Leadership amongst pigs when faced with a novel situation

Leadership is a risky behaviour that can impact individuals and groups. Leaders, i.e. individuals who perform or initiate a task while other individuals in the group follow, have been studied in different contexts, but there is still a lack of understanding on the role of individual characteristics that may predispose them to become leaders, such as dominance and personality. In particular, the characteristics of leaders in domestic animal populations has been poorly examined. We studied leadership within 32 groups of young pigs ( Sus scrofa domesticus , n = 366 individuals). Leadership was assessed during a group-based fear test (Human Approach Test) which was repeated three times. The first individual per group to touch the person was identified as leader. We assessed repeatability of leadership and characteristics of leaders as compared to followers. Leadership was marginally repeatable, with 6 out of 26 groups having a consistent single leader across all tests. Females had odds 4.13 times greater than males of being a leader, while there was no effect of body weight (a proxy of dominance) or coping style on leadership. The results indicate a similarity with wild populations, in which females lead the herd even though the males, which are superior in body weight, are often dominant.


Introduction
Individuals in a social group show various behavioural characteristics related to social life, such as leadership, dominance, boldness and sociability.Among these behavioural characteristics leadership is characterized as an individual's tendency to initiate movement at the front position (e.g., during a long journey), or to be the first to depart to perform a task or to recruit a partner (Fischhoff et al., 2007;King et al., 2008;Peterson, 2008).Mammalian literature, for example in horses, indicates that a single individual in a social group (wild or domesticated) eventually occupies a leadership position over time (Bourjade et al., 2015).The characteristics of leaders has been poorly studied in domesticated managed populations.
In wild populations, leadership is a trait expressed by a highly motivated, dominant, and knowledgeable individual (King et al., 2009), often with a particular set of morphological, physiological, or behavioural traits (Boinski and Garber, 2000).Although studies have linked leadership expression with different individual attributes and other behavioural correlates (Boinski and Garber, 2000), the relationship between leadership and coping style (i.e.responses to stressful scenarios) has rarely been examined.An individual's coping style can have implications for health (Korte et al., 2005) and fitness (Koolhaas, 2008), and therefore coping behaviours may be an important factor in the development of leadership.
Leadership traits in non-domesticated populations are linked with behavioural expression in other domains.For example, Kurvers et al. (2009) indicated, in barnacle geese (Branta J o u r n a l P r e -p r o o f leucopsis), a correlation between leadership traits and individual animal personality.
Similarly, in zebra finches (Taeniopygia guttata) more active and explorative individuals arrived early at a food patch (Beauchamp, 2000, Schuett andDall, 2009) and, in sticklebacks (Gasterosteus aculeatus), bolder individuals displayed more willingness to leave cover, both indicating a trait of leaders (Harcourt et al., 2009).Additionally, bolder pigeons (Columba Livia), occupied higher ranks in the hierarchy and influenced the collective movement of flocks (Sasaki et al., 2018).Further, exploratory or bolder individuals undertook the leadership role by initiating searching for food instead of depending on other individuals (barnacle geese: Kurvers et al., 2012).Studies from sticklebacks also indicate that within a pair an individual can be a leader in one situation while being a follower in another context (Harcourt et al., 2009;Webster and Ward, 2011).The literature suggests that individual experiences and socio-ecological factors are important in shaping many domains of behavioural expression including leadership traits.
Leadership has been mostly studied in natural settings, typically considering one or a few groups per site (King et al., 2009).This results in limited observations on the characteristics of leaders.Researching leadership in domesticated animals has the benefit that groups can be artificially formed, and henceforth many similarly composed groups can be studied for their leadership dynamics.The process of domestication and the provision of abundant resources in managed domesticated populations may have negated the benefits of leadership and relaxed selection pressure for leadership or even resulted in artificial selection against leadership.Additionally, the characteristics shown by leaders may have changed.Leadership traits in domesticated animal groups have been reported, although sparingly (ewes: Addison and Simmel, 1980;cattle: Dumont et al., 2005;Valenchon et al., 2022).For example, cattle (Bos taurus) show leadership characteristics in decision making, such as searching newer feeding sites (Dumont et al., 2005;Šárová et al., 2010).In domestic pigs (Sus scrofa J o u r n a l P r e -p r o o f domesticus), leadership was suggested as early as 1973 (Meese and Ewbank, 1973a) but has not been taken further.Pigs do form dominance hierarchies (Meese and Ewbank, 1973b) and while dominance rank has indicated individual leadership, (cattle: Šárová et al., 2010; dogs (Canis familiaris) Ákos et al., 2014; bonobos (Pan paniscus): Tokuyama and Furuichi, 2017) it cannot be generalised to all animal social structures.
The aim of this study was to investigate factors related to leadership traits in a domesticated population.This was studied in 32 groups of domestic pigs by assessing how coping style, sex and body weight influence leadership in a group-based novelty situation.Based on the existing literature, we expected that proactive and heavier individuals would have a greater likelihood of adopting a leadership role, as such individuals would theoretically suffer less from taking a risky behavioural strategy in some contexts (Wolf et al., 2007).

Statement of ethical review
This study followed the European Guidelines for accommodation and care of animals and was in accordance with the UK Government DEFRA animal welfare codes and adhered to the ASAB/ABS guidelines.The SRUC's Animal Ethics Committee and the UK Government Home Office (Project licence PPL60/4330) approved the experimental protocol (no.ED AE 21-2014).

Animals and housing
J o u r n a l P r e -p r o o f Male (n=184) and female (n=182) domestic pigs ((Large White × Landrace) × American Hampshire), originating from 53 litters, were studied across four batches (i.e., cohorts of litters born at approximately the same time).Piglets were raised in farrowing pens with the sow housed in a farrowing crate.At birth, they were weighed, vaccinated and identified with ear tags.Tails and teeth were kept intact and males were not castrated.Piglets were weaned from the sow at ~28 days of age and remained for one more week in the farrowing pen.The average weaning weight in males was 7.98 ± 1.41 kg (mean ± SD, range 3.7 to 12 kg) and in females 7.66 ± 1.48 kg (3 to 11.4 kg).Pigs had a positive daily body weight gain between week 2 and 4 of age (mean ± SD: 0.25 ± 0.07 kg / day, range: 0.05 -0.48 kg / day) and between week 4 and 9 of age (mean ± SD: 0.46 ± 0.07 kg / day, range: 0.23 -0.63 kg / day).
Further, the proportion in body weight gain from week 2 to week 4 (range: 0.5 -0.9, SD = 0.07) and week 4 to week 9 (range: 0.16 -0.38, SD = 0.03) suggests that most pigs were following a similar growth trajectory between weeks.
At 5 weeks of age, they were moved to the research building, where they were kept within the same litter (no change in group composition).Group size was on average 12 ± 2 pigs.
Pens had a solid floor with deep litter straw bedding and measured 1.9 m × 5.8 m (11 m 2 ), equating to 1.0-1.1 m² per animal.In winter, there was a covered kennel area made from straw.Each pen had a single-space feeder and a drinker, with ad libitum standard commercial pelleted feed and water.Lights were on from 07:00 -19:00 and the room had natural daylight from ceiling height windows.

Backtest
The backtest is a commonly used behavioural test in piglets in order to assess their coping style (e.g., Zebunke et al., 2017).In total 366 piglets were tested in the backtest, as a J o u r n a l P r e -p r o o f reflection of coping style, following the procedure of Hessing et al. (1993) and as described by Melotti et al. (2011).Piglets were tested at 14 days of age (15.37 ± 1.02; range 13 -18 days of age).The complete litter of piglets was taken to a room adjacent to the test room before the start of the experiment.The test order was balanced between litters of primiparous and multiparous sows (n = 14 and 20, respectively).For the test, the experimenter lifted at random one piglet out of the trolley and brought it to the test room.The experimenter placed the piglet in a supine position on a 25 kg feed bag which was located on a table, with a slight indent in the longitudinal direction (to support the piglet's spine).The piglet was held in this position for one minute.The experimenter placed one hand firmly on the piglet's thorax, with one of the piglet forelegs between thumb and index finger and the other foreleg between index and middle finger.The experimenter's other hand was placed loosely on the hind legs (Bolhuis et al., 2003).The handler counted the number of times a piglet resisted the experimenter's grip by struggling (i.e., escape attempts).One struggle was counted as movement of the body or legs by muscle tension, and lasted until the piglet relaxed the hind legs for at least a second.Another person counted the number of vocalizations.Each breath with a raised sound was counted as one vocalization.After the test the piglet was weighed and returned to the sow.Frequent vocalizations and struggles have been interpreted as a more proactive coping style (Spake et al., 2012), whereas during a reactive coping response piglets typically stay immobile and do not vocalize.The number of vocalizations ranged from 0 -64 (median 8), and the number of struggles from 0 -8 (median 1), showing sufficient variation in these traits for analysis.The distribution of these variables was skewed, with 24% (n = 95) of pigs not vocalizing at all and 29% (n = 114) not struggling.

Human Approach Test (HAT)
J o u r n a l P r e -p r o o f The HAT has been used to assess pigs' response to humans, especially fear of humans (Leliveld et al., 2017;Desire et al., 2023).In total, 366 pigs were tested at 6 weeks of age, when they had had limited contact with humans besides routine farm care and the backtest.
Three tests were carried out, once each during three afternoons (Thursday, Friday and Monday) in the same 7-day period, with the test order balanced between the groups.Pigs were marked on their back with animal marker spray for recognition, which was done the day or morning prior to the test.
One person, hereafter the 'experimenter', conducted all tests.The experimenter was not familiar to the pigs prior to the test and wore a red overall in order to differ in clothing from the animal care takers, who had blue overalls.The experimenter entered into the pen and walked slowly along the walls to make sure all pigs were awake and were paying attention.
Then the experimenter stopped at the centre of the wall opposite the entrance with her back against the pen wall, and stood idle in a half circle with a 0.5 m radius which was marked on the floor with blue spray paint.This position was maintained for ten minutes.The experimenter noted the latency in seconds for each pig to approach within the 0.5 m radius with either its head or fore limbs, as well as the latency to touch the experimenter.Pigs that did not enter the radius within the maximum time were given the latency of 600 s (10 min).
Tests were recorded by CCTV cameras attached to the ceiling and watched later for the pigs' behaviour.

Leadership
Within each group, leadership was assigned to individuals who were the first to touch the human in the HAT.When multiple individuals touched the human at exactly the same time, leadership was assigned to each of them.In case none of the pigs touched the human, no J o u r n a l P r e -p r o o f leader was assigned.Individuals who attained a leadership position across the three test days were categorized as "consistent leader" whereas the other leaders were categorized as "inconsistent leader".All other pigs were assigned as followers.For each test day a leader was assigned, and a note was made of changes in their role (leader or follower).The three repeated observations per individual were used to calculate repeatability of leadership.To analyse the factors that may predict leadership, the analysis was based on the first test out of three in which a leader was apparent.A group with a "distinct leader" was classified as a group in which one animal was the leader instead of several individuals touching the experimenter simultaneously.Six groups were discarded from this analysis due to having no distinct leader on any of the three tests (multiple consistent/inconsistent leaders on each test day).Thus, data by individual contained information on whether the individual was a consistent leader, inconsistent leader, or follower on each test day.Data by group classified whether groups had a distinct leader (consistent or inconsistent) on each test day.

Statistical Analysis
Data were analysed with R version 4.1.3(R Core Team, 2023) using two datasets.Dataset 1 had repeated measures per individual over three test days, which was used to assess repeatability (n = 366 individuals across 32 groups).Dataset 2 was analysed by individual with a single binary outcome (leader / follower) at the individual level for the first test in which a distinct leader was present (only for the groups that had a distinct leader in at least one HAT, n = 289 individuals from 26 groups).In dataset 2, six groups were discarded due to lack of a distinct single leader.P-values ≤ 0.05 were considered significant.
Repeatability of leadership (binary value) was analysed using Dataset 1 with "rptR" package (Stoffel et al. 2017) for repeatability analysis with 1000 bootstrapping and datatype as J o u r n a l P r e -p r o o f "Binary".The model included body weight at weaning (in kg, at week 4) and sex (male / female), coping response (number of struggles) as fixed factors and pig ID and group ID as random effects.The relationship between leadership, backtest outcome and body weight was analysed using Dataset 2. Leadership (yes / no) was the response variable in a generalised linear mixed model using the "lme4" package (Bates et al., 2015) with logit link for binomial data.Latency to touch the human was the response variable.The predictor variables were body weight at weaning, sex and coping response.For coping response both the number of struggles and vocalizations were entered in the model.Due to the correlation between both (Spearman correlation: r = 0.75; p < 0.001) only the number of struggles was retained, based on the best model fit (lowest AIC and BIC values).Batch (i.e., cohort) and group within batch, were included as random effects.Post-hoc pairwise comparisons were performed to compare the differences between the levels of the categorical variable (sex) using the "emmeans" package (Lenth, 2021) with response scale in R.

Repeatability of leadership
On the first test day, most groups (22 groups) showed no clear distinction between leaders and followers as 43% of the pigs did not touch the human.Leadership at the individual level was repeatable but only marginally (R = 0.095; p < 0.001), which was similar at the group level (R = 0.202; p < 0.001).
Across three days of tests, 18 pigs from 11 groups attained a consistent leadership position.
Six groups had a single consistent leader while five groups had multiple consistent leaders that were equally fast in the test.The remaining groups did not have pigs that attained a consistent leadership position.Post-hoc analysis showed that the latency to touch the human J o u r n a l P r e -p r o o f decreased across test days within followers and leaders (Fig. 1).For followers, all pairwise comparisons between days were significant (p <0.001).The followers were also quicker across test days in completing the task right after the leader (day 1: 39.3 ± 6.40, range: 1-487 s; day 2: 27.9 ± 5.93, range: 1-548 s; day 3: 16.2 ± 3.30, range 1-520 s).The latency to touch the human significantly differed, in the post-hoc pairwise comparison within leaders, between day 1 and 2 (p = 0.032) and day 1 and 3 (p = 0.008), but not between day 2 and 3 (p = 0.817).

J o u r n a l P r e -p r o o f
The number of struggles, as an indicator of coping style, was not a significant predictor of leadership (b = -0.361± 0.21; z = -1.7;p = 0.08).Similarly, body weight at weaning did not predict leadership (p >0.91).Body weight did not differ significantly between males and females (p = 0.246).However, females were more often leaders than males, as shown by a negative association between males and leadership (GLMM: estimate = -1.41;z-value = -2.91,n = 289; p = 0.003).From the 26 leaders (classified as leader based on touching the human first on day 1), only six were male.Post-hoc analysis showed that the probability for females to be classified as leader was 0.137 ± 0.029 while for males this was 0.037 ± 0.015, resulting in a 4.13 greater odds for females to attain a leadership position.

Discussion
Leaders take a risk by being, for example, the first to explore a new situation.In the Human Approach Test, pigs which were not deliberately socialized to humans took a risk by approaching an unfamiliar human.As seen from the latencies to touch the human, many pigs did not touch the human until the third exposure to the same situation, suggesting they perceive it as a risk.The leaders stood out with a remarkably shorter latency than the followers.Leadership was only marginally repeatable, with 6 out of 26 groups having a consistent single leader across the three test days.Leaders could not be predicted based on their earlier coping response in the backtest or by their body weight (as proxy of dominance).
However, females turned out to be much more frequently the leader than the males.
No a priori prediction was made regarding sex differences in leadership.Adult and sub adult pigs do show sexual dimorphism in physical appearance (Konjević et al., 2008) and behaviour (Camerlink et al., 2022) but in young pigs, as studied here at six weeks of age, there are only subtle differences between males and females.It was therefore not expected J o u r n a l P r e -p r o o f that females would take a leadership role more often.In contrary, based on the larger body weight of males and their dominant position in fights (Camerlink et al., 2022) it was rather expected that males would take the leadership role.The leadership role of females does correspond with the situation in the wild, where a mature wild boar female leads a small herd (Podgorski et al., 2014) while wild boar males are mostly solitary except for during the breeding season (Maselli et al., 2014).In other species, such as Orca and elephants, female lineages are also important in the leadership of groups, especially during food scarcity (McComb et al., 2001;Brent et al., 2015).While in the current study the females were still very young, it has been observed that young female wild boar may form social groups in the absence of an adult female (Bieber et al., 2019), suggesting that there is leadership in groupliving females.Females did not differ in body weight from males, which may explain why body weight did not predict leadership.
While the repeatability of leadership was significant but of negligible strength, 6 out of 26 groups (23%) had a consistent leader across the three days.The low repeatability may be because the data over which it is calculated contains predominantly followers.However, variation in leadership is found to be common across species, suggesting that individual's leadership position can be contextual.Georgopoulou et al. (2022) showed that across two trials stickleback fish show an emerging trend of repeatable leader-follower dynamics in a coordinated movement during an ordered state as compared to fish in uncoordinated (i.e.disordered) movement.Further, certain female mallard were consistent leaders within and between trials when moving in groups (Bousquet et al., 2017).The small but significant repeatability can be indicative that pigs might use distributed leadership within the group, in contrast to personal leadership (Leca et al., 2003).Alternatively, they may switch leadership position randomly as has been observed in sheep (Gómez-Nava et al., 2022), or switch between the leader and follower position based on the environment as was shown in captive J o u r n a l P r e -p r o o f house sparrows (Tuliozi et al, 2021).Understanding the dynamics of leader and followers beyond the individual level may contribute to a better understanding of the behavioural complexities found in animal groups.
In this study, we did not find an association between coping style and leadership.A reason for this might be that the response in the backtest relates to different traits than those seen in the HAT despite both involving a response to a potential predator, as discussed by Spake et al. (2012).The backtest, although frequently used as a personality test for young piglets (Finkemeier et al., 2018) due to its heritability repeatability over time (Zebunke et al., 2017), has been criticized due to the lack of correlation with other personality tests (O'Malley et al., 2019).Nevertheless, the backtest response has shown association with immune responses (Oster et al., 2015) and the response of the autonomic nervous system during general behaviour and in response to human contact (Krause et al., 2017).Krause and colleagues (2017) showed that reactive pigs in the backtest (no or few struggles, i.e. 'low resisters') had a longer latency to approach and touch a (familiar) human than proactive pigs, but only in the first three days of the experiment and not later on.This suggests that the effect of coping style may be context dependent.Knowledge of factors associated with leadership can increase the understanding of decision making processes across different contexts.Ramos et al. (2021) have suggested that, with the incorporation of leadership information, domestic and wild groups can be managed better.For example, in goats, old and experienced individuals lead the social herd during grazing, which has resulted in the recommendation to provide leading goats with bells to facilitate group cohesion (Miranda-de la Lama and Mattiello, 2010).Various cities in Europe and beyond face challenges to public health and well-being due to human-wild boar conflicts, and conflicts with other species (Licoppe et al., 2013;Massei et al., 2015;Linnell et al., 2020;Valente et al., 2020;Pokorny et al., 2022), for which management strategies are needed to reduce conflicts.Hence, understanding the J o u r n a l P r e -p r o o f

Figure 1 .
Figure 1.Boxplot of the predicted latency (fitted value of the response variable), in seconds,