Abstract: Agonistic intervention behavior was observed in captive groups of chimpanzees (Pan troglodytes), rhesus monkeys (Macaca mulatta), and stumptail monkeys (M. arctoides). Reciprocity correlations of interventions were determined while removing from the data the effects of several symmetrical relationship characteristics, that is, matrillineal kinship, proximity relations, and same-sex combination. It was considered likely that if significant reciprocity persisted after controlling for these characteristics, the reciprocity was based on cognitive mechanisms. Statistical significance was tested by means of recently developed matrix permutation procedures. All three species exhibited significant reciprocity with regard to beneficial interventions, even after controlling for symmetrical traits. Harmful interventions were, however, reciprocal among chimpanzees only. This species showed a “revenge system”, that is, if A often intervened against B, B did the same to A. In contrast, both macaque species showed significantly inversed reciprocity in their harmful interventions: if A often intervened against B, B rarely intervened against A. Further analysis indicates that the strict hierarchy of macaques prevents them from achieving complete reciprocity. Compared to chimpanzees, macaques rarely intervene against higher ranking group members. The observed contrast can be partially explained on the basis of differences in available space, as indicated by a comparison of indoor and outdoor living conditions for the chimpanzee colony. Yet, even when such spatial factors are taken into account, substantial behavior differences between chimpanzees and macaques remain.