Abstract
Violent conflicts between groups have been observed among many species of group living mammals and can have important fitness consequences, with individuals being injured or killed and with losing groups surrendering territory. Here, we explore between-group conflict among meerkats (Suricata suricatta), a highly social and cooperatively breeding mongoose. We show that interactions between meerkat groups are frequently aggressive and sometimes escalate to fighting and lethal violence and that these interactions have consequences for group territories, with losing groups moving to sleeping burrows closer to the centre of their territories following an intergroup interaction and with winning groups moving further away. We find that larger groups and groups with pups are significantly more likely to win contests, but that the location of the contest, adult sex ratio, and mean within-group genetic relatedness do not predict contest outcome. Our results suggest that intergroup competition may be a major selective force among meerkats, reinforcing the success of large groups and increasing the vulnerability of small groups to extinction. The presence of both within-group cooperation and between-group hostility in meerkats make them a valuable point of comparison in attempts to understand the ecological and evolutionary roots of human warfare.
1. Background
Many animals that live in social groups of stable composition compete intensely with their neighbours for territory, resources and mates. While much of the competition between groups is indirect (e.g. via scent marking behaviour or vocal exchanges), some species directly engage in physical aggression between groups. Although intergroup conflict was once thought to be rare outside of humans [1], violence between groups has now been described across a wide range of taxa including social insects [2–4], birds [5,6] and mammals [7–11]. These studies have provided an important comparative perspective on human warfare [12], which has been argued to have been frequent in our evolutionary history [1,13] and an important force in human social evolution [14,15] (although see [16]).
Some of the most valuable comparative insights on human violence come from observations of intergroup conflict in chimpanzees [17]. Unlike bonobos (who largely avoid aggression [17,18]), chimpanzees frequently patrol territory boundaries [19,20], have vocal exchanges with neighbouring groups [21] and engage in violent intergroup conflicts that have lethal results [7,17,21,22]. Chimpanzee males have been observed to form coalitionary parties for coordinated ‘raids’ on neighbours [23,24]. Although this behaviour may be the most analogous to the kind of violence seen in small-scale human societies [7,25], intergroup aggression has also been observed in many other primates including tufted capuchins [26] and spider monkeys [27] as well as carnivores including grey wolves [8], banded mongooses [9], lions [10] and spotted hyenas [28].
Across mammals, the intensity of territoriality and between-group aggression depends on factors such as resource distribution and population density [9,29], and success in intergroup interactions has a major influence on the survival and reproductive success of individuals and groups. Where lethal violence occurs, intergroup contests have clear fitness costs to those killed and benefits to winners, who reduce competition from rival groups. Intergroup contests may also determine group territories with corresponding fitness benefits. Among lions, for example, dominant groups have higher-quality territories which are associated with higher female reproductive success [10].
Given the fitness consequences of intergroup aggression, it is important to identify the characteristics that determine the outcome of contests between mammal groups. Several factors have been established from empirical research. First, larger groups tend to win intergroup encounters across a range of species [8,11,20]. Second, where individuals differ in body size, weaponry or gain different benefits from intergroup contests, the age and sex composition of groups may influence the outcome of aggressive encounters, with groups containing more adult males usually having the advantage in social mammals, as among grey wolves [8], lions [10] and several primate species [11,29,30]. Third, in some species, the location of the intergroup interaction has an effect on the outcome, with the group closer to the centre of their territory more likely to win, as seen among chimpanzees [31] and capuchins [32]. Additionally, because individuals will gain greater indirect benefits from intergroup contests when they are more closely related to their group mates, theory suggests that more closely related groups may be more formidable opponents [33]. Similarly, we may expect more tolerant interactions between more closely related groups, as seen in extra-group interactions among male mountain gorillas [34].
Here, we explore intergroup aggression in meerkats, Suricata suricatta. Meerkats live in stable and highly social groups within which individuals cooperate extensively [35], defend territories [36–38] and regularly have aggressive interactions with neighbours [38,39]. They also come from a clade containing other highly social species that regularly engage in intergroup conflict (i.e. the banded mongoose [9] and dwarf mongoose [40]). Meerkats live in cooperatively breeding groups where reproduction is monopolized by a dominant pair who produce more than 80% of offspring born in the group [37]. Like many carnivores, meerkats are highly territorial and deposit faeces and anal scent marks at sites both within their territories and at important locations on territorial boundaries [36]. Although cooperation and conflict within meerkat groups has been studied extensively, relatively little is known about intergroup aggression. Here, we (i) provide an empirical description of meerkat intergroup interactions, (ii) show that larger groups and groups with pups are more likely to win an intergroup interaction, and (iii) show that intergroup interactions have territorial consequences, with winning groups moving to sleeping burrows further away from their territory centre and losers retreating to burrows closer to their territory centre.
2. Methods
(a) Study population and data collection
The data presented here were collected as part of a long-term behavioural study of meerkats in the Kalahari at the Kalahari Meerkat Project, South Africa [37]. The Kalahari desert is a semi-arid environment where rainfall and temperature vary dramatically, meaning that resources are distributed patchily through time and in space [41]. Typically, the study population includes 10 groups of an average size of approximately 20 individuals. Groups are habituated to the presence of observers and visited every 1–3 days to record behavioural data, group composition, individual weights and life-history events such as pregnancy. Individuals can be identified during behavioural observations by dye marks on their fur.
Data relating to intergroup interactions were collected during sessions of ad libitum behavioural data collection. In these observation sessions, an observer follows a habituated meerkat group for several hours from their emergence from the burrow in the morning or leading up to their return to the sleeping burrow in the evening. During these observation sessions, interactions between the focal group and meerkats from other groups were recorded whenever seen. These interactions take three forms: (i) interactions between the focal group and dispersing individuals or small parties of dispersers where the focal group encounters an evicted female, a roving male or a small coalition of same-sex individuals from another group; (ii) cases where a focal group encounters the sleeping burrow of a neighbouring group containing pups and usually one adult ‘babysitter’; and (iii) intergroup interactions, where the focal group encounters another group of meerkats containing all or most of its members. Because behaviours related to interactions with single individuals or same-sex coalitions have different evolutionary causes and consequences and have been described elsewhere [38], and because burrow-encounters are relatively rare, we focus here on intergroup interactions.
When intergroup interactions occurred, observers recorded the following six behaviours exhibited by the group under observation (the ‘focal’ group): (i) the focal group observes another group (figure 1a), (ii) the focal group chases the rival group, (iii) the focal group ‘war-dances’ towards the rival group, with erect tails and puffed-out fur (figure 1b), (iv) the focal group retreats from the interaction, (v) the group excavates the rival group's burrow (after chasing the rival group back to it), and (vi) the group engages in aggressive physical contact with the rival group (a ‘fight’) (figure 1c). Fights were defined as when more than one meerkat was seen to engage in aggressive physical contact with a member of the rival group, usually attempting to bite the other. In most fights, most group members (with the exception of pups) engage in physical aggression. Interactions lasted, on average, 18.9 min (s.d. = 20.8) and typically consisted of a sequence of these behaviours. For example, a group might see a rival group when out foraging and ‘war-dance’ towards them before chasing the rival group away. Focal groups were deemed to have ‘lost’ an intergroup interaction if they retreated from an intergroup interaction, and ‘won’ an interaction if not. Because unhabituated meerkat groups were unlikely to attack a habituated group when an observer was present, we only include interactions between habituated groups.
Figure 1. Escalation of meerkat intergroup interactions from initial encounter to lethal violence. The numbers in parentheses represent the number of occurrences of each change in our in dataset. Interactions typically end when one of the groups retreat. ‘Aggression’ and ‘fight’ photographs courtesy of Dominic Cram, KMP. ‘Group seen’ and ‘lethal violence’ photographs taken by M.D., KMP.
Between January 2008 and February 2019, we recorded 330 intergroup interactions in which one of the two meerkat groups involved was being followed by an observer, and 92 interactions in which both groups were being followed by observers. In the latter cases, we randomly assigned one of the groups as the focal. Our final dataset comprised a total of 422 interactions involving 78 combinations of 36 different meerkat groups. The focal group was deemed to have won 212 of these interactions and lost 210. In order to determine whether relatedness between or within groups influenced the level of aggression and outcome of interactions, we estimated mean within-group genetic relatedness using a multigenerational pedigree produced from parentage data established through genetic analysis of 18 microsatellites derived from small tissue samples taken from the tail tips [42]. These genetic data were supplemented with observational field data on maternity where genetic data were missing.
At the beginning of morning sessions and the end of evening sessions, we recorded the GPS location of the sleeping burrow being used by the focal group. Although we did not visit every group every day, we were able to establish whether the group changed sleeping burrow in the evening after an intergroup interaction in 295 cases. Since meerkats typically move 0.5–1 km each day, but have territories of 2–5 km2, sleeping burrow changes usually result in the group foraging in a different set of locations [43]. We compared the post-interaction rates of burrow movement with a baseline figure calculated by randomly sampling 10 000 group-day observations across the study period and calculating whether the group moved sleeping burrow on that day. By taking the GPS locations of sleeping burrows, we were able to estimate the centre of a group's territory by taking the centroid of the polygon formed by the locations of the burrows used by the focal group in the three months prior to the intergroup interaction. This was estimated using the geosphere package [44] with a mean of 94.2 sleeping burrow locations in each estimation (s.d. = 24.3). Estimates of the group's territory centre allowed us to estimate whether sleeping burrow changes after an intergroup interaction resulted in the focal group moving further from or closer to the centre of their territory. The relative distance from territory centre is calculated as A–B where A is the distance between the morning burrow and the centre of the territory and B is the distance between the evening burrow and the centre of the territory. To determine whether dominant individuals or individuals of one sex were significantly more likely to initiate aggressive behaviours, and to establish whether observed rates of burrow change in winning and losing meerkat groups were different from the population baseline, we used proportions tests. To establish whether the distances moved by winning and losing groups when burrow changes occurred were significantly different, we used a permutation test written using a custom script with 10 000 permutations.
(b) Statistical modelling of intergroup aggression outcome
From the point of view of the focal group, each meerkat contest has a binomial outcome (win or loss). We analysed the data using the general approach for animal contests developed and presented by Wilson et al. [45] and using the ‘categorical’ family in MCMCglmm in R 3.5.1 [46]. For dyadic interactions, focal behaviours are probably influenced by both focal and opponent effects, and we thus model both among-focal and among-opponent variance components, allowing us to partition observed variance into its repeatable (among-group) and residual (among-observation) components. This approach also allows us to model the covariance between focal and opponent effects—i.e. the relationship between a group's value as a focal and its value as an opponent. Given that the phenotypic trait in this case is a binary ‘win/loss’ outcome and groups should be distributed randomly across focal and opponent designations, we expect a group that consistently performs poorly is likely to lose when acting as the focal, and to induce wins in the focal group when acting as the opponent. As illustrated by Wilson et al. [47], we, therefore, expect the correlation between repeatable focal and opponent effects in these conditions to be close to −1.
We first modelled the data without any fixed effects, to check that our assumptions regarding the covariance structure hold. We then ran the same model with the inclusion of fixed effects: relative group size, the presence of pups in the two groups, relative within-group genetic relatedness, relative adult sex ratio and the location of the contest relative to the centre of each group's territory. Relative group size was calculated as ln(FocalGroupSize/OpponentGroupSize) where the size of each group excludes pups (less than 3 months old). The presence of pups in the two groups is captured as a numeric variable with three values: (−1) the focal group does not have pups but the opponent group does, (0) both groups have pups or neither group has pups and (1) the focal group has pups but the opponent group does not. Relative within-group relatedness is calculated as ln(FocalGroupRelatedness/OpponentGroupRelatedness) where relatedness is calculated as the mean between all group members, including pups. The relative sex ratio excludes pups and is calculated as (numFocalFemales/FocalGroupSize) − (numOpponentFemales/OpponentGroupSize). The location of the contest relative to the centre of each group's territory is calculated as ln(DistanceFromFocalTerritoryCentre/DistanceFromOpponentTerritoryCentre). In addition to the main model of contest outcome, we also modelled the probability of the focal group exhibiting an aggressive behaviour. This model had the same structure but with estimated between-group genetic relatedness rather than relative within-group relatedness. This allowed us to test whether the degree of genetic relatedness between groups was associated with aggressiveness.
For all models, we used uninformative parameter-expanded priors, with models running for 1 050 000 iterations (discarding the first 50 000 and saving every 250th sample, producing an effective sample size of approx. 4000 for all parameters). Autocorrelation was low among consecutive thinned observations and fixed effects in all models; we inspected trace plots in addition to checking the Heidelberg and Geweke diagnostics to ascertain reasonable convergence, and checked that models were robust to different prior specifications. We present 95% highest posterior density credible intervals for fixed and random effects, and consider fixed effects to be nominally significant if their associated credible intervals exclude zero. We also estimated pMCMC, a measure of significance to a p-value.
3. Results
(a) What happens when meerkat groups meet?
Like many animal contests, meerkat intergroup interactions involve an escalation of aggression from posturing to physical violence. In the open semi-desert environment of the Kalahari, meerkat groups usually see one another before they are in close proximity, allowing one of the two groups to avoid the other before the interaction escalates to aggression or violence. Such avoidance was observed by the focal group in 149 (35.3%) intergroup interactions (figure 1). In the remaining 273 (64.7%) interactions, the focal group exhibited aggression, either chasing or ‘war dancing’ the opposing group. Two hundred and thirty-five (86.1%) of these aggressive interactions ended with one of the two groups retreating from the interaction after this initial chasing or war dancing behaviour and before direct physical violence occurred. However, 38 interactions (9.0% of all) resulted in fighting, with at least one meerkat being killed in 13 cases (3.1% of all interactions). In total, 22 meerkats were killed during these 13 interactions, 16 of which were from the group deemed to have lost the interaction. Two of the 22 individuals were adults (one male, one female) and 20 were pups (eight male, seven female and five of unknown sex).
In 233 occasions, observers were able to identify which individual had initiated an aggressive response towards another group. Males were significantly more likely to initiate an aggressive response (observed = 73.0%, 95% confidence intervals (CI) = (67.7, 78.2), expected proportion = 55.6%, χ12 = 27.79, p < 0.001). Among both males and females, the dominant individuals were more likely than subordinates to initiate a response (dominant males versus other males, observed = 66.5%, 95% CI = (58.8, 73.4), expected proportion = 15.4%, χ12 = 336.1, p < 0.001; dominant females versus other females, observed = 55.6%, 95% CI = (42.5, 67.8), expected = 20.4%, χ12 = 45.85, p < 0.001). Overall, 48.5% of aggressive responses were initiated by the dominant male, 15.0% by the dominant female, 24.5% by a subordinate male and 12.0% by a subordinate female.
(b) What factors determine the outcome of meerkat intergroup conflicts?
The relative size of groups had an important influence on the outcome of intergroup interactions. The results of our MCMCglmm model (electronic supplementary material, table S1; figure 2) show that the strongest predictor of the outcome of meerkat intergroup interactions is relative group size, with larger groups being more likely to win (posterior mean = 2.32, 95% credible intervals = (1.55,3.09), pMCMC < 0.0001). This advantage is equivalent to a group being 1.5 times more likely to win when they outnumber their opponent 2 : 1 as when they are equal in size, if all other effects are equal. Groups with pups (individuals less than three months old) were more likely to win intergroup interactions (posterior mean = 1.17, 95% credible intervals = (0.42,2.02), pMCMC = 0.005). The magnitude of this effect is equivalent to the focal group being 1.4 times more likely to win if they have pups and the other group do not, compared with a situation in which both groups or neither groups have pups.
Figure 2. Factors that determine the outcome of meerkat intergroup interactions. (a) Effects of our five predictors on the likelihood of the focal group winning an intergroup interaction and (b) predicted probability of the focal group winning an intergroup contest as a function of relative group size and of the presence of pups in the focal and opponent group.
We also tested whether meerkat groups were more likely to win interactions that were closer to the centre of their territory than to the centre of their rival's territory. Although the direction of our parameter estimate is consistent with a ‘home advantage’, with groups being more likely to win interactions closer to the centre of their range, the 95% credible intervals overlapped with 0 and are therefore not credible significant (posterior mean = 0.17, 95% credible intervals = (−0.18,0.55), pMCMC = 0.38). Although the direction of the effect size suggests that groups with a female-biased adult sex ratio may be more likely to win interactions, this result is also not significant (posterior mean = 0.75, 95% credible intervals = (−0.17,1.70), pMCMC = 0.13). We found no effect of within-group relatedness on the likelihood of the focal group winning an intergroup interaction (posterior mean = −0.10, 95% credible intervals = (−1.04,0.87), pMCMC = 0.85). Additional modelling of the probability of the focal group exhibiting an aggressive response during an interaction found no effect of between-group relatedness on aggression (posterior mean = −3.03, 95% credible intervals = (−8.38,2.24), pMCMC = 0.27) and otherwise supported the results described above, with larger groups and groups with pups being more likely to aggress their opponents (electronic supplementary material, table S2).
Our model structure also enabled us to estimate the variance in interaction outcome explained by consistent differences among groups, while also accounting for the expected negative covariance between focal and opponent effects. In a model without fixed effects, both focal and opponent variance components were non-zero, with the lower bound of their 95% credible intervals distinct from zero (Vfocal = 2.47, 95% credible intervals = (0.56, 4.95); Vopponent = 3.17, 95% credible intervals = (0.71, 6.63)). As predicted, the correlation between focal and opponent effects was not different from −1 (rfocal,opponent = −0.94, 95% credible intervals = −(1, −0.59)). In the full model, focal adjusted repeatability (i.e. the proportion of phenotypic variance explained by focal effects after accounting for fixed effects variance [48]) was moderate and significant (Rfocal = 0.24, 95% credible intervals = (0.11, 0.43)). Potential explanations for this consistency include the temporal span of different groups across the study period, an unmeasured or latent variable that explains group differences, or differences among groups in the presence of particularly aggressive individuals.
(c) What are the spatial consequences of losing an intergroup interaction?
The outcome of an intergroup interaction had an effect on range use. Losing groups were approximately 40% more likely to change sleeping burrow in the evening following an intergroup interaction than winning groups; 73 of 152 losing groups (47.6%) and 47 of 143 winning groups (32.9%) moved sleeping burrow in the evening after an intergroup interaction (proportions test, χ2 = 5.96, p = 0.015; figure 3). In the case of losing groups but not winning groups, this probability of changing sleeping burrow was higher than the baseline frequency of sleeping burrow change (baseline = 37.1%, baseline versus losers: χ2 = 6.66, p = 0.010, baseline versus winners: χ2 = 0.88, p = 0.35). When burrow moves occurred, we were able to look at the direction of these moves relative to the centre of that group's territory. When losing groups moved sleeping burrow, they moved towards the centre of their territory by an average of 155.9 m (s.d. = 646.6, n = 70; figure 3), while winning groups moved away from the centre of their territory by an average of 121.9 m (s.d. = 555.3, n = 47, permutations test, p = 0.007). These sleeping burrow moves also placed the winning groups closer to the centre of the rival group's territory with winners moving an average 115.7 m closer to the centre of the territory of their neighbour (s.d. = 614.6, n = 42) and losers moving an average of 185.6 m away (s.d. = 559.6, n = 66, permutations test, p = 0.005).
Figure 3. Territorial consequences of meerkat intergroup interactions (IGI). (a) Proportion of winning and losing focal groups that moved sleeping burrow in the evening after an intergroup interaction. (b) Distance moved by winning and losing groups relative to their territory centre. Positive values indicate a move away from the territory centre. Error bars show the standard error.
4. Discussion
We show that interactions between meerkat groups are never tolerant, that the majority involve some form of aggression and that a minority result in physical violence, with meerkats being killed in approximately 3% of all interactions, usually pups. Meerkat intergroup contests, therefore, take a form common to animal contests, with most contests being resolved after posturing or non-lethal aggression [49,50]. However, even when interactions between meerkat groups do not result in physical violence, they may have territorial consequences, with losing groups moving to sleeping burrows closer to the centre of their territory and winning groups moving to burrows further from the centre of their territory, on average.
We find that the most important factor determining the outcome of meerkat intergroup contests is relative group size, with most interactions being won by the larger group. We also find that groups which are foraging with pups are significantly more likely to win interactions. Although winning interactions is presumably always advantageous, this result could be a reflection of the fact that groups with pups gain more from maintaining or expanding their territory than groups without pups. Alternatively, groups with pups may be more likely to stand their ground, being wary of losing pups when retreating (groups have been observed to lose pups when running from a predator). Another (non-exclusive) possibility is that there is an underlying association between aggressive responses to meerkats from other groups and the pup care behaviours elicited by the presence of pups in the group.
Contrary to theoretical predictions [33], we found no effect of mean within-group genetic relatedness on the outcome of intergroup aggression or of between-group genetic relatedness on the probability of aggression during an interaction. However, it is important to note that, as with the lack of kin discrimination in within-group cooperative behaviours in meerkats [35], this does not necessarily indicate that the willingness of meerkats to engage in intergroup contests has not been selected via indirect fitness benefits. Although the effect of interaction location on contest outcome was not significant, the direction of the effect was consistent with the kind of ‘home territory’ advantage seen in other species [32]. Further work is required to establish the relative importance of intergroup interactions in driving territory changes in meerkats, and to establish the longer-term fitness consequences of winning or losing an intergroup interaction [43].
In most mammals, males are more aggressive towards members of other groups [50] and in the meerkat intergroup interactions analysed here, males are also more likely than females to initiate an aggressive response. However, females are not passive in intergroup interactions, with female meerkats initiating around one quarter of aggressive behaviours towards rival groups and contributing as much as males once aggression had started. In line with this, a male-biased sex ratio did not increase the probability of a meerkat group winning an intergroup interaction as in the case for lions, wolves and others [8,10,11]. In fact, the direction of this (non-significant) effect was more consistent with a female-biased sex ratio increasing group formidability. The engagement of both male and female meerkats in intergroup interactions is consistent with both the lack of sexual dimorphism [51] and high rates of intra-sexual aggression among both sexes in meerkats [52]. We also found that within each sex, the dominant individual was much more likely to initiate an aggressive group response. This is consistent with theoretical predictions [53] that the individuals contributing most actively to collective actions will be those who have the greatest fitness stake in the action or who have the largest capability—the dominant pair are typically the largest individuals of their sex as well as most closely genetically related to the group [54,55].
Although lethal violence between adults during intergroup interactions is rare in meerkats, losing an intergroup interaction has probable fitness costs in terms of territory retraction, losing pups during den moves, having less time to forage and increasing stress. These probable fitness costs combined with the fact that larger groups are much more likely to win intergroup contests suggest that intergroup interactions may reinforce the success of large groups and make small groups especially vulnerable to extinction [56]. For subordinate meerkats, the vulnerability of small groups may significantly reduce the fitness benefits of dispersing and attempting to start a new group, instead favouring remaining in the natal group as a subordinate helper [57]. Thus, intergroup interactions may play an important role in maintaining the system of cooperative breeding seen in meerkats.
The aggressiveness exhibited by meerkat groups towards their neighbours, combined with the largely indiscriminate altruism observed within meerkat groups [35], could lead us to consider them ‘parochial altruists’, a term applied to humans, who preferentially cooperate with ‘in-group’ members or increase within-group cooperativeness in response to between-group threats in some contexts [58] (although see [59]). Whether these two characteristics (within-group altruism and between-group hostility) have a shared physiological basis and are causally linked are questions of major theoretical interest. If between-group competition has been a driver of within-group cooperation in human social evolution, we see no reason why it could not also have been important in the evolution of cooperative behaviours in other highly social mammals, such as meerkats and work on the relationship between cooperation and between-group conflict in animal societies may be critical in determining the plausibility of this ‘parochial altruism’ process operating in humans [6,40]. More broadly, thinking about intergroup conflict and its fitness consequences raises important evolutionary questions relating to how individual-level phenotypes aggregate to determine group-level characteristics, and the extent to which we can consider selection to be acting on groups or lineages rather than individuals [60,61].
Ethics
Our work was approved by the Animal Ethics Committee of the University of Pretoria, South Africa (EC010-13) and by the Northern Cape Department of Environment and Nature Conservation, South Africa (FAUNA1020/2016).
Data accessibility
The data and code supporting the analyses are available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.0gb5mkkws [62].
Authors' contributions
M.D. and T.C.-B. conceived of the study; T.C.-B. and M.B.M. managed and oversaw data collection; M.D. and T.M.H. analysed the data; M.D. wrote the manuscript; all authors reviewed the manuscript and approved the final version.
Competing interests
We declare we have no competing interests.
Funding
This paper has relied on records of individual identities and/or life histories maintained by the Kalahari Meerkat Project, which has been supported by the European Research Council (Research grant nos 294494 and 742808 to T.C.-B. since 1 July 2012), the MAVA foundation, the University of Zurich (to M.B.M.) and the Mammal Research Institute at the University of Pretoria.
Acknowledgements
We thank the Kalahari Research Trust for permission to work at the Kuruman River Reserve and the neighbouring farmers for the use of their land. We thank Dave Gaynor and Tim Vink for organization of the field site and database and are grateful to all the volunteers, managers, students and researchers who have assisted with data collection. Thanks to Penny Roth for support and to members of the Large Animal Research Group and Behavioural Ecology group for helpful suggestions, especially Frank Groenewoud, Jack Thorley and Chris Duncan. We thank the Northern Cape Department of Environment and Nature Conservation for permission to conduct this research.