Objective. People have the intuition that hunger undermines social cooperation, but experimental tests of this have often produced null results. One possible explanation is that the experimental tasks used are not rich enough to capture the diverse pathways by which social cooperation can be sustained or break down in real life. We studied the effects of hunger on cooperation in two tasks of differential interaction richness. Methods. We manipulated hunger by asking participants to eat, or refrain from eating, breakfast. Participants in experiment 1 (n = 106) played a one-shot Ultimatum Game. Participants in experiment 2 (n = 264) played twenty rounds of a Public Goods Game in the same groups of four, ten rounds with the possibility of punishing other group members, and ten without. Results. In experiment 1, skipping breakfast had no significant effects on either amounts proposed or minimum acceptable offers. In experiment 2, there were multiple different significant effects of the manipulation. No-breakfast participants were more generous in the first round of the game without punishment, and in subsequent rounds, were more influenced by what other group members had done the round before. In the punishment game, no-breakfast participants were also less likely to punish their group-mates than breakfast participants. Consequently, the possibility of punishment was less effective in increasing group cooperation levels in no-breakfast groups. Conclusion. Replicating earlier findings, we found a null effect of hunger on cooperation in a one-shot Ultimatum Game. However, in our richer Public Goods Game, the dynamics of cooperation differed with hunger, in subtle ways not simply classifiable as hungry participants being ‘more’ or ‘less’ cooperative overall.