Abstract: The sexual selection hypothesis explains infanticide by males in many mammals. In our 11-year study, we investigated this hypothesis in a herd of Przewalski's horses where we had witnessed infanticidal attacks. Infanticide was highly conditional and not simply linked to takeovers. Attacks occurred in only five of 39 cases following a takeover, and DNA paternity revealed that, although infanticidal stallions were not the genetic fathers in four cases out of five, stallions present at birth did not significantly attempt to kill unrelated foals. Infanticide did not reduce birth intervals; only in one case out of five was the infanticidal stallion, the father of the next foal; mothers whose foals were attacked subsequently avoided associating with infanticidal stallions. Therefore, evidence for the sexual selection hypothesis was weak. The “human disturbance” hypothesis received some support, as only zoo bred stallions which grew up in unnatural social groups attacked foals of mares which were pregnant during takeovers.