1Many studies have investigated why males and females segregate spatially in sexually dimorphic species. These studies have focused primarily on temperate zone ungulates in areas lacking intact predator communities, and few have directly assessed predation rates in different social environments.
2Data on the movement, social affiliation, mortality and foraging of radio-collared African buffalo (Syncerus caffer) were collected from 2001–06 in the Kruger National Park, South Africa.
3The vast majority of mortality events were due to lion (Panthera leo) predation, and the mortality hazard associated with being an adult male buffalo in a male-only ‘bachelor’ group was almost four times higher than for adult females in mixed herds. The mortality rates of adult males and females within mixed herds were not statistically different. Mortality sites of male and female buffalo were in areas of low visibility similar to those used by bachelor groups, while mixed herds tended to use more open habitats.
4Males in bachelor groups ate similar or higher quality food (as indexed by percentage faecal nitrogen), and moved almost a third less distance per day compared with mixed herds. As a result, males in bachelor groups gained more body condition than did males in breeding herds.
5Recent comparative analyses suggest the activity-budget hypothesis as a common underlying cause of social segregation. However, our intensive study, in an area with an intact predator community showed that male and female buffalo segregated by habitat and supported the predation-risk hypothesis. Male African buffalo appear to trade increased predation risk for additional energy gains in bachelor groups, which presumably leads to increased reproductive success.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Sexual segregation, or the tendency for males and females to live in separate groups outside the breeding season, is widespread in many sexually dimorphic ungulates (for reviews see Main, Weckerly & Bleich 1996; Ruckstuhl & Neuhaus 2002; Bowyer 2004). Sexual segregation can result from either habitat segregation (different use of habitats) or social segregation (sexes live in different social groups). A number of hypotheses have emerged to explain sexual segregation that invoke behavioural, nutritional, environmental, reproductive and physiological–morphological arguments (Main et al. 1996; Barboza & Bowyer 2000; Conradt & Roper 2000; Ruckstuhl & Neuhaus 2000). The most promising of these hypotheses relate to two main concepts: optimizing reproduction relative to mortality, and the effects of size dimorphism on either activity budgets or forage selection.
The predation-risk hypothesis predicts that males and females will choose different habitats outside the breeding season in order to optimize reproduction relative to mortality. Females with offspring will choose predator-safe habitats even at the expense of forage quality or energy expenditure. Males, on the other hand, may sacrifice security for additional energy gains so that they may be more competitive for mating opportunities (Miquelle, Peek & Van Ballenberghe 1992; Main et al. 1996; Bleich, Bowyer & Wehausen 1997). This hypothesis implies a trade-off between predation risk and forage quality, otherwise both males and females should choose the habitat that is both safe and of the highest forage quality. Tests of this hypothesis, however, rarely directly estimate mortality rates in different habitats or social environments. The forage-selection hypothesis predicts that females will use areas of higher food quality than males because of differences in metabolic and digestive capabilities associated with body size (Ruckstuhl & Neuhaus 2000; Post et al. 2001). Finally, the activity-budget hypothesis predicts that males and females will be forced to associate in different social environments because body size differences will result in different foraging and digestion rates, which will then affect activity budgets (Conradt 1998; Ruckstuhl 1998). Key predictions of this hypothesis are that females will spend more time than males foraging, to compensate for their lower digestive efficiency, while males would spend larger proportions of time resting and ruminating. Poor synchronization of activity bouts of males and females would be expected to result in social, but not habitat, segregation. Note that the forage-selection hypothesis predicts the opposite outcome to the predation-risk hypothesis regarding sexual differences in forage selection. While the predation-risk and forage-selection hypotheses predict habitat segregation, the activity-budget hypothesis predicts social segregation.
Previous studies of sexual segregation in African buffalo (Syncerus caffer) have had mixed results. Halley & Mari (2004), in Chobe, Botswana, speculated about support for the activity-budget hypothesis in the early dry season. In Hluhluwe-iMfolozi Park, South Africa, however, Turner et al. (2005) did not find evidence to support the activity-budget hypothesis in buffalo. Their data showed that, even among females, there was limited synchrony of grazing times without compromising the ability to maintain cohesive herds. Activity synchrony was therefore unlikely to limit the ability of males to remain associated with mixed herds. In Manyara, East Africa, Prins (1996) showed anecdotally that bulls appeared to gain body condition while in bachelor groups and lost condition in breeding herds, which is a prediction that can be made from either the predation-risk or activity-budget hypothesis.
African buffalo are sexually dimorphic, and in the Kruger National Park (KNP) the average mature male (660 kg) weighs ≈19% more than mature females (532 kg) (Pienaar 1969). Buffalo bulls are found in two social environments: ‘breeding herds’ ranging from 50 to over 1000 animals of both sexes and young animals, and smaller groups of bulls commonly termed ‘bachelor herds’ (Sinclair 1977; Prins 1996; Cross, Lloyd-Smith & Getz 2005). Bachelor herds tend to use more riverine habitats than mixed herds, thus the sexual segregation of African buffalo is habitat segregation (this study; Prins 1996; Sinclair 1977). Adult bulls may be encountered in both social contexts throughout the year, but predominantly associate in bachelor herds out of the breeding season (Turner et al. 2005). In the KNP, the majority of breeding occurs between March and May, with most calves being born between January and April (Pienaar 1969).
To assess the costs of sexual segregation, we monitored known individuals and directly estimated predation rates, rather than a surrogate index of risk, in different social environments. We then compared the locations of mortality sites with the habitats frequently used by bachelor and mixed herds. To assess the relative benefits of different social groups, we measured faecal nitrogen content as an indicator of forage quality, daily movement rates and body condition. We conclude with a discussion of the relative merits of the different hypotheses used to explain sexual segregation in African buffalo.
This study was conducted in the central and southern regions of the KNP between Satara (24°23′ S, 31°46′ E) and Lower Sabie (25°07′ S, 31°55′ E) rest camps, from November 2000 to January 2006. Rainfall patterns in the KNP show a decreasing gradient from south to north, with a mean annual rainfall of 612 mm at Lower Sabie rest camp and 548 mm at Satara rest camp (Gertenbach 1980). Most of the annual rainfall occurs between November and April. The KNP is underlain by granitic soils in the west and basaltic soils in the east. In the study region, granitic areas tended to have more woody vegetation and less grass biomass than basaltic areas (Venter 1990). The Satara region, from which the majority of the data were collected, contained four to 12 buffalo herds and roughly 3000–4000 buffalo during the study period. In KNP, mixed herds tend to be large (mean 225·7, SD = 186·0) with a mix of bulls, cows and juveniles. Bachelor groups almost exclusively contained bulls and ranged in size from one to 50 animals (mean 4·9, SD = 5·4; Whyte 2004).
Buffalo were anaesthetized using both helicopter and ground darting with a combination of etorphine hydrochloride (M99; Logos Avet, Johannesburg, South Africa) and azaperone tranquillizer (Stresnil; Janssen Pharmaceutica, Sandton, South Africa) (Bengis 1993). Anaesthesia was reversed by intravenous administration of diprenorphine hydrochloride (M5050, Logos Avent). Over the study period from 2000–06, VHF radio-collars (Telonics Corp, Mesa, Arizona) were fitted to 163 unique individuals (Fig. 2). For the survival analyses, we limit the data set to 74 females and 55 males that were 5 years or older when mortalities occurred. For the analysis of where mortalities occur, however, all the data were used. Attempts to locate radio-collared animals were made on a weekly basis during daylight hours throughout the study period. Those individuals not found after 6 weeks were relocated from aircraft. All captured individuals were aged according to incisor eruptions patterns up to 5 years of age (Grimsdell 1973). Buffalo first captured over the age of 5 were aged subjectively according to a photographic reference collection in Pienaar (1969) using patterns of wear on the horns, hair loss on the ears, head and body, and the presence and size of a dewlap in males. These age estimates were also calibrated according to individuals of known age in a neighbouring reserve (C. U. Knechtel, personal communication). Although some of our age estimates are subjective, we believe they are sufficient to remove age as a potential confounding factor in our analyses of predation risk in different social environments.
We used a Cox proportional hazards (CPH) approach to assess the mortality hazards associated with being in different social environments (Cox 1972; Therneau & Grambsch 2000). The CPH framework allows the baseline mortality hazard to vary freely over time, thus removing any confounding due to seasonal or annual fluctuations in mortality. The CPH approach also accounts for the left-truncation (resulting from delayed entry) and right-censoring (due to collar failure) present in the data set (Heisey & Patterson 2007). Buffalo age was modelled in three ways: a linear effect, a categorical effect (5–8 years vs. 9+), and by fitting a penalized spline that was chosen based on the lowest AIC value (Therneau & Grambsch 2000; Burnham & Anderson 2002). As mortality increases with age, and older bulls spend more time in bachelor groups, we were concerned about confounding. We investigated only models where both age and herd type were included. This represents a more conservative test of the effects of herd type, as the variation due to age was already accounted for. Gender and herd affiliation were confounded. Only very rarely were females observed in bachelor groups. So the effect of sex and herd affiliation was modelled using three categories: females, males in bachelor herds, and males in mixed herds. We excluded all cases where the last sighting was over 60 days after from the mortality event. We also tested for differential mortality of bachelor groups compared with mixed herds using more traditional χ2 analysis using the number of radio-collar days in each herd type to generate the expected frequencies. The results remained significant and did not change our conclusions, thus we present only the CPH analyses because they account more rigorously for variation in mortality rates due to age and time. In addition, we conducted two additional CPH analyses using only those mortalities that were within 20 and 30 days of the last sighting and our conclusions remained the same. All CPH analyses were conducted using r ver. 2·5·0 (R Development Core Team 2005).
Data on herd affiliation (bachelor or mixed herds) were collected by direct observation of radio-collared individuals. As both male and female buffalo have horns, it was impossible for a radio collar to come off an animal without tearing through the belting. Therefore we identified mortality events by the belting still being intact, as well as the presence of remains of a carcass, and other evidence such as blood, rumen contents and lion spoor. We assumed that a radio collar had fallen off if the belting was severed and no carcass was present after an extensive search of the area. In seven out of the 12 presumed cases of collar failure, the individuals were resighted later (all captured individuals were uniquely branded). Of the remaining five collar failures, two were males last seen in bachelor groups and three were males last seen in mixed herds. In many cases, we removed collars prior to collar failure. These collars were then refurbished and placed on new individuals. Based on previous studies (Mills & Biggs 1993; Prins 1996), we assumed that all predation events on subadult and mature buffalo were from lions.
In addition to directly estimating mortality rates, we also investigated the visibility associated with predation sites compared with habitats used by bachelor and mixed herds. We used the presence of rumen content and dried blood to pinpoint the mortality site. If these signs were not available, we noted the location of the remains of the carcass. Lion attacks are often more successful at short ambush distances (Van Orsdol 1984), and increasing grass height and vegetation cover is an important variable in hunting success (Stander & Albon 1993; Funston, Mills & Biggs 2001). Habitats that have good visibility should therefore be relatively safe, as they provide for early detection and evasion of predators, and may reduce the risk of attack (Cowlishaw 1997). To measure visibility, field assistants walked away from the observer until they were no longer visible up to a height of 1·5 m, which approximates buffalo eye-level (Pienaar 1969). This measure was repeated for each of the four main cardinal compass directions and then averaged. In addition to predation sites, visibility was also measured at randomly chosen sites used by bachelor and mixed herds. Visibility distances were transformed using the natural logarithm for a two-way anova with geological substrate type (basalt vs. granite) and herd type (bachelor vs. mixed herd) used as explanatory variables. The data were categorized by geological substrate type because of the influence of these components on the vegetation and consequently visibility. Basalt-derived soils produce a flat landscape with open grassland savannah, whereas granite-derived soils produce an undulating topography with more closed woodland savannah (Venter 1990).
As proxy for energy expenditure, we estimated the average distances moved per day by buffalo in different social environments. Individuals were located at noon and again at noon the following day. We then calculated shortest distance between the two consecutive sightings. We linked each bachelor group movement with a paired sample from the nearest breeding herd within the same month and geological substrate. We then used a two-tailed Wilcoxon signed-rank test for matched pairs to determine if the two herd types differed in the distance moved. We assume that this minimum distance moved is correlated with the total distance moved during the day, which depends on whether mixed herds walk more (or less) tortuous routes than bachelor groups. We also recorded the position where we located buffalo in the topographical profile and classified the following hill slope units after Venter (1990): crest, mid-slope, foot-slope, valley bottom and drainage courses.
To assess forage quality consumed by buffalo in different social environments, faecal samples were collected from buffalo herds between 2001 and 2003 to analyse the percentage of faecal nitrogen (Nf). The use of faecal indices as indicators of ruminant diet quality has a long history (Arman, Hopcraft & McDonald 1963; Sinclair 1977; Erasmus, Penzhorn & Fairall 1978) and has been shown to be a useful adjunct to existing methods for assessing range quality in South Africa (Grant et al. 2000). Grant et al. (2000) were able to correlate these indices with body condition scoring in buffalo and impala. In a concomitant study of the same buffalo herds, (Ryan 2006) found a strong relationship between Nf and percentage N in vegetation utilized by buffalo.
In 2001, buffalo herds across the KNP were sampled for Nf in addition to the focal buffalo herds. Ten samples were taken from each herd and the values averaged to obtain a mean Nf value for the herd. Only large faeces were sampled, as these were assumed to be from adult buffalo. For bachelor groups, which frequently contain fewer than 10 individuals, samples were collected only when they were clearly from separate individuals. Samples were collected within 12 h of deposition to ensure consistent measurement (Leite & Stuth 1994). In order to conform to veterinary requirements, the faecal samples were dried in a convection oven at 65 °C for 72 h. The samples were analysed for N content at the Institute for Tropical and Sub-Tropical Crops, a division of the Agricultural Research Council, using standard Kjeldahl techniques (AOAC 1975).
We analysed the Nf data using a three-way anova with herd type, year sampled and season (wet season = November–April, dry season = May–October) as the independent variables. Herd type and season were considered fixed effects, and year a random effect. A comparison of Nf from adult bulls in bachelor groups to random adults in breeding herds (male or female) assumes there is no difference between sex and age in terms of differential digestive efficiency, which may be expressed as a difference in Nf levels even when foraging on similar material. In order to validate this assumption, Nf levels were compared from animals of known sex and age obtained during large-scale capture operations by a two-way anova, using size (adult male, adult female and subadult) and herd sampled as independent variables. The results indicated that the three age and sex classes did not have significant differences in Nf levels within a herd (two-way anova: F2,12 = 1·48, P = 0·267). Therefore we assume that any differences between bachelor and breeding herds reflect a difference in forage quality rather than sex- or age-related differences in digestive efficiency. Analyses were conducted at the 95% significance level (α = 0·05) using statistica (1997; Statsoft Inc., Tulsa, OK, USA).
Body condition was assessed visually from field and video observations between 2001 and 2004 using the same five-point scale (1 = very poor, 5 = excellent) as used in earlier studies on buffalo (Prins 1996; Caron, Cross & du Toit 2003). We investigated the change in body condition from the late dry season (September–November) to the following breeding season (March–May) for individuals that spent the majority of their time in either bachelor or mixed herds. Body condition is strongly correlated with sex and age, so body condition analyses were limited to bulls that were older than 7 years. Condition scores were averaged for each individual over the relevant time period. Individuals were considered to associate in bachelor groups or mixed herds if more than 60% of their sightings were in that social environment for that specific season. For these analyses, we also included individuals that were not radio-collared, but were known from brands placed during captures and marker collars. Due to the categorical and non-normal distribution of the condition data, randomization tests were used to investigate the statistical differences in condition gain leading up to the breeding season for bachelor bulls vs. herd bulls (Manly 1991). For the randomization test, the average difference in condition gain between bulls in the two herd types was compared with the results of 50 000 simulations of the same data where the herd type was randomly reassigned to individuals. We considered the null hypothesis rejected if less than 5% of the simulations resulted in values greater than the actual data. Simulations were conducted in matlab ver. 7·4 (2007; Mathworks Inc., Natwick, MA, USA).
The counts of bachelor groups over the study period indicated a seasonal change in average group size (Fig. 1). Average group size was largest towards the end of the dry season between September and November, and declined with the onset of the wet season to reach a minimum between March and June. Courtship activities occurred between March and June, and peaked in April, which coincided with the lowest bachelor group sizes. There was a significant inverse relationship between courtship activities and bachelor group size (Spearman's rank correlation coefficient: r = –0·85, N = 12, P < 0·01). The peak period of courtship activity coincided with the time of year when average bachelor group size was at a minimum, and conversely, the period of least reproductive activity occurred at the time of year when bachelor groups were largest (Fig. 1).
Adult males frequently moved back and forth between mixed herds and bachelor groups throughout the study period (Fig. 2). Bulls tended to spend less time associating in breeding herds with increasing age (Kruskal–Wallis anova: H3 = 469·76, P < 0·001; Figs 2–3). Immature bulls almost always associated with breeding herds and were seldom found in bachelor groups. Young bulls (5–7 years) spent a small proportion of their time associating in bachelor groups, but were mainly found in breeding herds (Figs 2 and 3). This age group did not show any pronounced seasonal pattern to their affiliation with either herd type (Mann–Whitney U-test: U = 25521·5, P = 0·11, N = 498). Middle-aged (8–10 years) and older bulls (11 years and over) spent the least proportions of their time in breeding herds. For these older age categories, a seasonal pattern was evident where they tended to associate proportionally more in the mixed herds over the breeding season between March and June (Mann–Whitney U-test: U = 17044·5, P < 0·001, N = 460; Fig. 3). Unlike a previous study by Sinclair (1977), we observed very few older males that did not spend any time within mixed groups (Fig. 2).
We observed 38 adult buffalo mortalities during the study. Fifteen of 21 adult males that died were last seen in bachelor groups. Fifteen adult females died in mixed herds, and two were last seen alone prior to death. Two statistical models were tied for the lowest AICc value (Table 1). These models included age as a penalized spline, and either sex or herd type (Table 1). Sex and herd type could not be included in the same model due to confounding (no females died while in bachelor groups). The mortality hazard associated with being in a bachelor group appeared to increase over time (Fig. 4), violating the proportional hazards assumption, but the mean hazard for bachelor bulls was still significantly greater than for adult females (Fig. 4). In particular, the mortality hazard was almost four times greater for bachelor males compared with adult females (exp(β) = 3·79, 95% CI = [1·85, 7·79], P = 0·0003). While the mortality rates of adult males and females within breeding herds were not statistically different (exp(β) = 1·67, 95% CI = [0·62, 4·52], P = 0·31).
Table 1. Cox proportional hazards models using 38 deaths of 130 adult African buffalo in different social groups
AgeCat represented age as two categories (5–8 years and 9+).
Penalized spline fit to age. Degrees of freedom for the penalized spline were based on AIC.
Visibility at mortality sites, mixed herd and bachelor bull sightings was significantly related to herd type and geology (anova, substrate: F1,396 = 9·4, P = 0·002; herd type: F2,395 = 34·8, P < 0·0001; herd : substrate: F2,395 = 0·18, P = 0·83). Mixed herds used more open habitats than bachelor groups [Tukey's honestly significant difference (HSD) test, P < 0·001; Fig. 5]. Bachelor groups, however, used habitats with visibility similar to that of habitats where buffalo mortalities occurred (Tukey's HSD test, P = 0·605; Fig. 5). Finally, areas used by buffalo on basalt soils were more open than areas used on granitic soils (Tukey's HSD test, P = 0·006). If buffalo died in habitats similar to that in which they occurred regularly, one would expect that adult cows, which associate in breeding herds, would tend to be killed in areas of high visibility (more open areas) than would bulls. We found, however, that there was no difference with respect to the visibility at sites where males and females died (one-way anova, F1,55 = 1·62, P = 0·21). In addition to the differences in visibility, we also found that bachelor groups preferentially selected the bottomlands (valley bottoms and drainage lines) compared with breeding herds, which spent relatively more time on the crests P < 0·01).
Benefits of bachelor groups
Faecal Nf levels were associated with year, season and geological substrate (Fig. 6; Table S1 in Supplementary material). Using data from all seasons, there was no effect of herd type on Nf (Table S1). During the dry season, however, mean Nf on granitic soils for bachelor groups tended to be higher than for breeding herds, although the difference was significant only in 2001 (Tukey's HSD test, P = 0·005, Table S1; Fig. 6). On basaltic soils, average Nf for bachelor groups was higher than for breeding herds in 2001, but not significantly different (Tukey's HSD test, P = 0·320; Table S1), and similar to breeding herds in the subsequent 2 years. Thus bachelor bulls tended to have Nf levels similar to or higher than mixed herds during the dry season when nutrients are most limiting, which is the opposite pattern to what would be expected for the forage-selection hypothesis (Ruckstuhl & Neuhaus 2000; Post et al. 2001).
The minimum daily movements for bachelor groups were significantly less than for breeding herds (Wilcoxon signed-rank test, T = 13, N = 13, P < 0·02). On average, bachelor bulls moved 2·35 km (SD ± 2·21 km) per day compared with 3·33 km (SD ± 2·23 km) for animals in breeding herds. The reduction in movement and equal or higher forage quality appeared to increase the body condition of bulls in bachelor groups. Bulls that spent the larger portion of their time associating in bachelor groups over the late dry season (September–November) gained more body condition (0·77 ± 0·09 SE on a five-point scale) in the period up to the breeding season (March–May) than bulls that spent their time in mixed herds over the dry season (0·23 ± 0·1 SE, randomization test, P = 0·0043).
Adult male buffalo appear to trade increased risk of predation in bachelor groups in return for net energy gain. Our mortality data showed that bulls had almost four times the risk of predation in bachelor groups compared with adult females in mixed herds, while the predation rate of adult males and adult females in mixed herds was not statistically different. Bachelor groups tended to use riverine habitats that were similar in visibility to those areas where buffalo tend to be killed by lions, while mixed herds used upland areas with greater visibility. The low-lying areas used by bachelor groups tend to retain more moisture, which enables grasses to remain green for longer (Macandza, Owen-Smith & Cross 2004). Faecal Nf levels were equal or higher in bachelor bulls compared with mixed herds during the most nutrient-limiting season (Fig. 5; Table S1 in Supplementary meterial). Finally, bachelor bulls moved approximately 29% less per day than bulls in breeding herds. The reduced energy expenditure, combined with equal or better forage quality, translated into additional gains in body condition for adult males in bachelor groups compared with those in breeding herds. These data support the predation-risk hypothesis as an explanation for sexual segregation in buffalo. Males incurred greater risks in return for gaining body condition, which we assume leads to a competitive advantage in securing mating opportunities (Clutton-Brock, Guinness & Albon 1982), while females inhabited more secure habitats that are likely to improve calf survival.
To the authors’ knowledge, this is the first study on sexual segregation that directly measured predation in different social environments. We assumed that individuals were killed in the same social environment as that in which they were last seen. Allocating mortalities to the wrong herd type will reduce statistical power and the observed effect size, but is unlikely to change the direction of the effect. Of the 38 mortalities we used in the survival analyses, 15 were within 7 days of the last sighting, and 26 mortalities occurred within 20 days of the last sighting. We also conducted two further survival analyses on those mortalities that were within 20 and 30 days of the last sighting, and our conclusions remained the same. The model including herd type remained the best by AICc, and herd type remained a significant variable in all analyses.
Similarly to this study, Turner et al. (2005) also found that bachelor bulls increased grazing time and moved less than breeding herds. Sinclair (1977) also concluded that bulls in bachelor groups are more sedentary than their counterparts in breeding herds, and calculated that it could result in an energy saving of 4–7%. Our results are also supported by Prins (1996), who, using data from buffalo carcasses in Manyara, found that the chance of being killed dropped steeply with increasing distance from potential cover for lions. Our data on buffalo mortality sites demonstrate that habitats of lower visibility are more dangerous for buffalo. Further, bachelor groups used low-visibility habitats similar to where mortalities usually occur, whereas breeding herds were found in more open, safer habitats.
We did not explicitly test the activity-budget hypothesis, but we did observe an increasing amount of segregation with age (Figs 2 and 3), which is predicted by the activity-budget hypothesis, as males continue to increase in size even after sexual maturity (Pienaar 1969). Turner et al. (2005), however, found that even among females there was little activity synchrony, yet they were still able to maintain cohesive herds. More importantly, though, the activity-budget hypothesis does not explain why the buffalo bulls of this study would move to riskier habitats when they segregate from females. For this reason, we prefer the explanatory power of the predation-risk hypothesis, but recognize that it may not be exclusive in this system.
The forage-selection hypothesis as an explanation of sexual segregation has received varying support (Main et al. 1996; Ruckstuhl & Neuhaus 2002), and has not been specifically tested in buffalo. Our results, however, contradict the predictions of this hypothesis, as males obtained diets of quality similar to or better than those of females. This hypothesis may, however, not be independent of the predation-risk hypothesis. As pointed out by Main (1998), males with larger body size may have an advantage in terms of efficiency of digestion, and may be able to subsist more effectively by expanding their diet to include lower quality forage when necessary (Demment & Van Soest 1985). Therefore, when forage is equally poor across the landscape, males may still be foraging better than females. Lastly, the ‘social affinity’ hypothesis (Bon et al. 2001), which predicts that individuals of similar social motivation and behaviour will form cohesive units, does not provide an adequate explanation of the benefits that would be necessary in buffalo to offset the high predation rates in bachelor groups.
A meta-analysis conducted by Ruckstuhl & Neuhaus (2002) supported activity synchrony as an overarching explanation for sexual segregation and, based on previous studies, assumed that African buffalo supported the activity synchrony hypothesis. Our data show that males risked higher predation rates in bachelor groups in return for better body condition, as predicted by the predation-risk hypothesis. The activity-budget hypothesis also predicts that males are likely to gain more body condition when they are in same-sex groups, but fails to explain why males would incur greater predation risk. Studies on other species that directly estimate predation rates in different social environments may shed more light on this subject.
This research was supported by the US National Science Foundation (Ecology of Infectious Disease Grant DEB-0090323 to W.M. Getz) and the National Research Foundation, South Africa. Wayne Getz and Johan du Toit provided supervisory guidance. Martin Haupt provided logistical support. We are grateful to South African National Parks and Drs Ian Whyte, Markus Hofmeyr, Peter Buss and Roy Bengis for their support and for this opportunity to conduct this research in the Kruger National Park. Many thanks to Justin Bowers, Julie Wolhuter, Robert Dugtig, Kutani Bulunga, Augusta Mabunda, Fernando Muhlovo and Simeon Muhlovu for their help collecting field data. We appreciate the valuable comments by John Winnie, Anna Jolles, Mike Ebinger and two reviewers that improved previous versions of this paper.