This paper provides evidence that free riders are heavily punished even if punishment is costly and does not provide any material benefits for the punisher. The more free riders negatively deviate from the group standard the more they are punished. As a consequence, the existence of an opportunity for costly punishment causes a large increase in cooperation levels because potential free riders face a credible threat. We show, in particular, that in the presence of a costly punishment opportunity almost complete cooperation can be achieved and maintained although, under the standard assumptions of rationality and selfishness, there should be no cooperation at all. We also show that free riding causes strong negative emotions among cooperators. The intensity of these emotions is the stronger the more the free riders deviate from the group standard. Our results provide, therefore, support for the hypothesis that emotions are guarantors of credible threats. The paper presents an experiment with and without punishment opportunities in public goods games. In the treatment without punishment opportunities, complete free riding is a dominant strategy. In the treatment with punishment opportunities, punishing is costly for the punisher. Therefore, purely selfish subjects will never punish in a one-shot context. This means that if there are only selfish subjects, as is commonly assumed in economics, the treatment with punishment opportunities should generate the same contribution behavior as the treatment without such opportunities. However, the results show that in the no-punishment condition contributions converge to very low levels, while in the punishment condition, average contribution rates between 50 and 95 percent of the endowment can be maintained. In the punishment condition, subjects can implicitly coordinate on a common group standard, leading to almost full cooperation. The results suggest that punishment opportunities significantly increase cooperation levels and that emotions play a key role in this process. The paper also shows that punishment opportunities can lead to significant payoff gains for groups, even though they may initially cause a relative payoff loss. The results indicate that the presence of punishment opportunities eventually leads to pecuniary efficiency gains. The paper concludes that emotions and social norms are important in maintaining cooperation and that punishment opportunities can significantly influence behavior in social interactions.This paper provides evidence that free riders are heavily punished even if punishment is costly and does not provide any material benefits for the punisher. The more free riders negatively deviate from the group standard the more they are punished. As a consequence, the existence of an opportunity for costly punishment causes a large increase in cooperation levels because potential free riders face a credible threat. We show, in particular, that in the presence of a costly punishment opportunity almost complete cooperation can be achieved and maintained although, under the standard assumptions of rationality and selfishness, there should be no cooperation at all. We also show that free riding causes strong negative emotions among cooperators. The intensity of these emotions is the stronger the more the free riders deviate from the group standard. Our results provide, therefore, support for the hypothesis that emotions are guarantors of credible threats. The paper presents an experiment with and without punishment opportunities in public goods games. In the treatment without punishment opportunities, complete free riding is a dominant strategy. In the treatment with punishment opportunities, punishing is costly for the punisher. Therefore, purely selfish subjects will never punish in a one-shot context. This means that if there are only selfish subjects, as is commonly assumed in economics, the treatment with punishment opportunities should generate the same contribution behavior as the treatment without such opportunities. However, the results show that in the no-punishment condition contributions converge to very low levels, while in the punishment condition, average contribution rates between 50 and 95 percent of the endowment can be maintained. In the punishment condition, subjects can implicitly coordinate on a common group standard, leading to almost full cooperation. The results suggest that punishment opportunities significantly increase cooperation levels and that emotions play a key role in this process. The paper also shows that punishment opportunities can lead to significant payoff gains for groups, even though they may initially cause a relative payoff loss. The results indicate that the presence of punishment opportunities eventually leads to pecuniary efficiency gains. The paper concludes that emotions and social norms are important in maintaining cooperation and that punishment opportunities can significantly influence behavior in social interactions.