The minimal group paradigm (Tajfel, Billig, Bundy and Flament, 1971) has been influential in the study of intergroup relations. Thus far, most minimal group experiments have divided the subjects either into two groups, or have categorized them on two separate dichotomous dimensions in cross-categorization experiments. This study examines the minimal group paradigm using three distinct and independent groups. Comparison of the results with three minimal groups with those of a baseline two-group experiment shows that with a three-group structure there is no significant ingroup bias. It is suggested that the two-group minimal group experiment shows ingroup bias because subjects access a dichotomous categorization, and that this dichotomous categorization primes a competitive orientation. A two-group context may be particularly efective in evoking an ‘us versus them’ contrast. Self-categorization as a group member is more likely to occur in the presence of two groups whereas three minimal groups renders an ‘us–them’ contrastive orientation less salient. The absence of intergroup discrimination found in the present minimal group study may be limited to the behaviour of minimal or artificially created groups. In the real world of intergroup relations discrimination towards multiple outgroups is a well-known phenomenon. While this study should be regarded as only preliminary research, further elaboration and specification of the conditions under which multiple group contexts may hinder intergroup discrimination is required.