Abstract: tEye and limb preferences were scored in the closest undomesticated relative of Equuscaballus using the same methods as used previously to study laterality in feral horses.Observations were made of 33 Przewalski horses (Equus ferus przewalskii) (male N = 20,female N = 13) living under natural social conditions on a large reserve in France. Signifi-cant left-eye/side biases were found in agonistic interactions within harem bands (M ± SEbias to left 58% ± 0.01 for threats, P < 0.001; 68% ± 0.05 for attacks; P < 0.001) and in stallionfights (threats, 52% ± 0.01 left, P < 0.001; attacks, 63% ± 0.02 left, P < 0.001): as many as 80%of the horses were significantly lateralized in attack responses within harem bands. Lat-erality of vigilance was measured as lifting up the head from grazing and turning it to theleft or right side: a directional bias to the left was found (M ± SE 53% ± 0.02 left, P < 0.001).Side bias in reactivity was calculated as the percent of head lifts above the level of thewithers on the left or right side and this was also left side biased (M ± SE 73% ± 0.03 left,P < 0.001). These results indicate right-hemisphere specialization for control of aggressionand responses to novelty. The left bias in attack scores within harem bands was strongerin males than females (P = 0.024) and in immature than adult horses (P = 0.032). Immaturehorses were also more strongly lateralized than adults in vigilance scores (P = 0.022), whichmay suggest that experience reduces these side biases. Our results show that Przewalskihorses exhibit left eye preferences, as do feral horses, and do so even more strongly thanferal horses. Considering feral and Przewalski horses together, we deduce that ancestralhorses had similar lateral biases. Also similar to feral horses, the Przewalski horses showedno significant forelimb preference at the group level or in the majority of horses at theindividual level, confirming the hypothesis that previously reported limb preferences indomestic breeds are entrained or generated by breed-specific selection.

Abstract: Hawk-dove games have been extensively used to predict the conditions under which group-living animals should defend their resources against potential usurpers. Typically, game-theoretic models on aggression consider that resource defense may entail energetic and injury costs. However, intruders may also take advantage of owners who are busy fighting to sneak access to unguarded resources, imposing thereby an additional cost on the use of the escalated hawk strategy. In this article we modify the two-strategy hawk-dove game into a three-strategy hawk-dove-sneaker game that incorporates a distraction-sneaking tactic, allowing us to explore its consequences on the expected level of aggression within groups. Our model predicts a lower proportion of hawks and hence lower frequencies of aggressive interactions within groups than do previous two-strategy hawk-dove games. The extent to which distraction sneakers decrease the frequency of aggression within groups, however, depends on whether they search only for opportunities to join resources uncovered by other group members or for both unchallenged resources and opportunities to usurp.