Agonistic behavior and social hierarchy in female domestic rabbits kept in semi-groups

Abstract

In semi-natural mixed-sex groups, rabbits (Oryctolagus cuniculus) organize their social life by forming a hierarchy, which is characterized by linearity and stability. Compared to the natural environment, many factors are different in contemporary breeding systems, such as space allowance and, in the case of semi-group housing, a period of single housing around parturition. Aggressive interactions and the resulting injuries are frequently regarded as negative. However, there is not much information about the social dominance among does in these breeding groups. Therefore, the aim of this study was to generate knowledge about the hierarchy in female rabbits kept in “semi-groups” on farms by using sociometric measurements. Sixteen semi-grouped does of three or five rabbits each with various degrees of familiarity were first mixed one week before parturition. After a separation period of either two or three weeks for giving birth, they were remixed with the same group mates. Both mixing procedures were video-analyzed for agonistic interactions. The observed agonistic behavior was mainly characterized by one-sidedness. Mixing rabbits before parturition did not prevent aggression after the separation period. The resulting hierarchies were neither found to be totally despotic nor egalitarian, as linearity was revealed in most groups, but steepness levels varied between them. While dominant does turned out to be rather despotic, showing high win proportions, the other relationships among lower-ranked does were more balanced. In contrast to the stable hierarchical structure in natural breeding colonies, rank order stability was not confirmed in the studied semi-groups. A larger group size negatively influenced the rank stability. Our results point out some characteristics of the social life in semi-grouped rabbit does in contemporary breeding systems and the differences to semi-natural habitats.

Introduction

The rabbit (Oryctolagus cuniculus) retained most of the behavioral features of its wild ancestor during domestication apart from the fact that the domestic rabbit shows a somewhat changed circadian rhythm with shorter intervals of activity and resting and more staying above ground during the daytime (Stodart and Myers, 1964, Kraft, 1977, Selzer and Hoy, 2003). One substantial attribute of rabbits’ behavior is that in natural or semi-natural environments, they form social bonds that exceed mating purposes (Southern, 1948, Mykytowycz, 1958, Lockley, 1961). The rabbit may therefore be considered as a gregarious species (McBride, 1964). The vast majority of rabbits are organized in small mixed-sexed groups with several females and males (Myers and Poole, 1959, Cowan, 1987b). Depending on the habitat, they may have a tendency toward monogamous relationships (Parer, 1977, Roberts, 1987).

When living in breeding colonies, rabbits organize their social life by forming a hierarchical structure, establishing rank orders sorted by sex (Mykytowycz, 1958, Vastrade, 1986). Thereby, dominant rabbits claim their territory and demand a safe distance from their neighbors, ready to defend their home range (Southern, 1948, Lockley, 1961). While male aggression is more obvious to the observer and is associated with mating (Mykytowycz, 1958, Myers and Poole, 1959, Lockley, 1961, von Holst et al., 1999), the female's fighting intensity may range from very low (Myers and Poole, 1959, Vastrade, 1986) to quite fierce (Mykytowycz and Hesterman, 1975). However, aggression among females is associated with several factors such as scarcity of food (Vastrade, 1986) or increasing densities (Myers et al., 1971, Schuh et al., 2003) and is mainly displayed near the breeding burrows even among members of the same social group (Myers and Poole, 1959, Mykytowycz, 1959, Cowan, 1987a, Rödel et al., 2008). Moreover, female aggression might be directed not only toward other adults but also to the kits of other does (Southern, 1948, Mykytowycz, 1959, Mykytowycz and Dudziński, 1972), whereas the contact to their own offspring is seldom aggressive (Mykytowycz and Dudziński, 1972) but rather rare and not particularly affectionate (Lehmann, 1991). The young themselves start to show increased aggression toward adolescents at the age of approximately three months (Dudziński et al., 1977, Lehmann, 1991, Vervaecke et al., 2010).

To overcome possible problems with upcoming aggression, the most common solution is to keep rabbits in single pens after they have reached maturity (EFSA, 2005, McNitt et al., 2013). However, this treatment is strongly opposed to the natural gregarious behavior, which is described previously, and such an impairment of this behavioral pattern can lead to compromised welfare, reflected in stereotypies and behavioral disorders (Podberscek et al., 1991, Gunn and Morton, 1995, Dal Bosco et al., 2004). However, even if rabbits are group-housed, they are forced to face some challenges, which play an outstanding role regarding their biology. They are mixed and remixed according to our management ideas or experimental purposes, often forming groups of a similar age pattern. In addition, sexually mature rabbits are housed in single-sex groups to avoid undesired pregnancies. Even when reproduction is desired, single-sex groups are a common practice as well because they allow control of the breeding activity by ensuring specific mating and synchronization of litters. Hence, rabbits are often confronted with a severe space constraint, sometimes making it very difficult to avoid each other in their required individual distance.

When rabbit does are housed in groups during the reproduction period, semi-group housing is the most established system. In semi-groups, does are housed in single pens shortly before parturition until after insemination. This treatment provides some advantages, for example, inhibiting direct competition for nest sites and protecting kits from infanticides in their most vulnerable phase until they are ten days old (Rödel et al., 2008). Beyond that, the grouping after insemination prevents pseudo-pregnancies induced by female-female mounting (Rommers et al., 2006, Andrist et al., 2012). In the most widespread 42-day rhythm, insemination takes place 11 day after parturition (EFSA, 2005), and housing in groups lasts for approximately three to four weeks of the six-week reproduction rhythm. When group housing is applied, the contact with new unknown does may be necessary. This happens either before (Machado et al., 2016, Zomeño et al., 2017) or more commonly after (Andrist et al., 2012, Rommers et al., 2014, Buijs et al., 2015) the separation period. It is known that after these groupings, aggression might be displayed, which can result in mild to severe injuries (Andrist et al., 2012, Buijs et al., 2015). However, little information exists concerning the social structure and effects of the separation period in semi-groups of rabbit does.

Owing to the divergent conditions in modern animal husbandry (space allowance and group structure) compared with the mostly studied semi-natural enclosures, knowledge about social hierarchy in rabbits should not be transferred directly from these semi-natural environments to that in the semi-group housings. At the same time, sociometric measurements may contribute to a better understanding of the nature and dynamics of the does’ agonistic behavior in semi-groups and as a consequence could lead to a greater elucidation of its effects (Rommers and de Greef, 2018). Therefore, the aim of this study was to collect and present data on agonistic behavior and on the characteristics of the hierarchical structure of female rabbits housed in a semi-group system in single-sex groups of different group compositions and sizes and with a different duration of separation.

Data were collected from May to October 2017 on a commercial rabbit farm in Germany, where approximately 600 breeding rabbits (HYPLUS PS 19 genetics of HYPHARM S.A.S., France) and their offspring were reared.

The housing system used for the present study consisted of 36 single units, which could be joined to group pens for a variable count of does. Each single unit was approximately 80 cm × 80 cm (length × width), surrounded by wire mesh walls, open at the top and equipped with a plastic slatted floor with 11-mm wide slots and slats. At the back of the pen, an elevated platform 58 cm × 52 cm (length × width) was attached to the wall. A nonperforated plastic mat covered 40 cm × 52 cm (length × width) of the platform and was slightly inclined (9%) to drain off liquid waste. In addition to the elevated area, each single unit was enriched with a plastic tube (placed above the platform), a chain (hanging from the platform), and gnawing material (a piece of wood attached to a chain hanging from the platform and a second piece of wood as well as a cotton rope, each of them being fixed to the wall). Nestboxes measuring 25 cm × 39 cm × 32 cm (length × width × height) filled with wood shavings were attached to the outside of each single unit and opened up for nest building the morning after the does had been housed in the system.

Single units were connected by winding doors measuring approximately 25 cm × 22 cm (width × height) that were opened to allow three or five single units forming a group pen. Figure 1 gives an impression of the housing system.

Does were fed a diet for lactating does (Fok Lapin, Victoria Mengvoeders, The Netherlands) ad libitum as well as chopped hay. Water was freely available from two nipple drinkers per single unit.

Manure was collected in a slurry pit, and the building was vacuum ventilated. During the study period, the mean temperature was 19.5 ± 2.4°C, and the mean humidity was 77.8 ± 9.8%. LED light strips were switched on from 05:30 hours to 19:00 hours, with dimming sequences of half an hour each. Windows served as an additional light source.

Before being housed in the experimental pens, does had been reared either in groups of four to five animals or in single pens until one week before their first parturition. In the first experimental cycle, only primiparous does, that had grown up together in the same rearing groups, were selected.

Three or five animals were collected from a rearing group to form a starting group. When does were housed in the new pen for the first time, doors of single units were kept close for a couple of hours to allow the animals to acclimatize to the new environment. Subsequently, the doors were opened up to form a group pen (prepartal mixing). Does remained in these groups until two to three days before kindling. After a period of single housing, they were returned into their groups (postpartal mixing). The regrouping after birth took place either at the day of insemination, that is, 11 day postpartum (p.p.) (regrouping treatment 11 day) or one week after insemination, that is, 18 day p.p. (regrouping treatment 18 day). Controlled suckling was practiced once a day until the day of insemination. Thereafter, nestboxes were opened for free suckling. On the twenty-first day after parturition, all nests were closed to encourage kits to eat solid food. Nonpregnant does were removed 30 day after kindling, and kits were removed 31 day after kindling.

In the following cycles, new primiparous does were integrated as new group members in the experiment to replace nonpregnant, culled or deceased does (unstable groups) when the following cycle started (one week before kindling). Does that became pregnant again stayed as established group members in their group in the same pen. If no doe in a group became pregnant again, the whole group was replaced with a new starting group. If every doe became pregnant again, the group remained as it was for the following cycle (stable groups).

Four consecutive reproduction cycles were evaluated, investigating a group of three and a group of five from each regrouping treatment per cycle. Owing to this grouping management, 40 individuals participated in the study.

Based on this, groups differed in the following characteristics: experimental cycle (1-4); group size (n₃ = 8; n₅ = 8); regrouping treatment (n_11d = 8; n_18d = 8); group stability (n_{starting group} = 5; n_stable = 1; n_unstable = 10); the percentage of new and established group members; and the percentage of members from group rearing or single rearing (see Table 1 for details of each group composition).

Does were individually marked with animal spray paint (RAIDEX GmbH, Dettingen/Erms, Germany) on their backs before being housed in the pens. Behavioral data were recorded using a digital video recorder (EverFocus AHD ECOR-FHD-16x1-NH) and first person view cameras (resolution 700 TVL, lenses 2.1 mm) placed into each single unit. All lit hours within a 72-hour period after the first and the second mixing were analyzed by counting the frequency of agonistic interactions every hour. The 72-hour period was divided into days: day one for the first 24 hours after mixing, day two and three accordingly. Furthermore, full hours were assigned to specific periods of daytime: early morning (05:01-08:00), mid-morning (08:01-10:00), late morning (10:01-12:00), midday (12:01-14:00), early afternoon (14:01-16:00), late afternoon (16:01-18:00), evening/night (18:01-05:00). For example, when a group was mixed at 11:45:00, then the first hour (11:45:01-12:45:00) and the second hour (12:45:01-13:45:00) were regarded as midday, whereas the third hour (13:45:01-14:45:00) was assigned to early afternoon.

If an animal died during the sampling period of 72h after mixing, behavioral analysis of this group was stopped from this moment on. If an animal had to be removed by animal carers during daily husbandry routines, behavioral observations were interrupted until the group was complete again.

Interactions were recorded according to the ethogram described in Table 2. Based on the description of Graf (2010), aggressive elements were divided into one-sided and both-sided. As all collected behaviors were considered to be of short duration, only the frequency was recorded. Interactions were counted as new interactions after an interruption of two seconds occurred in which both animals exhibited another behavior, for example, feeding, grooming, or being occupied with enrichment material. Interactions were recorded separately for each day, capturing the active and the receiving animal for attacks, mounting, and avoidance behavior or the beginner, loser, and winner for fights. If more than two animals participated in a fight, the interaction was divided into separate dyadic interactions.

Statistical analyses were conducted using the statistic software R (R Core Team, 2019). Odds ratios (OR) were calculated, and the level of significance was set at 0.05.

The occurrence of the behaviors per hour and group was analyzed separately for avoidance and mounting. As fights occurred very rarely, the total number of fights and attacks were referred to as aggression when analyzing the data. In a considerable part of the observed period, no interactions were recorded. Therefore, by implementing the R packages lme4 (Bates et al., 2015) and glmmTMB (Brooks et al., 2017), data were analyzed by applying a zero-inflated mixed Poisson model consisting of the following two parts: in the first step, data were dichotomized (behavior was recorded: no or yes), and we tested how different effects influenced the probability of the behaviors not occurring using mixed-effects logistic regression with logit-link function. In the second step, we tested how the effects influenced the number of interactions per hour using a mixed-effects Poisson regression with log-link function.

The following fixed effects were tested in the models: group size, regrouping treatment, mixing procedure, daytime, hour, group stability, percentage of new group members, and percentage of group members from group rearing. In addition, a random group effect was included. As mounting was not recorded in a sufficient number of hours, the effects of the percentage of new group members and percentage of group members from group rearing were left out of the model for mounting.

To analyze the hierarchical structure within groups, interactions that were recorded in a 72-hours period were added up to make one matrix per group and mixing procedure (prepartal or postpartal) containing all wins (actor) and losses (receiver). Therefore, we added up the behaviors attacks, fights, and avoidance to make dominance-matrices. Fights without a clear winner were excluded from sociometric analysis. Mounting was not included in dominance analysis, as Albonetti and Dessí-Fulgheri (1990) found female-female mounting not to be clearly rank determined.

To analyze linearity, the dominance-matrices were converted into binary matrices, whereby the dominant animal within a dyad was given the value 1 and the submissive animal 0. Both animals in a tied relationship were given the value 0.5, as proposed by de Vries (1995).

In the case of matrices containing no unknown relationships, linearity was analyzed calculating Landau's linearity index h in accordance with Landau (1951), where N represented the number of individuals in the group and v the number of individuals dominated by the individual i:

$$h=\frac{12}{N^3-N}\ast \sum _{i=1}^N{\left(v_i-\frac{\left(N-1\right)}{2}\right)}^2$$

In the case of matrices containing unknown relationships, we followed the procedure to calculate a modified Landau's linearity index h’ as described by de Vries (1995). The calculations were made using the following formula, with u representing the number of unknown relationships:

$$h^{\text{'}}=h+\left(\frac{6}{N^3-N}\right)\ast u$$

Levels of linearity can range from 0 in case of no linearity to 1 in case of a complete linear hierarchy. According to Appleby (1983), linearity levels should not be tested for significance in groups of less than six animals.

In addition, the maximum possible number of circular triads (d_max) and the actual number thereof (d) (Kendall and Babington Smith, 1940) were calculated according to the following formula with S_i being the sum of the row i (i = 1 to N). No number of circular triads was calculated in case of tied relationships, and modified d’ was calculated in case of unknown relationships (de Vries, 1995).

$$d_{\max }=\frac{N^3-N}{24}$$

$$d=\frac{N\left(N-1\right)\left(2N-1\right)}{12}-\frac{1}{2}\sum \left(S_i\right)²$$

$$d^{\text{'}}=d-\frac{1}{4}\ast u$$

The R package steepness (Leiva and de Vries, 2014) was applied to analyze the steepness of hierarchy within the groups. In a first step, we calculated dyadic dominance indices corrected for chance as suggested by de Vries et al. (2006) for matrices containing different interaction frequencies. Based on the dyadic dominance indices, normalized David's Scores (NormDS) were computed afterward and sorted by size to assign a rank to each animal. Low numbers were assigned to higher-ranking animals, starting with 1 indicating the highest rank (dominant doe). The highest number (3 or 5, depending on the group size) indicated the lowest rank. Afterward, the steepness of the dominance hierarchy was measured by using linear regression to fit a straight line through the NormDS plotted against the rank. The slope of the fitted line can vary between 0 in a totally egalitarian hierarchy and −1 (which equals a steepness of 1) in a despotic hierarchy with perfect linearity (de Vries et al., 2006). Afterward, 10,000 randomizations were conducted to test steepness levels for significant differences from expected steepness.

In addition, the percentage of unknown relationships (No interaction was observed within the dyad.), one-way relationships (one individual winning all interaction within the dyad), two-way relationships (both individuals winning some interactions within the dyad), and tied relationships (both individuals winning the same number of interactions) were determined.

The directionality of behaviors was further analyzed, calculating the Directional Consistency Index (DCI) with the following formula:

$$DCI=\frac{\left(H-L\right)}{\left(H+L\right)}$$

Thereby, H stands for the total number of times the behavior occurred in all individuals in the direction of higher frequency and L the total number in the less frequent direction. If the occurrence of a behavior were equally balanced between animals, the DCI would be 0, and if a behavior is completely one-sided, it would be 1 (van Hooff and Wensing, 1987). DCI was calculated for dominance-matrices and separately for aggression-matrices and avoidance-matrices.

To find out in which direction the behaviors aggression, avoidance, and mounting occurred in each dyad within the group, we calculated win proportions sorted by rank. In addition, we calculated win proportions for the dominant does separately for each observation day. The slopes of linear regression lines were used to differentiate upward and downward trends in the dominant does win proportions. We therefore defined slight upward trends (slope m ≥ 0.02 and m < 0.1), strong upward trends (m ≥ 0.1), and slight downward trends (m ≤ −0.02 and m > -0.1), as well as strong downward trends (m ≤ −0.1). Win proportions were also calculated in the case of a divergent winner of the day for the respective animal.

As there was the possibility for the individual to attain another rank in the prepartal and the postpartal mixing within their group or in the following prepartal mixing in the consecutive cycle, we further analyzed the rank stability by applying the R package lme4 (Bates et al., 2015). A mixed logistic regression model was used to test for significant effects on the probability of having a rank swap from the first to the second mixing or from the second mixing procedure to prepartal mixing procedure in the following cycle. Group size, regrouping treatment, mixing procedure, percentage of new group members, the social status of the individuals (being part of a starting group, being an established group member, or being a new group member), and the individual's rearing type (group or single rearing) were included as fixed effects, and the individual and the group were considered as random effects.

To analyze the differences between the ranks, we subtracted the NormDS of each lower-ranked doe from the NormDS of the next higher ranked doe and thus obtained the difference ΔNormDS_x,y between ranks x and y (y = x + 1). In a perfectly linear and steep hierarchy, all these rank differences would be ΔNormDS_x,y = 1. In a first step, the null hypothesis if all rank differences would have the true expected value of ΔNormDS_x,y = 1 was tested, applying the Wilcoxon rank-sum test. Afterward, we tested the global null hypothesis that all differences (ΔNormDS_1,2, ΔNormDS_2,3, ΔNormDS_3,4, and ΔNormDS_4,5) would be the same by implementing the Kruskal-Wallis test. Subsequently, pairwise comparisons were applied using the Dunn-Posthoc test, and P values were adjusted according to Bonferroni-Holm method to account for the problem of multiple testing.