Subjects
Thirty-one privately owned horses and ponies took part in the study, all of them housed at the Aktivstall Mauerbach complex in the Vienna Woods, Austria. There were four groups, each of different social composition.
Group 1, riding horses (N = 10), comprised 4 geldings and 6 mares, aged from 2 to 22 years. They included 4 Warmbloods, 1 Sorraia Mustang, 1 Pryor Mountain Mustang, 2 Quarter Horses, 1 Icelandic horse, and 1 Haflinger. The group was housed in a “Hit Aktivstall”, designed to cover the needs of horses as well as possible. The stabling covered approximately 2.5 hectares, (2500 m2 per horse) and included a rest and sleeping shelter (300 m2, enclosed on three sides, with three open doorways on the eastern side). The horses had 24-h access to grass pasture, straw fodder, and water from an automatic dispenser. An automatic group hay feeder opened for 15 min, 16 times per day. There was also an individual automatic hay feeder and an individual automatic pellet feeder, which were programmed according to each horse’s needs, with individual rations varying between 500 g and 2 kg per day. The feeders automatically portioned and dispensed the hay or pellets when activated by a transponder, worn either on a collar around the horse’s neck, or woven into the horse’s mane. To reach the pellet dispenser, the horses had to walk around a track of approximately 700 m. The stabling area also included three grass pastures, covering a total of 2 hectares, which were open to the horses 24 h a day. The horses shared their living quarters with two female donkeys, but as only one interaction was observed between a horse and the donkeys, the donkeys were not considered in the analysis.
Group 2, mares and foals (N = 8), comprised 5 Mini-Shetland pony mares, 3 with foals at foot. The foals, 2 fillies and 1 colt, were all between 3 and 6 months old at the time of observation, while the mares ranged from 1 to 20 years old. The group was housed on approximately 6400 m2 of grass pasture and woodland (800 m2 per pony) with two shelters, each of 20 m2, which were enclosed on three sides. There was a covered hay station providing ad libitum hay, and water was supplied in large buckets. The area was divided into two grass pastures, a sand enclosure where the hay station was positioned, and an area of woodland which offered shade. The ponies also received approximately 150 g of grain once a day.
Group 3, harem (N = 8), comprised 1 stallion and 7 mares, all Mini-Shetland ponies, aged 3–14 years old. The group was housed on 0.7 hectares of mixed grass pasture and woodland, (970 m2 per pony). Hay was provided in hanging dispensers and nets, as well as in fixed stands. Fresh water was available from a stream, as well as in large buckets. The ponies also received approximately 150 g of grain once a day. There were two shelters, each enclosed on three sides: one of 72 m2, one of 48 m2.
The stallion was removed from the group for management purposes the evening before the final observation period; however, as the absence of a stallion has been shown to slightly increase social interaction in mares (Sigurjónsdóttir et al. 2003), we continued to collect data on the mares.
Group 4, stallions (N = 5), comprised 3 mature stallions and 2 yearling colts, all Mini-Shetlands, aged from 1 to 20 years old. The group was housed on a 2-hectare grass pasture (4000 m2 per pony), with shade provided by trees along one side and a small grove in the centre. There were two shelters each measuring 48 m2 and enclosed on three sides. The ponies shared this pasture with nine sheep, but there was very little contact between the sheep and the ponies. The grass was so plentiful that additional hay was not considered necessary, but the ponies did receive approximately 150 g of grain once a day. Water was supplied in buckets and automatic drinkers.
Observation
Groups 1, 2 and 3 were observed for 12 h each, and group 4 (which had fewer individuals) for 10 h, between July 4th and July 21st, 2017. Observation was carried out between 10 a.m. and 6 p.m., in periods of between 1.5 and 2.5 h. The observation periods for each group were randomised across the times of day, and no group was observed more than once on any 1 day. Observer 1 (KF) recorded each observation verbally on the voice recorder of a Samsung A3 mobile phone, and Volunteer 1 made video recordings of the observations on an iPhone6 as a backup and cross reference. The data from the recordings were transferred to an Excel 2013 sheet on a Packard Bell “Easy Note” laptop immediately after the observation period. The recordings and data sheets were then backed up on USB sticks. Volunteer 2 transcribed the voice recordings into text.
All the horses and ponies were already acclimatised to the presence of people, and the observation points were based between 10 and 30 m away from each group, although the precise distance depended on the movement of the horses/ponies, and whether the observers had to move in closer to see the details of an interaction. There was no point at which the horses and ponies appeared to be disturbed by the presence of the observers. When horses or ponies spontaneously approached the observers, they were gently encouraged to move away and return to other members of their group.
Volunteer 1 simply recorded the video and did not make any rating or comment and so could not be used for an inter-observer rating. Therefore, a sample of 10% of the videos was shown to volunteer 3, who made an independent assessment of the behaviour. There was a high level of agreement between observer 1 and volunteer 3, with a Cohen’s Kappa coefficient of k = 0.932.
Data collection
Affiliative interactions: approaches and interactions
We defined affiliative approaches by considering the behaviour of the approached horse. If the approached horse retreated more than two metres from the approaching horse, the approach was considered non-affiliative. If the approached horse did not move, moved towards the approaching horse, or moved less than 2 m to make room for the approaching horse, the approach was considered affiliative, as described by Schneider and Krüger (2012). Affiliative interactions typically included allo-grooming, swishing flies for each other, and standing in a proximity of less than 2 m for at least 15 s while grazing or resting. The side placement of equine eyes makes it easy to see whether one eye or the other is being preferred in any interaction. One horse approaching another with its left eye to the approached horse scored one point under “affiliative left” for the approach, or “affiliative right” if the approach was with the right eye. A further point was allocated both to the approaching and to the approached horse if the approach led to allo-grooming, nose to tail fly swishing, or just relaxing and standing within 2 m of each other for at least 15 s, according to the side of the interaction.
If a pair of horses switched sides, further points were allocated to each horse accordingly, and if a pair positioned themselves side by side, affiliative points were awarded to each horse according to the eye used for viewing the conspecific.
Interactions where a lateral choice could not be established (for example, a head-on approach) were not scored for the sensory laterality data but were included in the rank dominance calculations if appropriate. As head-on approaches only occurred in agonistic encounters, these were excluded from the affiliative laterality analysis.
Rank dominance observations
Rank dominance points were awarded based on retreats by either the approaching or the approached horse. The retreating horse was allocated one point under “lose” for a retreat, and the horse that was retreated from was awarded a point under “win”. Non-affiliative interactions were defined as approaches with the ears pinned back and the nose extended, retreats, threats to bite or kick, bites, kicks and chases as described by McDonnell and Haviland 1995, and McDonnell 2003.
Approaches and interactions were scored under the categories (1) affiliative left, (2) affiliative right, (3) win, and (4) lose.
Data and statistical analysis
Affiliative laterality index (ALI): an ALI was calculated for each horse, using the standard formula of (right eye score − left eye score)/total lateral interactions, as used by Austin and Rogers (2012). This gives scores between − 1 and + 1 with negative scores showing a left bias, and positive scores a right bias.
Social index (SI) an index was calculated for each horse using all interactions, where the SI = (affiliative interactions − non-affiliative interactions)/total interactions. This gives a number from − 1 to + 1, with positive numbers indicating relatively more affiliative behaviour.
Dominance Index An average dominance index (ADI) was calculated as recommended by Hemelrijk et al. (2005). ADI = 1/N ∑j[xij/(xij + xji)]; N the number of interaction partners, xij the number of times the individual i won against conspecific j, xji the number of times individual i lost against conspecific j. ADI values range from 0 to 1, with a high value indicating a high rank in the group. Individuals were counted as a winner when their interaction partner retreated one step or more. Pairs that were not involved in an encounter with each other were excluded from the analysis.
The R Studio and R commander (version 3.4.1, 2017) were used to analyse the data and compare the laterality indices across groups, gender, rank, and social index. Figures and tables were compiled in Microsoft Excel 2016.
The ALI was not normally distributed (Shapiro Wilk test). We therefore continued using non-parametric tests. We considered the numbers of literately indices to the left and the right for each individual and used a binomial test to analyse the level of bias on population and individual levels. Multivariate factor analysis [GLM, formula = ALI ~ age + phenotype + gender + group + rank + social index, family = Gaussian (identity)] was used to compare the four groups with respect to the variables of phenotype and group composition, and to compare the variables of age, gender, social index, and rank within the groups. All the tests used were two sided and the significance level was set at 0.05.