Abstract: We explore how herd composition and management factors correlate with frequencies of social interactions in horse groups. Since the welfare of horses correlates with low aggression levels and social contact opportunities, information of this kind is important. The data are a collection of records of social interactions of 426 Icelandic horses in 20 groups of at least eight horses. The complexities and limitations of the data prohibit useful statistical modelling so the results are presented descriptively. Interesting and informative patterns emerge which can be of use both in management and in future studies. Of special interest are the low levels of agonistic behaviours in breeding groups where one stallion was present. The horses were less agonistic when in groups with young foals and where group membership was stable. Unfamiliar yearlings in peer groups were especially aggressive. Allogrooming was most frequent in groups with relatively more young horses and in unstable and small groups. Interestingly, the horses allogroomed more if they had few preferred allogrooming partners. The findings show that composition (age/sex) and stability of groups are of great importance with respect to aggression levels and opportunities for establishing bonds.

Abstract: Social bond analysis is of major importance for the evaluation of social relationships in group housed horses. However, in equine behaviour literature, studies on social bond analysis are inconsistent. Mutual grooming (horses standing side by side and gently nipping, nuzzling, or rubbing each other), affiliative approaches (horses approaching each other and staying within one body length), and measurements of spatial proximity (horses standing with body contact or within two horse-lengths) are commonly used. In the present study, we assessed which of the three parameters is most suitable for social bond analysis in horses, and whether social bonds are affected by individual and group factors. We observed social behaviour and spatial proximity in 145 feral horses, five groups of Przewalskiâ&#65533;&#65533;s horses (N = 36), and six groups of feral horses (N = 109) for 15 h per group, on three days within one week. We found grooming, friendly approaches, and spatial proximity to be robust parameters, as their correlation was affected only by the animalsâ&#65533;&#65533; sex (GLMM: N = 145, SE = 0.001, t = â&#65533;&#65533;2.7, p = 0.008) and the group size (GLMM: N = 145, SE < 0.001, t = 4.255, p < 0.001), but not by the horse breed, the aggression ratio, the social rank, the group, the group composition, and the individuals themselves. Our results show a trend for a correspondence between all three parameters (GLMM: N = 145, SE = 0.004, t = 1.95, p = 0.053), a strong correspondence between mutual grooming and friendly approaches (GLMM: N = 145, SE = 0.021, t = 3.922, p < 0.001), and a weak correspondence between mutual grooming and spatial proximity (GLMM: N = 145, SE = 0.04, t = 1.15, p = 0.25). We therefore suggest either using a combination of the proactive behaviour counts mutual grooming and friendly approaches, or using measurements of close spatial proximity, for the analysis of social bonds in horses within a limited time frame.