Can chimpanzees learn the reputation of strangers indirectly by observation? Or are such stable behavioral attributions made exclusively by first-person interactions? To address this question, we let seven chimpanzees observe unfamiliar humans either consistently give (generous donor) or refuse to give (selfish donor) food to a familiar human recipient (Experiments 1 and 2) and a conspecific (Experiment 3). While chimpanzees did not initially prefer to beg for food from the generous donor (Experiment 1), after continued opportunities to observe the same behavioral exchanges, four chimpanzees developed a preference for gesturing to the generous donor (Experiment 2), and transferred this preference to novel unfamiliar donor pairs, significantly preferring to beg from the novel generous donors on the first opportunity to do so. In Experiment 3, four chimpanzees observed novel selfish and generous acts directed toward other chimpanzees by human experimenters. During the first half of testing, three chimpanzees exhibited a preference for the novel generous donor on the first trial. These results demonstrate that chimpanzees can infer the reputation of strangers by eavesdropping on third-party interactions.

Previous research has shown that many social animals follow the gaze of other individuals. However, knowledge about how this skill differs between species and whether it shows a relationship with genetic distance from humans is still fragmentary. In the present study of gaze following in great apes, we manipulated the nature of a visual obstruction and the presence/absence of a target. We found that bonobos, chimpanzees, and gorillas followed gaze significantly more often when the obstruction had a window than when it did not, just as human infants do. Additionally, bonobos and chimpanzees looked at the experimenter's side of a windowless obstruction more often than the other species. Moreover, bonobos produced more double looks when the barrier was opaque than when it had a window, indicating an understanding of what other individuals see. The most distant human relatives studied, orangutans, showed few signs of understanding what another individual saw. Instead, they were attracted to the target's location by the target's presence, but not by the experimenter's gaze. Great apes' perspective-taking skills seem to have increased in the evolutionary lineage leading to bonobos, chimpanzees, and humans.

The prevalence of cooperation among humans is puzzling because cooperators can be exploited by free riders. Peer punishment has been suggested as a solution to this puzzle, but cumulating evidence questions its robustness in sustaining cooperation. Amongst others, punishment fails when it is not powerful enough, or when it elicits counter-punishment. Existing research, however, has ignored that the distribution of punishment power can be the result of social interactions. We introduce a novel experiment in which individuals can transfer punishment power to others. We find that while decentralised peer punishment fails to overcome free riding, the voluntary transfer of punishment power enables groups to sustain cooperation. This is achieved by non-punishing cooperators empowering those who are willing to punish in the interest of the group. Our results show how voluntary power centralisation can efficiently sustain cooperation, which could explain why hierarchical power structures are widespread among animals and humans.