Within the crossed categorization paradigm we examined the consequences of cooperative and personalized contact under conditions that promoted attention to unique attributes of team members (decategorization) versus conditions that emphasized their category memberships. As predicted, when the rule for composing teams emphasized one or both of the experimentally induced dimensions of category distinction, ethnocentric bias generalized to members of another team as a function of category similarity. When, instead, the rule emphasized team members' unique attributes, shared in-group memberships no longer moderated bias towards members of another team. Instead, there was an equivalence pattern in the evaluative ratings of the four targets (in-group/in-group; in-group/out-group; out-group/in-group; out-group/out-group) of the crossed categorization paradigm. Copyright © 2001 John Wiley & Sons, Ltd.