Equine Social Behaviour: Love, War and Tolerance

Simple Summary Horses are highly social animals that preferably live in stable social groups and form long-term affiliative bonds. However, although their need for social interaction has not changed with domestication, domestic horses are often housed in individual stables with limited social contact with other horses or in group housing with regular changes in their group composition. Thus, this review aims to provide an overview of social ethograms to facilitate the inclusion of social behaviour in equine welfare assessment. A literature review yielded 27 papers that studied equine adult social behaviour using a well-defined ethogram. Social interactions were observed in 851 horses living in groups of 9.1 (mean +/− 6.8 s.d., range: 2–33) horses. A total of 40 (mean: 12.8/paper, range: 2–23) social behaviours were described, of which 60% (24/40) were agonistic, 30% (12/40) affiliative, 7.5% (3/40) investigative and 2.5% (1/40) neutral. The 27 papers focused predominantly on socio-negative interactions by including 67.7% agonistic and only 26% affiliative, 5.1% investigative and 1.2% neutral social behaviours in their research. The strong emphasis on agonistic behaviour contrasts sharply with the rarity of agonistic behaviour in stable horse groups and the well-established importance of affiliative interactions for equine welfare. Therefore, to advance the assessment of horses’ welfare, the ethogram needs to be refined to reflect the nuanced and complex equine social behaviour better and consider more affiliative and also ambivalent and socially tolerant interactions. Abstract Sociality is an ethological need of horses that remained unchanged by domestication. Accordingly, it is essential to include horses’ social behavioural requirements and the opportunity to establish stable affiliative bonds in equine management systems and welfare assessment. Thus, this systematic review aims to provide an up-to-date analysis of equine intraspecific social ethograms. A literature review yielded 27 papers that met the inclusion criteria by studying adult (≥2 years) equine social behaviour with conspecifics using a well-defined ethogram. Social interactions were observed in 851 horses: 320 (semi-)feral free-ranging, 62 enclosed (semi-)feral and 469 domesticated, living in groups averaging 9.1 (mean +/− 6.8 s.d., range: 2–33) horses. The ethograms detailed in these 27 studies included a total of 40 (mean: 12.8/paper, range: 2–23) social behaviours, of which 60% (24/40) were agonistic, 30% (12/40) affiliative, 7.5% (3/40) investigative and 2.5% (1/40) neutral. The 27 publications included 67.7% agonistic and only 26% affiliative, 5.1% investigative and 1.2% neutral social behaviours in their methodology, thus focusing predominantly on socio-negative interactions. The strong emphasis on agonistic behaviours in equine ethology starkly contrasts with the rare occurrence of agonistic behaviours in stable horse groups and the well-established importance of affiliative interactions for equine welfare. The nuanced and complex equine social behaviour requires refinement of the ethogram with a greater focus on affiliative, ambivalent and indifferent interactions and the role of social tolerance in equine social networks to advance equine welfare assessment.

Horses' social organization is based on a stable, complex dominance hierarchy reflecting resource-holding potential, and a female defence polygyny [4,6,7,11,15,21,22,26,31,32,40]. Equine groups have overlapping home ranges and aggregate, forming multilevel societies (herds) with synchronized daily movement and seasonal migration and stable spatial and hierarchical positioning of the various groups within the herd [17,41]. The social complexity of maintaining long-term affiliative relationships and navigating multilevel societal structures requires the ability to recognize and remember individuals and their relative rank [17,42,43]. Indeed, horses are capable of cross-modal individual recognition using visual, auditory and olfactory cues even after a year's absence and transitive inference of dominance relationships through observation [44][45][46][47][48][49][50][51]. Horses' social cognition is further demonstrated by third-party interventions in agonistic and affiliative dyadic interactions of group members and increased affiliative behaviour after a conflict [34,52,53]. As food-related aggression is not typically relevant in grazers that feed on widely dispersed and undefendable resources, agonistic interactions occur mainly to establish a dominance hierarchy and maintain personal space, in which horses only allow affiliative associates [13,14,39,54]. Dominance typically depends on age, physical characteristics, experience, and length of residency in the herd [13,14,39,55]. The stable composition and hierarchy of (semi-)feral equine groups and the long-term social bonds result in social cohesion and a low frequency of agonistic interactions, most (80%) of which are ritualized and do not involve physical contact [54,56].
Comparisons of the behaviour of feral and domesticated horses indicate that the species-specific social behaviour of horses has remained qualitatively relatively unchanged by domestication [13]; however, the environment of domestic horses has changed dramatically compared to naturalistic conditions. Although management systems that accommodate equine sociality exist, most domestic horses are confined to individual stables with limited contact with conspecifics [12,13,15,22,35,[56][57][58][59][60][61][62][63][64][65][66]. Lack of social contact is thought to be one of the most serious stressors for horses, as evidenced by significant increases in faecal corticosterone metabolites, and it triggers stress-related behaviours and stereotypies such as weaving, cribbing and box-walking in horses kept without adequate opportunities to socialize with conspecifics [12,54,56,58,60,[67][68][69][70][71][72][73][74][75][76][77][78]. Indeed, social contact, specifically the possibility to engage in affiliative behaviours such as allogrooming, which has been shown to lower the heart rate, has been identified as an ethological need and essential for equine welfare [15,22,56,79]. In addition to the limitation in social contact, managed horses also do not have the opportunity to choose their group affiliation. They are faced with frequent changes in group composition and social companionship, which limits their opportunities to establish long-term social bonds and a stable hierarchy, resulting in higher aggression and frequency of agonistic encounters [56,58,63,80]. The space restrictions inherent to domestic conditions, which limit the opportunities for subordinate individuals to escape or provide dominant conspecifics with their required individual distance, further compound the social challenge [56,58,62,80,81]. As horses do not adapt to repeated regrouping and a stable hierarchy is achieved only after 2-3 months [56,58], the common disregard of equine social group dynamics in equine husbandry poses a significant welfare concern [12,[56][57][58]61,62,81].
Thus, it is essential to include horses' social behavioural needs and the opportunity to establish stable affiliative bonds in equine management systems and welfare assessment [82]. However, to facilitate evidence-based optimization of equine husbandry practices and their evaluation, the influence of different environmental and management factors on equine social interactions needs to be further elucidated [83]. As differences in the sampled behaviours currently complicate the comparison of equine social behavioural studies, this systematic review aims to analyse the literature on equine social ethograms to distil a well-defined social behavioural repertoire as a basis for further studies.

Data Sources and Searches
This review was carried out according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [84]. Scientific peer-reviewed articles focused on adult (≥2 years) equine intraspecific social behaviour were identified through a systematic search in the PubMed (National Institutes of Health. PubMed [Database]. Bethesda, MD, USA: National Library of Medicine; https://pubmed.ncbi.nlm.nih.gov, accessed on 25 January 2022) and Scopus (Elsevier, Amsterdam, The Netherlands; https: //www.scopus.com) electronic databases on 25 January 2022. The search was conducted by combining the search strings ("horse" OR "equine" OR "equus") in the title and ("social" OR "ethogram" OR "agonistic" OR "affiliative") in the title or abstract with the Boolean operator "AND", with no restriction on publication date. The following exclusion criteria were set a priori ( Figure 1): (a) non-peer-reviewed publication, dissertation, thesis, review, commentary, or single case report; (b) only a conference/seminar abstract published; (c) the article was not written in English; (d) the study did not include equine intraspecific social behaviour but focused on interspecies interaction or learning behaviour; (d) no ethogram of observed social behaviour was provided; (e) the observations were limited to a specific subset of behaviours (reproductive behaviour, dominance behaviour, mare-foal interaction); or (f) the observed horses were <2 years of age.

Data Extraction and Risk of Bias Assessment
The study selection process was carried out by U.A. and F.J. following the procedure detailed in Figure 1. Any disagreement between the authors on the studies included in the review was resolved during a consensus meeting.
Information on the population, intervention, comparison, outcome and study design (PICOS) was retrieved from the articles, and the risk of bias in selected studies was assessed using a modification of the Evidence Project risk-of-bias tool [85][86][87].

Agonistic submissive
Avoidance/ Withdrawal [14,22,26,30,37,64,90,92] Movement that maintains or increases the distance to an approaching horse (which does not threaten). While making way, the subordinate usually lays its ears back Only head turn away from the initiator [14,30] 8 These three terms are used interchangeably → clarification of the definitions is required.
We propose using avoidance/withdrawal to indicate increase/maintenance of distance to a non-threatening approaching horse, retreat as a reaction to an agonistic approach at the walk or trot, and flight as a rapid increase/maintenance of distance in response to an attack Retreat [27,34,37,41,52,80] One individual immediately moves away from an animal that approaches to within 2 m of it to maintain or increase the distance Abrupt halt or reversal of direction with movement of the head and neck in a rapid sweeping dorsolateral motion away from an apparent threat while the hind legs remain stationary. The forelegs may simultaneously lift off the ground 1 Snapping [5,12,13,22,37,54,55,58,79] Corners of the open mouth are pulled back, showing teeth and gums, making chewing motions. Hindlegs may be slightly bent in a cringing position. Head and neck are extended, the ears are oriented back or laterally An appeasement act delivered to older/higher-ranked animals [55] 9 affiliative (Affiliative) Approach [5,14,27,30,55,64,80,90,93]  Grooming approach [34,41,52] Approach with subsequent mutual grooming 3 Mutual approach [34,41,52] Both animals approach each other Table 3. Cont.

Quantitative Assessment
Formula Reference

Social interactions
Total number of affiliative and agonistic behaviours observed/per horse and or per hour [12][13][14]34,55,56,64,80,89,[92][93][94] Mean number of social interactions per week [58] Social interactions in %: (Number of observations of a behaviour/total number of observed behaviours) × 100 [5,54,55] Aggressiveness, aggressive score Total number of agonistic behaviours observed [27,29,79,90] Count of agonistic acts received and given [55] Aggression rate per group per horse/total number of aggressions per group per horse [27] Activity similarity (Number of observations including A and B)/(Total number of observations of A + total number of observations of B) [22,89] Time percentage when two horses were first neighbours and engaged in the same activity [5] Nearest neighbour Nearest neighbour per activity = Time that individuals were first neighbour to each other when engaged in same activity/time that individuals were first neighbour to each other × 100 [5,89] Frequency of two individuals being observed as "being close" or "being far" [79] Number of observations including A and B/total number of observations of A + total number of observations of B [89,92] Observations of an individual at a specific distance/Total observations of that individual at any distance × 100 [22,54] Total number of observations in which an individual was either the first or second neighbour of a particular one (single link cluster analyses) [5]

Duration Duration of a behaviour/interaction
Time in seconds from start to end [37,90] Dominance Relationships

Ranking index
Number of agonistic encounters won by A against B/total number of agonistic encounters in which A and B were involved [27,29,34,52,55,90,92] Highest rank = individual with least threats possible from individuals below it [5,55] Comparison between the number of threats received by individuals and the number of threats initiated [22,91] [

Discussion
Aiming to provide an up-to-date analysis of equine social ethograms, this systematic review included all original studies of equine social behaviour that detailed the ethogram underlying the reported research. Surprisingly, the equine social ethograms primarily reference the "agonistic ethogram of the equine bachelor band" (includes 23/40 behaviours, is referenced in 15/27 papers [12,13,22,27,34,37,41,52,54,56,58,80,[92][93][94]), which was based on a literature review of equine behaviours and 50 daylight hours of observation of 15 pony stallions (2-21 years old) pastured together in a semi-natural enclosure of 9 acres [37]. While this landmark publication provides an excellent ethogram, it describes interactions in an equine bachelor group, which is neither under (semi-)feral nor domestic conditions the prevalent social group structure and hence may not suffice as a comprehensive behavioural catalogue for horses living in harems or human-managed groups. The strong focus on agonistic behaviours has persisted in equine ethology, with 67.7% of the social behaviours studied in the 27 papers focussing on the socio-negative spectrum and only 26% on affiliative, 5.1% on investigative and 1.2% on neutral behaviours. The rare occurrence of agonistic behaviours in stable horse groups (0.2-1.5 agonistic interactions/h per horse [13,56,79,[95][96][97] and the well-established importance of affiliative interactions for equine welfare [12,15,22,31,54,56,58,60,[67][68][69][70][71][72][73][74][75][76][77][78][79] further emphasize the necessity to expand and diversify the equine social ethogram to include a broader spectrum of behaviours of horses living in different group compositions and environments. The binary division in agonistic and affiliative social interactions as bipolar opposites belies the reality that many relationships are not one-dimensionally positive or negative but more multifaceted and may entail social tolerance, coactivated feelings of positivity and negativity toward a relational partner (ambivalence) or lack affective valence (indifference) [80,90,[98][99][100][101][102][103][104][105][106][107][108]. In human social sciences, the impact of ambivalent relationships on social networks and quality of life is increasingly recognized. Studies have shown them to be prevalent in both personal and professional networks and cause increased stress, blood pressure and detrimental health outcomes [98,109,110]. The social behaviour of horses similarly includes ambivalent interactions and relationships, such as the more frequent but less violent aggressive interactions among preferred associates and the predominant initiation of affiliative interactions by dominant individuals [5,22,26,29,30,56,80,90,107]. In addition, horses show social tolerance (defined as proximity to conspecifics around valuable resources with little or no aggression [108]) depending on space availability and their social experience [5,22,26,29,30,56,80,90,107]. As the current positive-negative dichotomy does not sufficiently reflect the nuanced and complex equine social behaviour, equine ethology can build upon the human social science approach of assessing the valence generated by social interactions along the continuum from negative to positive in combination with the elicited autonomic activation intensity (arousal) to refine equine ethograms [98,99,[110][111][112][113][114].
The quantification of social interactions incorporated by 81.55% of the included papers also assists in assessing dyadic relationships and group dynamics. These quantitative criteria are primarily based on the reported interindividual distance of 2 m to two horse lengths, within which horses only tolerate close affiliates [12,13,27,38,39,54,[115][116][117] and include measurements of spatial proximity between two horses, the number and duration of affiliative or agonistic interactions per hour and recording the nearest neighbour. In addition, recent studies have incorporated social network analysis to examine indirect connections beyond the dyad level and analyse the patterns of individual and group-level social interactions [41,52,[118][119][120][121].
The combination of a more nuanced qualitative assessment of equine social behaviour with quantitative approaches may greatly assist equine welfare assessment and optimization, as poor welfare conditions, such as high population density (< 331 m2 per horse), may reduce equine sociality and skew horses' social behavioural repertoire toward agonistic interactions [80,122,123]. Changes in social behaviour have also been associated with disease in various group-living species ranging from honeybees, zebrafish and mice to calves and humans [121,[124][125][126][127], but the link between social networks and health has not yet been explored for horses. The changes seem species-specific, as mice reduce social interactions, while rhesus macaques and calves increase affiliative interactions with familiar conspecifics [121,128,129]. More detailed studies in sick humans found increased social interactions with familiar support figures but withdrawal from strangers and a strong correlation between pain intensity and interpersonal distance to strangers in patients with lower back pain [121,127,130,131]. Expanding social behavioural research in horses to also include assessment of the effect of acute and chronic disease on social interactions may further advance equine welfare by facilitating early detection, treatment and monitoring of disease and pain.

Conclusions
Horses, as social non-territorial equids that preferably live in stable, hierarchically structured social groups, have developed complex cognitive skills, ritualized communication signals and nuanced social behaviour. In these stable groups, the frequency of agonistic interactions is low under species-appropriate housing and welfare conditions (e.g., adequate enclosure size, stocking density and resource availability). However, our systematic review reveals a strong focus of current social ethograms on socio-negative interactions with 67.7% agonistic and only 26% affiliative, 5.1% investigative and 1.2% neutral social behaviours. The traditional equine ethology approach focusing on univalently negative social interactions does not sufficiently reflect the complexity of equine social behaviour and requires the development of a more refined ethogram, which also considers ambivalent and indifferent interactions and relationships and the role of social tolerance in equine social networks. A standardized comprehensive social ethogram combined with quantification of social interactions and social network analysis would facilitate research into the effect of disease and pain on equine social behaviour and constitute a valuable tool for equine welfare assessment and optimization.