A normative conflict arises when multiple plausible rules exist, specifying how one ought to behave in a given situation. In such cases, enforcing one normative rule can lead to a sequence of mutual retaliatory sanctions, which we refer to as a feud. We investigate the hypothesis that normative conflict enhances the likelihood of a feud in a public-good experiment. Normative conflict in our experiment arises from the fact that individuals derive different benefits from the public good. We find that punishment is much more likely to trigger counter-punishment and start a feud when there is a normative conflict, than it is in a setting in which all individuals derive the same benefit from the public good and no such conflict exists. While the possibility of a feud sustains cooperation, the cost of feuding fully offsets the efficiency gains from increased cooperation. Our results point to an important limitation of self-enforcement mechanisms.