Foraging efficiency, social status and body condition in group-living horses and ponies

Animals and ethical statement

The work was approved by the University of Bristol Animal Welfare and Ethical Review Board (University Investigation Number UB/10/049) and all methods were carried out in accordance with relevant guidelines.

The study sample was drawn from a population of outdoor, group-living horses based at Redwings Horse Sanctuary (UK), that had been living together for at least 3 months and had established social relationships. All of the individual animals were managed similarly, fed forage from identical sources, lived in outdoor environments and were not ridden, meaning that structured exercise could be removed as a potential confounding factor. Herds that included pregnant or lactating mares were not considered for the study. Twenty study herds were selected randomly from all remaining suitable herds within the sampling frame.

The policy of the sanctuary was to house horses in relatively compatible groups with shared characteristics. Thus, larger horses were housed in separate herds from smaller ponies, all stallions were housed in one ‘bachelor’ herd, while youngsters were also housed together, with the few horses under 1 year of age (three individuals) accompanied by older ‘nanny’ mares. Herd size was 2–10 (mean 6 ± 0.56 individuals). 116 individuals (84 ponies of height <148 cm, and 32 horses of height ≥148 cm) from within these herds were observed between 2 December, 2013 and 23 January, 2014. Ages ranged from 5 months to 32 years (11.83 ± 0.63 years). Breeds were native ponies (51.72%), native cobs (17.24%), lightweight horses (12.07%), heavy horses (5.17%), sports horse breeds (5.17%) and other (8.62%).

Study period and horse management

The winter months were chosen for observation as natural food resources were at their minimum and therefore food based social interactions were likely at their highest due to the close proximity of individuals. All horses lived in an outdoor paddock environment for 24 h a day and were fed from circular hay feeders provided at a fixed ratio of feeder space (30 cm) per animal. Horses were fed twice daily with fresh hay replenished once at the start of morning observation (between 08:00 and 09:00) and once at the start of afternoon observation (between 11.30 and 13:00). Any uneaten hay remained in the hay feeder throughout the day. Twelve study horses received additional supplementary feed from a bucket once a day, and this was recorded as a potential confounder.

Time spent foraging

Each study herd was observed for 6 h to assess overall time spent foraging, and interruptions occurring during foraging bouts, once during a 3 h morning session (08:00–09:00 until 11:00–12:00) and once during a 3 h afternoon session (11:30–13:00 until 14:30–16:00) on a different day within the same week, by a single trained observer. Due to the time of year, these times were chosen based on daylight hours.

Time spent foraging was recorded using scan sampling at 5 min intervals throughout each 3-h observation period. A random number generator was used to determine the order in which individuals were observed. Once this order was determined, all individuals were observed in sequence, in 5-min intervals. At each interval, it was recorded which individuals were foraging and which were not. Foraging was defined as the horse ingesting either hay or grass, with intermittent periods of the head down ingesting forage and the head up chewing this forage material. The horse could be foraging from either the hay feeder or eating grass (although the latter was rare as there was little grass available). The percentage of time spent foraging was then calculated based on the number of intervals that each individual was foraging within the full 6 h of observation per herd.

Alongside this, continuous 5 min focal animal observations were scheduled for each horse during each 3 h recording period. Each individual animal was independently observed for at least 20 min (4 × 5-min) in total. These observations were predominantly used to record foraging interruptions and social interactions (as detailed in “Foraging Efficiency—Duration and Frequency of Foraging Interruptions” and “Dominance Rank” below), however they were also used to more accurately estimate the total foraging time for each individual. If an individual was not foraging for more than 1 min during the 5-min observation period, it was considered to have stopped foraging. The number of minutes it had stopped foraging for were then subtracted from the total 5 min.

Foraging efficiency—duration and frequency of foraging interruptions

During the continuous 5-min focal animal observations, described above, observations relating to foraging interruptions were also conducted. Interruption to foraging was defined as an activity that was short in duration (less than 1 min) and prevented the individual from selecting, biting or chewing hay or grass. Both the frequency and overall duration of any interruption was recorded and interruptions were categorised as one of the following:

Vigilance: Head raised from foraging and ears pricked in the direction of interest, the head is higher and the ears upright distinguishing vigilance from raising the head to chew.

Movement whilst foraging: a short movement resulting in a change in foraging location, either following a displacement by another individual or simply changing location at a walk.

Displacements given: interaction directed towards another individual, with the head outstretched and ears flat back against the head resulting in recipient raising head, or taking a step away in any direction.

Displacements received: interaction received from another individual defined as above, causing recipient to raise head, move sideways or take a step away in any direction.

Scratching: Using either the mouth or the hoof to scratch the body.

Startle response: A quick reaction to an unexpected stimulus, the startle usually involved a quick movement, either jump backwards or sideways followed by looking up with ears pricked.

If any interruption lasted for over 1 min then the individual was classed as having stopped foraging. Note that individuals were only observed in detail when they were foraging, if an individual was not foraging when it was due to be observed, this was recorded (to calculate total foraging time, as described in “Animals and Ethical Statement”) and but also counted as ‘missed’ in terms of recording interruptions. Once a missed individual was foraging again it was observed next as a priority (only if it had not yet already been observed for 20 min), but just for a single 5-min interval, before resuming the original order. This was to maximise the collection of data on foraging efficiency for each individual.

The frequency of foraging interruption (a proxy for foraging efficiency) was calculated as the number of instances of all interruptions per minute foraging. Separate frequencies were also determined for each interruption category (Table 1). The duration of interrupted foraging referred to the total percentage of time spent interrupted per individual.

Dominance rank

Although the concept of dominance lacks universal explanatory power in describing social structure, it is a useful construct when considering the specific context of competition for a limited food resource. Under such conditions, horses generally follow a linear ranking hierarchy, with occasional triangles and some influence of third-party interactions (Houpt, Law & Martinisi, 1978; Van Dierendonck, De Vries & Schilder, 1995; Hartmann, Christensen & McGreevy, 2017).

Here we defined dominance ‘an asymmetry in the outcome of dyadic interactions between individuals, or a priority of access to resources’ (Drews, 1993) and assessed it by measuring outcomes between dyadic pairs when feeding from hay feeders. Agonistic interactions were recorded continuously throughout the 3-h observation period (these were easily measurable alongside other observations). An agonistic interaction was defined as one individual approaching or displaying to another with the neck outstretched and ears back flat against the head and, crucially, the second individual moving away. Dominance rank was then calculated using the methods described by Appleby (1980). The number of agonistic interactions both given and received was recorded for each herd individual, and then the number of other individuals that a focal individual both dominated and was dominated by was calculated.

Once an Appleby rank had been given, this was then adjusted to take into account herd size (as in Giles et al. (2015)). Adjusted dominance rank was calculated as 1 − (a − 1)/(h − 1), where a is the Appleby rank and h is the herd size. Where dominance rank or dominance status is referred to in this manuscript, this refers to this adjusted dominance rank.

Body condition score

Measurements were taken immediately after the second set of observations on the herd had been completed. All study animals were accustomed to being handled. Body condition score was measured using the Henneke nine-point scale (Henneke et al., 1983) by a single trained observer (SLG). Six areas of the horse were scored between 1 and 9 and then averaged and rounded to the nearest 0.5, to obtain a single score. A score of five on the scale was taken to indicate an ideal body condition.

Statistical analyses

Results were analysed using Stata 12.1 (Statacorp, College Station, TX, USA). Univariable relationships were assessed using mixed effects linear regression, the clustered study design was controlled for by including herd group and herd size as a random effects, on the basis that herd size or other herd specific factors such as environment could plausibly have some influence on foraging and interactive behaviours. Univariable relationships of primary interest were:

1. The relationship between dominance rank (adjusted for herd size) and body condition score.

2. The relationship between dominance rank (adjusted for herd size) and interruptions to foraging (as a proxy for foraging efficiency).

3. The relationship between body condition and interruptions to foraging (as a proxy for foraging efficiency).

Following an initial univariable exploration of these relationships, relationships between the separate foraging interruption variables were also considered. In addition, breed, age, height, sex and whether or not the individual received supplementary feed were recorded as potential confounding variables. To be considered a potential confounder the variable had to be associated with both the explanatory and outcome variable, and not on the causal pathway between the two (Petrie & Sabin, 2009). Statistical significance was defined using p ≤ 0.05 with a screening p-value for multivariable models of p ≤ 0.07.

Mixed effects multivariable linear regression was then used to build a best-fit explanatory model for both adjusted dominance rank and body condition. The foraging interruption variables (see Table 1 for list) were added to the model one at a time, based on the strength of univariable association, starting with a minimal model. A likelihood ratio test was used to assess the contribution of each variable to the model fit and variables were retained on the basis of this and the adjusted p value.

Multivariable analysis using a mixed effects linear regression model was also used to make predictions regarding interruptions to foraging—to explore whether this could be a possible mechanism linking dominance status and body condition. Duration of foraging interruption was associated with both dominance status and body condition, therefore this was added to a model containing adjusted dominance rank and body condition. Its explanatory contribution to the model was then assessed using both the adjusted p and estimates and a likelihood ratio test.

Univariable Analysis

Multivariable analysis

Vigilance and body condition

Vigilance frequency was the individual interruption behaviour most strongly associated with body condition score—it showed a strong negative association. However, vigilance was not associated with dominance status. These results suggest that certain individuals may be more likely to conduct vigilance, perhaps on behalf of the group, regardless of their social status. These results do seem to support the suggestion that vigilance is an inherently costly activity (Elgar, 1989; Fritz, Guillemain & Durant, 2002; Fattorini & Ferretti, 2019; Pacheco & Herrera, 1999) as demonstrated by the negative association with body condition. However, lower body condition individuals may also be more stressed or nervous individuals, which would also explain the association with increased vigilance.

The complexity of vigilance as a single trait may somewhat explain the lack of observed association with dominance status. Vigilance may serve a range of functions in group living animals (Fattorini & Ferretti, 2019), including anti-predatory behaviour (Elgar, 1989; Hunter & Skinner, 1998), monitoring of other herd members and scanning the environment for resources (Underwood, 1982). Ungulate mammals that are unexposed to predation have been observed to greatly reduce their vigilance behaviour (Hunter & Skinner, 1998). Horses, unexposed to predation, may therefore show relatively low levels of vigilance, with reasons other than anti-predatory vigilance having a proportionally larger role.

Alongside the association between dominance status and body condition, the association between body condition and vigilance provides evidence of two separate behavioural traits associated with body condition in group living animals. Behavioural predictors of body condition have so far received little attention in horses (for exceptions, see Ingólfsdóttir & Sigurjónsdóttir, 2008; Giles et al., 2015) and may warrant continued investigation, especially as obese horses (BCS > 7) may show differences in activity and eating behaviour when compared to lean horses (BCS 4-5) (Moore, Siciliano & Pratt-Phillips, 2019).

Dominance status, movement during foraging and displacement interactions

Subordinate horses showed more movement whilst foraging, and were (as expected) more likely to receive displacements. Indeed, statistical analysis revealed that displacement was strongly associated with movement during foraging in our study population, with subordinate animals forced to move foraging location. Theoretical models and empirical studies have proposed that subordinate individuals may be forced to foraging positions carrying a greater risk of predation (Hamilton, 1971; Hemelrijk, 2000). Future studies could examine whether subordinate animals showed increased vigilance specifically when in displaced locations, and during non-foraging periods.

Overall our results therefore appear to support predictions that displacement reduces foraging efficiency for the recipient (Bautista, Alonso & Alonso, 1998; Stillman et al., 2002). Valuable foraging time is wasted not only over the initial dispute, but also in relocating to a new foraging location. In contrast, dominant horses tended to interrupt their own foraging to displace others, but these interruptions tended to be of short duration, allowing the dominant animal to return quickly to foraging. As our study herds were feeding from hay feeders, potentially displacement and movement occurred more often than would occur during foraging on pasture, due to the artificially close proximity of herd members (Hoffmann, Bockisch & Kreimeier, 2009).