Honor killings, which occur when women are perceived to have broken purity norms and bring “dishonor” to family, pose profound moral and public health problems and deep sociological puzzles. Given the immense cost of having murdered one’s own daughter, niece, or cousin, why do families do this? And, conversely, given its potentially tragic consequences, why does norm breaking happen at all? Drawing on detailed accounts of honor killings, we characterize the key actors, actions, and plausible incentives, and create a model that captures its key elements. Our model analytically distinguishes honor norms from incidences of killings and yields novel and sharp predictions for both outcomes. It predicts that honor norms are stronger the lower the costs to punishing deviators, the higher the community-imposed costs from not punishing the deviators, and the stronger the community closure. Moreover, it counterintuitively predicts that honor killings occur most frequently when honor norms are contested; not when they are strictest. Analyzing a large-scale dataset (25 countries, 26,000 individuals) on Muslim’s attitudes towards honor killings, we find support for the first three hypotheses. We also find support for the counterintuitive prediction on the incidences of honor killings, building a unique a dataset from Turkey. Our work advances the theoretical understanding of honor killings and provides one of the most comprehensive empirical analyses of the factors that increase or decrease it.