Humans are uniquely cooperative and form crucial short- and long-term social bonds between individuals that ultimately shape human societies. The need for such intense cooperation may have provided a particularly powerful selection pressure on the emotional and communicative behaviours regulating cooperative processes, such as guilt. Guilt is a social, other-oriented moral emotion that promotes relationship repair and pro-sociality. For example, people can be more lenient towards wrongdoers who display guilt than towards those who do not. Here, we examined the social consequences of guilt in a novel experimental setting with pairs of friends differing in relationship quality. Pairs of participants took part in a cooperative game with a mutual goal. We then induced guilt in one of the participants and informed the other participant of their partner’s wrongdoing. We examined the outcome using a dictator game to see how they split a joint reward. We found that guilty people were motivated to repair wrongdoing regardless of friendship. Observing guilt in others led to a punishment effect and a victim of wrongdoing punished close friends who appeared guilty more so than acquaintances. We suggest, therefore, that guilt has a stronger function between close friends as the costs of relationship breakdown are greater. Relationship context, therefore, is crucial to the functional relevance of moral emotions.

Facial expression. The stimulus videos of Player 1 were coded using the Facial Action Coding System

Dictator game - reward split

In the dictator game presented to Player 1 (Fig. 1), players were forced to make an uneven split and attribute at least £1 more to themselves or to their partner. Therefore the dependent variable was the overall proportion of coins given to Player 2 (of the 15 available: percentage < 0.5: more money retained for self; percentage > 0.5: more money to Player 2).

In conditions 1 and 2, Player 2 was presented with the reward money already split (10 to Player 1, 5 to Player 2). Here, the dependent variable was the magnitude of change from the original split to the final split:  10-x15 , with x  being the final number of coins given to Player 1. In condition 3, Player 2 was presented with an un-split reward and the dependent variable was calculated as for Player 1.