Two experiments are reported which test Rokeach's belief congruence theory (BCT) against predictions from Tajfel's social identity theory (SIT). It is the basic assumption of BCT that discrimination of out-group members is caused by perceived attitudinal dissimilarity between in-group and out-group members. Since, according to SIT, discrimination serves the function of establishing a positive differentiation between in-group and out-group, there should be more discrimination against similar rather than dissimilar out-groups. In the first of two experiments using the minimal group paradigm, attitudinal similarity was manipulated among individuals. This led to more liking for other similar rather than dissimilar individuals, regardless of group membership. On the other hand, both attitudinal similarity and group membership affected reward distribution among members of in-group and out-group: a similar out-group member was discriminated against compared with a similar in-group member, but favoured above a dissimilar one. Manipulating intergroup attitudinal similarity instead of interpersonal attitudinal similarity in the second experiment led, with regard to the allocation of rewards between an in-group and an out-group member, to even more discrimination against members of the similar rather than of the dissimilar out-group. Implications of these findings for BCT and SIT are discussed.