Measurement of reproductive skew in social groups is fundamental to understanding the evolution and maintenance of sociality, as it determines the immediate fitness benefits to helpers of staying and helping in a group. However, there is a lack of studies in natural populations that provide reliable measures of reproductive skew and the correlates of reproductive success, particularly in vertebrates. We present results of a study that uses a combination of field and genetic (microsatellite) data on a cooperatively breeding mongoose, the meerkat (Suricata suricatta). We sampled 458 individuals from 16 groups at two sites and analyzed parentage of pups in 110 litters with up to 12 microsatellites. We show that there is strong reproductive skew in favor of dominants, but that the extent of skew differs between the sexes and between different sites. Our data suggest that the reproductive skew arises from incest avoidance and reproductive suppression of the subordinates by the dominants.
In cooperatively living vertebrates, groups are typically composed of a dominant pair who do all or most of the breeding and a number of subordinates that breed little or not at all but still help rear the offspring in the group. Subordinates face the decision of either remaining as a helper in their natal group or dispersing in an attempt to breed elsewhere (Creel and Waser, 1994; Emlen, 1991; Komdeur et al., 1995; Rabenold, 1985; Stacey and Ligon, 1987). The realized benefit of adopting either of these strategies depends strongly on the probability of (1) breeding as a subordinate or attaining the dominant role in their original group, and (2) emigrating successfully and breeding in a new group. These probabilities are influenced by a number of factors that influence the breeding success of individuals such as reproductive competition (Clutton-Brock, 1998), reproductive concessions made by dominants to retain subordinates as helpers in the group (Keller and Reeve, 1994; Vehrencamp, 1984), and/or the fitness costs of mating with a relative (Charlesworth and Charlesworth, 1987).
Group members can be categorized according to whether they have been resident in a group since birth or have migrated into it from another group, as well as sex, age, and position in a dominance hierarchy. Only natal subordinate helpers stand to gain indirect benefits from helping raise kin. These individuals may have very little reproductive success either because they have no access to unrelated mating partners (incest avoidance) or because dominants are not required to concede reproductive incentives to them to retain them as helpers in the group (Keller and Reeve, 1994).
Although there is considerable observational data on the distribution of breeding success in females, direct genetic analyses have shown that observational data on male mating success can be unreliable (Hughes, 1998). For example, genetic data show that subordinate male dwarf mongooses obtain a much greater share of paternities than thought from observation (Keane et al., 1994), and female Ethiopian wolves (Canis simensis) copulate with males from neighboring groups as often as with their own mate (Sillero-Zubiri et al., 1996). Genetic studies of avian species have also revealed that although the frequency of extrapair paternity is generally low (Cockburn, 1998), exceptions exist (e.g., in Seychelles warbler, Acrocephalus sechellensis [Richardson et al., 2001] and Australian fairy wrens [Brooker et al., 1990; Mulder et al., 1994]). Genetic information also reveals that groups are not always made up of family members as is assumed (e.g. in white-nosed coatis, Nasua narica [Gompper et al., 1998] and gray wolves, Canis lupus [Lehman et al., 1992]). Even in cases in which there are genetic data, it is often difficult to draw firm conclusions about the breeding success of different categories of males because (1) there is a lack of ecological data on relevant factors such as the immigration status of subordinates; (2) the genetic data are limited in their scope from having either screened too few loci or too small sample sizes (for review, see Gompper and Wayne, 1996); (3) only one or a small number of groups is studied, and so, statistical analyses usually treat individuals as independent data points, which they are not, leading to pseudoreplication (Hurlbert, 1984).
In this article, we use microsatellites to assess the breeding success of different categories of males and females in the cooperative meerkat (Suricata suricatta), combined with extensive long-term ecological and behavioral data. Meerkats live in arid regions of southern Africa in packs of three to 20 adults and subadults accompanied by dependent young (Clutton-Brock et al., 1999b; Doolan and MacDonald, 1996a). Each group is composed primarily of a dominant pair and subordinates of both sexes that were born in that group. In addition, some groups contain subordinate males who have immigrated into the group (sometimes in small coalitions) (Clutton-Brock et al., 1999a; Doolan and MacDonald, 1996b). Females may eventually attain the dominant position in their natal territory, but males must disperse to an existing neighboring group to do so (alternatively, either sex may do so by founding a new group) (Clutton-Brock et al., 1999a). Successful breeding by meerkats is completely reliant on helping from other members of the group, who help feed and guard the young at the burrow while the rest of the group spend the day foraging elsewhere in the territory (Clutton-Brock et al., 1998b, 1999c, 2001). Observational data suggest that the dominant female produces around 80% of litters (Clutton-Brock et al., 1999a). The dominant pair in a group are easily distinguished as they are usually older and heavier (Clutton-Brock et al., 2000), are rarely threatened or displaced by other group members, receive more grooming than other animals, and mark territory boundaries an average of 10 times more frequently than the other group members (Clutton-Brock et al., 1998a).
In the present study, we use molecular and ecological data to determine the distribution of reproductive success (reproductive skew) in females and males, and the correlates of breeding success within groups. Our specific aims are to determine the (1) accuracy of previous observational measures of female breeding success, (2) extent to which dominant males monopolize access to dominant and subordinate breeding females in their groups, (3) breeding success of subordinate mature males before dispersal from their natal group, (4) breeding success of mature males that immigrate as subordinates into a new group, and (5) relatedness between breeders.
Data were collected on meerkat populations at two sites in the South African Kalahari basin: one at Nossob in the Kalahari Gemsbok National Park (hereafter referred to as “Park”) and one on ranch land close to the park, approximately 120 km to the southeast (hereafter “Ranch”). These sites and the populations are described in detail elsewhere (Clutton-Brock et al., 1999b). An important difference between the sites is that the meerkat population is almost twice as dense at the Ranch (Clutton-Brock et al., 1999a). In the Park site, wildlife is protected from hunting, and so there are high densities of raptors, which are the primary predators of meerkats (Clutton-Brock et al., 1999b). Groups were visited for observation daily or at least every 2 weeks (Clutton-Brock et al., 1999a).
Samples for genetic analysis were collected from both study sites over a period of 5 years (1993–1998). A layer of skin, 2–5 mm in diameter, was removed from tail tips of as many individuals as possible from groups monitored in each study site. We extracted DNA from 548 samples (236 [six groups] from the Park and 312 [10 groups] from the Ranch). Twelve microsatellite loci (three derived from other species of carnivores and nine cloned directly from meerkat DNA) were then amplified and screened as described by Griffin et al. (2001).
Only individuals typed at six or more loci out of a possible 12 were included in the analyses described below. From the sampled population, we genotyped 166 individuals as pups and a further 64 as adults from the Park (44 litters from six groups) and 220 as pups and 108 as adults from the Ranch (66 litters from 10 groups). The mean number of alleles per locus was 10, and mean observed heterozygosity was 0.75 (SE = 0.02) at the Park and 0.77 (SE = 0.02) at the Ranch.
We performed parentage analysis by using the parentage inference package CERVUS (Marshall et al., 1998). This program calculates a statistic, Δ, that is defined as the difference between the log-likelihood ratio (LOD score) of the most likely candidate parent of an offspring. The statistic Δ is then used to assign the most likely parent of an offspring at a specified level of confidence. The analysis uses locus-specific allele frequency data for the population of interest, in our case, different data for meerkats at the Park and the Ranch because allele frequencies differed between sites (Griffin, 1998). It also requires a number of population parameters, for which the following values were used in this study. Number of candidate parents was 10, a conservative value for the number of candidate mothers and fathers of a litter at any time in either site. Mean proportion of loci typed was 0.760 (Park) and 0.752 (Ranch), observed values for each site. Proportion of candidate parents sampled was 0.6, estimated value at both sites. Error rate was 0.02, estimated from cases in which we were particularly confident of mother-pup relationships. Number of simulation cycles was set at 10,000. We sought to identify parents of pups at 80% and 95% confidence.
Using CERVUS we first sought to identify the mother of a pup. Pregnancy can be detected with some confidence in meerkats, because the abdomens of females swell visibly. However, it was possible that subordinate females could breed without being observed to be visibly pregnant, so all mature females in a group were also considered. Identifying one parent, in this case the mother, greatly improves power to identify the other parent. In total, 302 pups were assigned maternity at more than 80% confidence out of 348 (87%) that were analyzed.
Second, we sought to identify the father of each pup, using maternal genetic information (see above) and a CERVUS analysis of candidate fathers. The candidate fathers were made up of (1) the dominant male in the group, (2) all other males in the group, and (3) any solitary males that were observed to enter the group containing estrous females. In total, 239 pups were assigned paternity at more than 80% confidence out of 348 (69%) that were analyzed. There is no reason to suppose that parents of unassigned pups belonged to a different distribution of category from assigned pups, and we, therefore, omitted these pups from further analysis. A further 27 (8%) pups were assigned paternity to males outside of their groups (the last category described above) after exclusion (i.e., LOD = 0) all males present in their group at the time of conception. Allele frequencies were sufficiently different between neighboring groups to make paternities by males from outside the group possible to detect (Griffin, 1998).
Analysis of relatedness
We measured relatedness between individuals by using the package KINSHIP 1.0 (http://gsoft.smu.edu/Gsoft.html) (Goodnight and Queller, 1999), which estimates relatedness by the extent to which individuals share alleles identical by state based on the method of Grafen (1985). Relatedness data were generated for each group in the form of matrices containing all individuals ever born, seen, or known to have been in contact with that group. The latter were usually males from neighboring groups that had been assigned paternity of pups using CERVUS, either immigrants or solitary males mentioned in paternity analysis section.
Lists of individuals present at the time of conception were available for 46 litters at the Ranch, and matrices generated by KINSHIP 1.0 were used to calculate mean relatedness values between breeders and nonbreeders.
Two possible statistical problems in the analysis of the sort of data described in this article are those of pseudoreplication and the treatment of proportion data. First, data from different animals or different categories of animals living in the same group are often treated as independent data points, usually because there is only information available from one or two groups. This results in pseudoreplication because members of the same group are not independent (Hurlbert, 1984). In this study, unless otherwise stated, the unit of analysis used was an average across groups (n = 16 over both sites).
Second, proportion data, such as proportion of pups sired by dominants, usually have non-normally distributed error variance and unequal sample sizes. To avoid these problems while retaining maximum power, all proportion data were analyzed with a general linear model analysis of deviance, assuming binomial errors, and a logit link function in the GLIM statistical package (Crawley, 1993). Importantly, this form of analysis weights each data point according to its sample size (e.g., number of pups in the group) and, so, controls for the fact that different numbers of pups were sampled from different groups and that the error variance is greater with small samples. Significance testing was performed with χ2 tests. The appropriateness of the assumption of binomial errors was checked by comparing the residual deviance with the residual degrees of freedom after fitting the explanatory variable. Large relative values of the residual deviance indicate overdispersion, which may result in overestimation of significance levels. To account for this, we rescaled the deviance by the heterogeneity factor (HF), the ratio of the residual deviance to the degrees of freedom (McCullagh and Nelder, 1983). When correcting for overdispersion was necessary (when HF > 3) the significance of a term was then tested for using an F test (Crawley, 1993).
Accuracy of observation of female reproduction
Previous observational work suggests that in virtually all groups, dominant females are the mothers of the majority of pups born (Clutton-Brock et al., 1998a). Our genetic analysis confirms that when field workers are able to observe and monitor pregnant females in a group, they are able to identify mothers of litters. Overall, 89% of 90 litters and 93% of 302 pups were the offspring of dominant females. The remaining pups were born to subordinate females that had also been identified correctly as the mother by observers. Within a group, there are on average 1.2 sexually mature subordinates and one dominant female. Considering the reproductive success as per individual, dominant females produced 8.7 times as many offspring as did each subordinate female (Figure 1a). Looking at the two sites separately, the dominant female was identified as the mother of 100% (n = 6 groups) of the pups for which maternity could be assigned at the Park (total of 114 pups from 44 litters), and a mean of 88% (SE ± 0.03%, n = 10 groups) at the Ranch (total of 188 pups from 66 litters). The difference between sites was significant (HF = 2.24, F1,14 = 9.57, n = 16, p <.01).
Reproductive dominance of dominant males
Overall, dominant males fathered the majority of pups born in all groups (80%, 65 of 81, of litters; 77%, 203 of 255, of pups). Within a group, there are on average 2.6 sexually mature subordinates and one dominant male. Considering the reproductive success of individuals, dominant males produced 11.8 times as many offspring as did each subordinate male (Figure 1b). This skew was significantly greater than expected by chance (p <.01, n = 16 groups). Considering the two sites separately, the dominant males were estimated to be the fathers of 88% (+0.03%, −0.04%, n = 6 groups; SEs are asymmetric owing to back-transforming a maximum likelihood analysis based on a logit link; see Statistical analyses) of pups for which paternity could be assigned at the Ranch (97 of 143 pups from 48 litters) (Figure 2a) and 72% (±0.04%, n = 10 groups) of pups on average at the Park (106 of 121 pups from 33 litters) (Figure 2b), and this difference between sites was not significant (HF = 3.08, F1,14 = 2.48, n = 16, p >.05).
The breeding success of dominant males, measured as the percentage of all paternities achieved, was not significantly lower in groups containing immigrant subordinate males (χ2(1) = 0.94, n = 12), even though immigrants achieved an average of 31% of paternities in the six groups in which they were present at the Ranch (average of 14% paternity to subordinate immigrants over the 10 Ranch groups). This was because outside males only bred when immigrant subordinate males were not present in the group (achieving 22% of paternities in the four groups in which immigrant males were not present, and an average of 13% of paternities overall in the 10 groups). However, the paternities lost to subordinate and outside males differed: Subordinate males tended to share paternities with dominant males in litters produced by the dominant female (69%, nine of 13, of subordinate males' litters had mixed paternity), whereas outside males tended to father whole litters (27%, three of 11, of outside males' litters had mixed paternity).
Immigrant males always achieved fewer paternities than did the dominant male in a group (considering the five groups with immigrants present, sign test, p =.03). Overall, 16 litters were produced by the dominant female at a time when there was at least one subordinate immigrant male present (mean number of males present = 2.6 per group). Of these, nine litters were fathered by the dominant male and two by a subordinate immigrant male, and paternity was split in the other five between the dominant and subordinate male. Of the 36 pups born into mixed paternity litters, 26 (72%) were fathered by the dominant and 10 (28%) by the immigrant.
Who fathers the offspring of dominant females?
Overall, dominant males were the fathers of the majority of pups born to dominant females in all groups (88%, 72 of 86, of litters; 83%, 197 of 236, of pups) (Figure 3). Considering the two sites separately, dominant males were estimated to be the fathers of 79% (±0.04%, n = 10 groups) of pups at the Ranch (91 of 115 pups from 53 litters) (Figure 2a) and 88% (+0.03%, −0.04%, n = 6 groups) of pups for which paternity could be assigned at the Park (106 of 121 pups from 33 litters) (Figure 2b). This difference between sites was not significant (HF = 3.08, F1,14 = 2.48, n = 16, p >.05).
Who fathers the offspring of subordinate females?
Relatively few of the pups born to subordinate females were fathered by a dominant male. Considering the four Ranch groups in which subordinate females bred, dominant males fathered a significantly lower percentage of the subordinate females' pups (18%, three of 17, of pups, with paternity in one of 8 litters) compared with the dominant females' pups (79%, 91 of 115, of pups, with paternity in 39 of 49 litters), (t = 4.89, p =.02, n = 4) (Figure 3).
Breeding success of subordinate males before dispersal
Subordinate philopatric males never fathered offspring in their own natal group. They fathered none of the 264 pups from 81 litters for which paternity was assigned even though there were on average two subordinate philopatric males per group. Subordinate philopatric males may father offspring in other groups. Older philopatric males not infrequently leave their natal groups for a few days at a time (Clutton-Brock et al., 1999a; Doolan and MacDonald, 1996b) and are often seen in the vicinity of other groups. Although 13% of pups are fathered by males not belonging to the same group as the mother, we currently do not know what proportion of these are males still resident in their natal groups.
Reproduction by subordinate males after dispersal
Immigrant subordinate males commonly bred in their resident group: in around half of the groups we studied (three of six at the Park; six of 10 at the Ranch, an average of one immigrant male per group). All three subordinate males that were known to be immigrants at the Park bred. At the Ranch, two groups contained the same band of three brothers that had moved from one group to another together; the other four contained single immigrant males whose origins were unknown. In addition to these seven males, there were an additional five males who had not bred by the time this analysis was completed.
Although most of the pups fathered by immigrants were produced by the dominant female, subordinate males fathered a greater percentage of subordinate females' pups (Figure 3). Of 36 pups produced by dominant females when subordinate immigrants were present, 10 were fathered by the subordinate immigrant (28%); of seven pups produced by subordinate females, four were fathered by the subordinate immigrant (57%). Subordinate females only produced one litter over the study period in each of two groups when an immigrant male was present.
When immigrant males bred they were more likely than dominant males to share paternity of a litter (eight of 14, 57%, subordinate male litters had mixed paternity compared with 10 of 39, 26%, dominant male litters; sign test on groups in both sites where both dominant and immigrant males bred; p = .016, n = 9).
Relatedness between breeders and nonbreeders
Figure 4 shows the average meerkat group composition at the Ranch site (similar data from the Park are incomplete and, therefore, not shown), indicating the average relatedness between different sex/status categories. Breeding pairs were never related: The average relatedness of breeding pairs did not differ significantly from zero (t = 0.13, p >.2, n = 15; average relatedness between Ranch pairs = −0.03, SE = 0.04, n = 10 groups; and Park pairs = 0.068, SE = 0.07, n = 5 groups) (Figures 4 and 5). Furthermore, the average relatedness of breeding pairs was significantly less than the average relatedness between adult males and females within groups (p <.004, sign test, n = 15 groups) (Figure 5). In no case was a dominant male observed to mate with a subordinate female who was his daughter. In the only case in which the dominant male fathered pups born to a subordinate female, the dominant male was thought to be the half brother of her father (r between this breeding pair = −.12).
Subordinate male breeding success can be predicted by access to an unrelated opposite sex breeder, but female subordinate breeding cannot (Figure 6). Subordinate philopatric males, who never bred (Figure 2), were commonly the sons of the dominant female (Figure 6). This difference between philopatric and immigrant males in relatedness to the breeding female was significant (Figure 6).
Nonbreeding females, on the other hand, had highly variable relatedness to the breeder of the opposite sex (Figure 6). This difference was almost certainly because most subordinate females were the offspring of dominant males. Relatedness among members of a group appeared to be driven by the higher turnover of dominant males than of females and higher migration rate between groups of males than of females. The higher turnover in dominance in males compared with females is reflected in higher relatedness in all classes of females relative to males (Figure 4).
Figure 4 also shows that immigrant males at the Ranch were often related to the dominant male in their new group. In three out of the four groups containing immigrant males at the Ranch, the immigrant males present were related as brothers to the dominant. These males were, therefore, related to pups fathered by the dominant but not to pups that may have been born in the group before a successful usurpation.
Dominant males father most of the pups born (87% Park, 72% Ranch) (Figure 2a,b) on average 12 times as many as each subordinate male in their group (Figure 1b). The remaining pups were fathered either by immigrant subordinate males (11% Park, 14% Ranch) or by males from outside the group (3% Park, 13% Ranch). Philopatric subordinate males never bred in their own group, nor did they ever inherit dominance. Different categories of males obtained paternity of pups produced by different categories of females: Dominant females' pups were fathered mostly by dominants and immigrant subordinate males (Figure 3), whereas subordinate females' pups were fathered mostly by males from other groups—possibly older subordinate philopatric males on forays (Clutton-Brock et al., 1999a).
Immigrant males and reproductive skew
Despite the ability of immigrant subordinate males to obtain paternities, the presence of immigrant males in a group did not significantly reduce the breeding success of dominant males relative to males in groups in which there were no immigrants present. This is because males that wander into a territory without becoming group members (outside males) father a similar percentage of pups to immigrant males, but only did so in groups in which there are no immigrant subordinates present. However, the pattern in which immigrant and outside males bred differed (Figure 3). Immigrant males father pups in dominant female litters that also contain pups fathered by the dominant (if the immigrant male was not present, presumably the dominant would have sole paternity of her litters). Outside males generally father whole litters produced by a subordinate female. Dominant males father relatively few pups born to subordinate breeding females (Figure 3), presumably because subordinate females are commonly their daughters. In both cases in which dominant males fathered offspring of subordinate females, the subordinate females were unrelated to the dominant male. This suggests that immigrant subordinate males may affect the reproductive success of dominants to a greater extent than males from outside the group.
If immigrant males are able to obtain paternities in litters, which would otherwise be fathered solely by the dominant male, why do dominant males tolerate their presence? It is possible that dominants are not able to evict subordinates, but then we might expect that “outside males” would join groups in which they had mated subordinate females. Another possibility is that breeding subordinate males are commonly related to dominant males and have immigrated into the group at the same time as a coalition (Figure 4). Formation of coalitions as a strategy for increasing the chance of breeding in a new group has been documented in several other species of cooperative mammal, for example, in lions (Packer et al., 1991), white-nosed coatis (Gompper et al., 1997), and dwarf mongooses (Waser et al., 1994). Once a group has been formed or taken over by a coalition of males, the presence of subordinates may continue to benefit the dominant male by increasing group size (group augmentation benefits [Kokko et al., 2001]), outweighing the cost of losing paternities.
Our results show that there is a significant lack of incestuous mating compared with that expected by chance. Several points suggest that this may be owing to incest avoidance: (1) dominant males do not breed with subordinate females who are generally their daughters, and (2) subordinate females only breed when there are unrelated males in their group or on rare occasions when solitary males roam through their territories. (Solitary roaming through unfamiliar territory is likely to be a risky strategy for males, which may give rise to the statistically significant difference [see above] in subordinate female breeding success between the Park [in which predator density is relatively high] and the Ranch [cf. Figure 2a and b].) (3) Subordinate philopatric males never breed with the dominant (usually their mother) or subordinate (usually their sisters or half-sisters) females in their own group. Further support for a role for incest avoidance comes from the observations that (1) in the cases in which a dominant female died, leaving an unmated dominant male and one or more mature subordinate females who were his daughters, none of these mature philopatric females bred with their father (in most cases the dominant male soon left the group); (2) in the small number of cases in which both dominant breeders died and there were no immigrant males in the group, the subordinate females did breed, but never with a male from their own group; and (3) hormonal studies suggest that subordinate females are not constrained from breeding (O'Riain et al., 2000).
More generally, data suggesting a role for incest avoidance in the evolution of reproductive skew and dispersal in cooperatively breeding species come from a variety of observational and experimental studies on cooperative vertebrates. Breeding pairs were found to be unrelated relative to average relatedness in populations of white-tailed prairie dogs (Cynomys ludovicianus; Dobson et al., 1997), gray wolves (Smith et al., 1997), and acorn woodpeckers (Koenig et al., 1999). In addition, experimental studies on naked mole-rats (Heterocephalus glaber; Ciszek, 2000) and the Damaraland mole-rat (Cooney and Bennett, 2000) show that mating with nonrelatives is preferred. Furthermore, the level of inbreeding depression that has been shown to occur in another mole-rat species (Cryptomys darlingi) is sufficient to select for strong incest avoidance behavior (Greeff and Bennett, 2000). In contrast to our results, the dwarf mongoose (Helogale parvula, the nearest relation to the meerkat that has been examined for evidence of incest avoidance) is one of the most cited examples for its lack of incest avoidance (Keane et al., 1996). However, this study did not use direct genetic estimates of pedigrees and relatedness, and so, incest avoidance cannot be ruled out. For example, when assessing a role for incest avoidance, (Keane et al., 1996) assume that the dominant pair produced all pups, despite their previous work having shown that subordinate males produced approximately 24% of pups (Keane et al., 1994).
Dominance and dispersal
Not all patterns of reproductive success in meerkats can be explained by incest avoidance. In particular, dominant males still monopolize the paternity of pups even when there are unrelated immigrant males in the group. This suggests either that the dominant males are exercising some control of who mates (mate guarding) or that females prefer to mate with dominant males rather than with subordinate immigrant males. Mate guarding may be relatively easy in meerkats as members of a group are rarely out of sight from one another as they forage and sleep together. Furthermore, incest avoidance alone does not explain why subordinate females often do not breed even when there is an unrelated male present in the group (Figure 5). Behavioral suppression of subordinate females by dominant females has been well documented previously by Clutton-Brock et al. (1998a). Given the ability of the dominant female to successfully suppress breeding attempts made by her subordinate daughters and sisters, suppression may be less important for males than for females.
Given that subordinate males never breed in their natal group, an obvious question is why don't all males disperse when they reach maturity? In meerkats, subordinate philopatric males (1) gain direct fitness benefits by living in a group in which their survival rate is greater until they are older and experienced enough to successfully immigrate into a new group (Clutton-Brock et al., 1999a,b,c; Kokko and Ekman, 2002), and helping may be a form of rent-paying to stay in the group (Gaston 1978, Kokko et al., 2002); (2) obtain some direct fitness by mating with females from other groups; and (3) gain indirect (kin-selected) fitness benefits by helping raise pups who are their siblings and half siblings (Brown et al., 1982; Creel and Waser, 1994; Hamilton, 1964a,b; Mumme et al., 1989).
We compare breeding success of immigrant subordinate males with philopatric subordinate males by counting the number of pups produced in a group by each category of male (Figure 2). However, to make an accurate comparison of the relative fitness pay-off of the two strategies—dispersing versus staying at home—it would be necessary to measure (1) the chance of emigrating into a new group successfully by dispersers, and (2) the chance of gaining paternities in groups without becoming a member (as an outside male). Given the difficulties in measuring this in the field, we must presume that our measure of immigrant male breeding success is an overestimate of the fitness pay-off from dispersing from the natal group.
To conclude, our results provide a clear example of how multiple factors can influence the distribution of reproductive success in cooperative breeders. We have suggested that reproductive skew is primarily determined by (1) incest avoidance and (2) reproductive suppression of subordinates by the dominants. In turn, the importance of these two factors, as well as when they differ between sexes and sites, will be determined by population demographics (e.g., dispersal rates, age structure within groups). Data on reproductive skew show the immediate direct fitness consequences for subordinates of staying and helping in a group. A major task for future research is to determine the relative importance of immediate and future fitness consequences (direct and indirect) to staying and helping in a group, as well as the underlying causal factors (Griffin and West, 2002; Hatchwell and Komdeur, 2000). However, this work will require detailed long-term studies in order to accurately determine (1) the future fitness benefits of helping, such as the chance of inheriting dominance and the average tenure of dominance; (2) the chance of breeding in another group for males that disperse from their natal groups; and (3) the extent to which competition between relatives reduces the kin-selected benefits of helping raise relatives (Frank, 1998; Queller, 1994; West et al., 2001, 2002). The relative importance of all these factors will depend fundamentally on population demographics.
We are grateful to the National Parks Board of the Republic of South Africa for permission to work in the Kalahari Gemsbok national Park, and to Mr. and Mrs. H. Kotze for permission to work on their land at Van Zyl's Rus. This study would not have been possible without the support of members of the Mammal Research Institute, University of Pretoria. A.G. and J.P. thank the many volunteers and workers in the field who collected our samples. For statistical advice, we thank T. Marshall and S. West. For advice and comments on the manuscript, we thank S. West, L. Kruuk, C. Faulkes, and three anonymous referees for thoughtful and constructive criticism. Our research is funded by grants from the Natural Environment Research Council (NERC).