One of the aims of social psychologists interested in group processes is the development of intervention strategies designed to reduce prejudice, improve social inclusion, and create more harmonious intergroup relations (Gaertner & Dovidio, 2000; Pettigrew, 1998). Although the study of multiple social categorization is varied and wide-ranging, part of its appeal is its potential for attaining such goals. The recognition that multiple and cross-cutting social categorizations may reduce intergroup bias has, from early work, been a key motivation (for reviews see Crisp & Hewstone, 1999a; Crisp & Hewstone, this volume). In particular, the common ingroup identity model (Gaertner & Dovidio, 2000; Gaertner, Mann, Dovidio, Murrell, & Pomare, 1990; Gaertner, Mann, Murrell, & Dovidio, 1989) and the crossed categorization model (Crisp, Ensari, Hewstone, & Miller, 2002; Deschamps & Doise, 1978) have developed as potentially viable methods of encouraging more positive intergroup relations. Recent findings, however, suggest that such interventions do not invariably attenuate intergroup bias, and that there may be important moderating conditions that are yet to be fully specified. In this chapter a program of research that has focused on one such moderator, social identification, is reviewed. I argue that social identification is a key determinant of when and whether creating common ingroups can improve intergroup relations and, on the basis of this, suggests an agenda for future research seeking to maximize the bias-reducing potential of multiple categorization.