Third-party punishment by preverbal infants

Abstract

Third-party punishment of antisocial others is unique to humans and seems to be universal across cultures. However, its emergence in ontogeny remains unknown. We developed a participatory cognitive paradigm using gaze-contingency techniques, in which infants can use their gaze to affect agents displayed on a monitor. In this paradigm, fixation on an agent triggers the event of a stone crushing the agent. Throughout five experiments (total N = 120), we show that eight-month-old infants punished antisocial others. Specifically, infants increased their selective looks at the aggressor after watching aggressive interactions. Additionally, three control experiments excluded alternative interpretations of their selective gaze, suggesting that punishment-related decision-making influenced looking behaviour. These findings indicate that a disposition for third-party punishment of antisocial others emerges in early infancy and emphasize the importance of third-party punishment for human cooperation. This behavioural tendency may be a human trait acquired over the course of evolution.

Fig. 1. Schema of Experiments 1 to 4.

a, Gaze-contingent events. b, Sequence of experiments. Baby silhouette from openclipart.org.

Fig. 2. Results of the sensitivity analysis for BF_incl for each effect across experiments (Experiments 1 to 5).

The natural logarithm of BF_incl for each effect (small dots) is plotted as a function of the scaling factor of the prior for the effect size. The natural logarithms of BF_incl reported in the main text are shown as large dots. The thick coloured lines are fitted using LOESS smoothing regression. When the logarithm of BF_incl of the effect is a positive value, a model with the effect is, on average, supported across all candidate models. When the logarithm of BF_incl of the effect is a negative value, a model without the effect is, on average, supported across all candidate models. The dashed and dotted horizontal lines show conventional evidence thresholds for interpreting BFs. The interpretations of the logarithms of BFs are shown on the right y axis.

Fig. 3. Results from Experiments 1 to 5.

a–e, Results of selective looks in the pretest and posttest across experiments (Experiments 1 to 5). The dot represents the proportion of infants who selectively look toward an aggressor (or a causer) for each trial. The solid lines represent the estimated probabilities for selective looks at an aggressor (or a causer), which are based on the full model. The shaded regions represent 95% CIs.