and after subjects were constrained to make fair choices, prosocials' self-esteem would be higher than competitors' self-esteem. Experiment I supported the first of these predictions when a measure of personal self-esteem was used as a dependent variable. Experiment 2 attempted to extend the generality of the findings of Experiment 1 by defining and measuring self-esteem in collective terms. The expected prior pattern of results did not occur again. Constraining subjects to make discriminatory choices increased their collective self-esteem regardless of their social values.