The effects of female group-changing behavior on female-female aggression, female rank, and female hierarchy stability

It is unclear how habitat features alter animal response to social instability. Only by uncovering such interactions can we fully understand the evolutionary drivers and fitness consequences of sociality. We capitalize on a management-induced manipulation of social stability in an island population of free-ranging feral horses (Equus caballus), living across three distinct habitat types. We tested whether female group-changing behaviour (a reliable measure of social instability) affected 1) female-female aggression, 2) rank within female dominance hierarchies, 3) stability of female hierarchies (in the groups they joined and/or left), and 4) how habitat characteristics shaped these responses. Female group-changing behaviour positively predicted aggression from other females, but only when habitat features such as visibility and freshwater distribution were considered. We found no strong association among female group changing behaviour and the aggression initiated, female rank, or the stability of female dominance hierarchies. Our work reveals that animal responses to social instability are nuanced and impacted by the surrounding habitat. A better understanding of these impacts can help elucidate evolutionary drivers of sociality and mitigate unintended effects of management practices.

Methods
Behavioural and demographic sampling

Comparable to previous studies, data were collected across 7.5 weeks in 2021 (June – July) and 8 weeks in 2023 (June – August), totaling 697.15 hours of behavioural observation (379.9 hours, 2021; 317.25 hours, 2023), averaging 4.72 ± 0.61 hours per mare (range = 0.5-14.25 hours). We recorded all instances of female-female aggression including displacements, bites, kicks, and pushes [25]. Each female’s aggression rate was calculated as the total number of aggressive interactions (either initiated or received) / the total number of hours she was observed. Only reproductive mares (aged 4 years and older) for which we collected at least three hours of data were included in our analyses (n = 51). When possible, animals were observed during both years of the study; two main factors precluded this for some animals: 1) field logistics (finding and then observing animals on foot) required that only mares that could be located on a consistent basis were observed each year and 2) several animals did not meet our age cutoff in 2021 but did in 2023. Consequently, 35 mares were observed in both seasons; the remaining were observed in only 2021 (n = 10) or 2023 (n = 6). Observed animals represented 89% and 91% of all reproductive females in 2021 and 2023, respectively. We excluded 2021 data for one mare that changed groups and initiated aggression at rates that were 4.86 and 5.41 standard deviations above the mean, respectively (Grubbs test: group changes per hour, G = 4.87, U = 0.72, P = 6.63 x 10-6; aggression initiated per hour, G = 5.42, U = 0.66, P = 1.03 x 10-7).

On average, we located each study group once every 7.39 ± 1.09 (range = 3-17) days and once every 9.72 ± 1.14 (range = 5-23) days in 2021 and 2023, respectively. We recorded each group’s GPS location and composition, noting the presence or absence of individual mares. Horses were identified individually by colour, sex, physical condition, and other distinguishing markings, including scars and freeze brands. Ages are known from long-term records and from NPS data (Stuska, unpublished data). We observed groups from at least 15-20 meters away (as per NPS guidelines). Though such distances may be disruptive with other animal populations, the horses of Shackleford Banks are tolerant of human activity. Upon our approach, horses typically continued their activity or looked to the observer for several seconds before returning to their previous behaviour (personal observation, C.M.V.N.).

Mare group changes were rarely witnessed directly (2021, n = 1; 2023, n = 1); therefore, mare absence from a band was an important metric with which we measured the number of changes between groups. We remained with each group for at least 30 minutes to ensure that individuals recorded as absent were not nearby, but out of sight. Group-changing behaviour was confirmed by the mares’ presence in new bands. In 2021, confirmation was made within 6.62 ± 1.03 days on average (range = 0-24); in 2023, confirmation was made within 8.75 ± 1.17 days on average (range = 0-34). Typically, females changed groups individually. In one case, it seemed that two females had changed groups together, though this could not be confirmed as the change was not witnessed directly. Data regarding that change were excluded from our analyses.

Dominance Hierarchies and Elo scores

For each aggressive interaction, a “loser” and “winner” were determined based on whether a female maintained her position or left the immediate area. From these data, we calculated Elo ratings using the randomization process in the aniDom package in R (version 4.2.0)  to determine females’ dominance ranks within each band in each year of the study. To compare ranks across bands of different sizes, we calculated individual ranks relative to the highest-ranking female in each band. We also estimated the stability of female hierarchies within each band using Elo uncertainty scores [45, 46]. Overall, hierarchies were steep, i.e. had a high probability that higher ranking animals would win a given interaction (mean probability = 0.94 ± 0.02, range = 0.67 - 1), and stable (mean Elo uncertainty = 0.91 ± 0.18, range = 0.71 - 0.99 [45]). 

Body condition

Because physical condition can play a role in aggression and dominance, we measured mare condition with rump scores, assessing the curvature of the line between the tailbone and the point of the hip. Only one observer (C.M.V.N.) measured animal condition, referencing a rump score guide regularly to ensure consistency of measurement during the study. Scores were based on a scale from 1 to 5; a score of 1 being the poorest. Females were scored an average of 3.46 ± 0.58 times (range = 2.0-4.9) and 3.11 ± 0.59 times (range = 1.0-4.1) in 2021 and 2023, respectively.