Grooming might be a resource that is offered in exchange for some benefit (e.g. access to a feeding site or coalitionary support) or it might be a mechanism for building and servicing social relationships, whose function, in turn, is to facilitate the exchange of resources and services. Bi-directional (or simultaneous mutual) grooming is unusually common among chimpanzees (though rare in other primates) and we suggest that this might be because it is an especially strong indicator of social bonding. Because the bonding role of bi-directional grooming offers substantially different predictions from the interpretation offered by the models based on reciprocal altruism (RA), we use a critical tests methodology (i.e. tests that unequivocally support one hypothesis at the expense of the other) to differentiate between the bonding and RA hypotheses. We use data on the dynamics of grooming interactions from a captive group of chimpanzees (Pan troglodytes) to show that dominant individuals tolerated the individuals with whom they performed bi-directional grooming more than they did those who typically provided them unidirectional grooming. Dominants rejected and terminated grooming sessions more often with the individuals who provided them with mostly unidirectional grooming than with those with whom they groomed bi-directionally. In addition, animals engaged in bi-directional grooming more often with both relatives and those with whom they were often in proximity. These results support the bonding model of mutually reciprocated grooming at the expense of the RA model, and suggest that, at least in chimpanzees, simultaneous mutual grooming may play a particularly important role in social bonding.