Much of the debate about reciprocity in humans and other primates hinges on proximate mechanisms, or more precisely, the contingency of one service on another. While there is good evidence for long-term statistical contingencies of services given and received in primates, results for short-term behavioral contingencies are mixed. Indeed, as we show here controlled experiments using artificial tasks and explicit turn-taking were unlikely to find short-term effects. We therefore used more naturalistic experiments to test for short-term contingencies of grooming on food sharing and vice versa in one group of chimpanzees and two groups of bonobos. Overall, we found significant effects of grooming on food sharing and vice versa, however, in the chimpanzees these effects disappeared when controlling for long-term characteristics of the dyad including services exchanged over the whole study period. In the bonobos, short-term contingencies remained significant which was likely a consequence of considerable tension surrounding monopolizable food resulting in higher rates of grooming and other affiliative behaviors around sharing sessions. These results are consistent with the fact that previous evidence for short-term contingency often involved grooming and that long-term contingency is more commonly observed in primates. We propose that long-term contingency is proximately regulated by a ‘relationship score’ computed through a tally of past interactions which tend to outweigh recent single events. We therefore suggest that future research into the proximate mechanisms of reciprocity should trace the development of such a score by focusing on newly formed dyads with no history of interactions.