Intra-specific variation in social organization of gorillas: implications for their social evolution
- First Online:
- Cite this article as:
- Yamagiwa, J., Kahekwa, J. & Basabose, A.K. Primates (2003) 44: 359. doi:10.1007/s10329-003-0049-5
- 407 Views
We analysed intra-specific variation in the social organization of gorillas and ecological and social factors influencing them, based on recent data on diet, day journey length, home range size, group size and proportion of multi-male groups in three subspecies [western lowland gorillas (WLG); eastern lowland gorillas (ELG); mountain gorillas (MG)]. Median group size was similar across subspecies and across habitats, but the extraordinarily large group including >30 gorillas was only found in habitat with dense terrestrial herbaceous vegetation. Within-group competition may determine the upper limit of group size in frugivorous WLGs and ELGs in lowland habitats with scarce undergrowth. A frugivorous diet may be a causal factor of subgrouping in multi-male groups of WLGs and ELGs, while a folivorous diet may prevent subgrouping in multi-male groups of MGs. Social factors, rather than ecological factors, may play an important role in the formation of multi-male groups and their cohesiveness in MGs. High gregariousness of female gorillas and their prolonged association with a protector male are explained by their vulnerability to both infanticide (MGs) and predators (ELGs). Comparison of long-term changes in group composition and individual movements between ELGs in Kahuzi and MGs in the Virungas suggest that the occurrence of infanticide may promote kin-male association within a group. Threat of infanticide may stimulate MG females to transfer into multi-male groups to seek reliable protection and maturing MG males to stay in their natal groups after maturity. By contrast, the absence of infanticide may facilitate ELG females to associate with infants and other females at transfer and ELG males to establish large groups in a short period by taking females from their natal groups, by luring females from neighbouring groups, or by takeover of a widow group after the death of its leading male. These conditions may prevent ELG and WLG maturing males from remaining to reproduce in their natal groups and possibly result in a rare occurrence of multi-male groups in their habitats. Similar reproductive features of MG and ELG females suggest both female strategies have been adaptive in their evolutionary history.
KeywordsGorillaSubspeciesDietGroup sizeMulti-male group
Demographic and life history parameters are important when considering intra- and inter-specific variation in the social organization of the African great apes in relation to ecological factors (Goodall 1986; White 1996; Boesch 1996; Doran and McNeilage 1998, 2001; Sugiyama 1999; Yamagiwa 1999). However, our knowledge of the social organization of gorillas has been limited to a single population of mountain gorillas (MGs, Gorilla gorilla beringei). Most data on their life history come from long-term records of the Karisoke Research Center in the Virungas.
Mountain gorillas (MGs) form a cohesive group consisting of one or more adult males and several females with their offspring (Schaller 1963; Fossey 1983). Both males and females tend to emigrate from their natal groups around maturity, and only females immigrate into other groups or join solitary males to form new groups (Harcourt et al. 1976; Harcourt 1978). Unlike females, who do not travel alone, males lead a solitary life or join all-male groups after emigration until they get female partners (Caro 1976; Yamagiwa 1986, 1987a). The high cohesiveness of a group is due to the attractiveness of the leading male to females, who are usually unrelated and rarely have physical contact with each other (Harcourt 1979; Stewart and Harcourt 1987). Ecological features of MGs are considered to support their social organization. Their folivorous diet may reduce feeding competition and contribute to their high degree of aggregation (Fossey and Harcourt 1977; Watts 1984). It may cause the low site fidelity of females and the extensive overlap of home ranges between groups to facilitate female transfer (Robbins 1995; Watts 1996).
However, recent findings on the ecological features of western (WLG; Gorilla gorilla gorilla) and eastern lowland gorillas (ELGs; Gorilla gorilla graueri) show large differences between their diet and that of MGs. WLGs exhibit a frugivorous diet, as do chimpanzees, and regularly feed on ants and termites (Sabate Pi 1977; Tutin and Fernandez 1992, 1993; Nishihara 1995). ELGs also show a frugivorous diet seasonally in both the lowland tropical forest and the montane forest (Yamagiwa et al. 1994, 1996a).
How do such ecological variants influence the social organization of WLGs and ELGs? Harcourt et al. (1981a) pointed out that the median group size of WLGs, five, was significantly smaller than that of ELGs and MGs, i.e. nine. WLGs tend to form less cohesive groups than MGs (Tutin 1996) and occasionally form subgroups (Remis 1994, 1997; Goldsmith 1996). Within-group scramble competition may stimulate female WLGs to spread to forage and within-group competition may result in stronger female philopatry and more differentiated female relationships (Doran and McNeilage 1998). Subgrouping may possibly be caused by the limited availability of fruit, as observed in chimpanzees and bonobos. However, these perspectives are based on insufficient data and are still anecdotal.
Recently we have analysed long-term data on several groups of ELGs who have been habituated since the 1970s in the Kahuzi-Biega National Park, Democratic Republic of Congo. Information on birth rate, infant mortality, age at first parturition, inter-birth interval and migration of individuals can now be compared to those of MGs. Ecological surveys on ELGs in both lowland and montane forest have been conducted to elucidate ecological factors influencing their social organization. The aim of this paper is, therefore, to illustrate intra-specific variations in gorilla social structure, using recent information from ELGs for comparison with both WLGs and MGs, and to discuss the causal factors of such variations.
We made vegetation surveys using a line-transect method (Yamagiwa et al. 1993a, 1996b) to estimate the diversity and density of tree species in 1989 (Itebero: lowland) and in 1994 (Kahuzi: highland). Tree density and species diversity in the Itebero region were about 3 times higher than those in the Kahuzi region (Yamagiwa et al. 2003).
The mean annual rainfall from 1994 to 2000 at the meteorological station (1,600 m above sea level) was 1,586 mm (range 1,409–1,809) with a distinct dry season in June, July and August, in which the mean rainfall was <50 mm. The mean monthly temperature was 20.1°C (mean maximum 26.4°C; mean minimum 13.2°C).
Gorillas inhabiting the Itebero and Kahuzi regions are classified as the same subspecies (Gorilla g. graueri) because the two populations have had genetic exchange until recently as a consequence of the corridor between these regions. Both populations are also sympatric with chimpanzees (Pan troglodytes schweinfurthii). The estimated densities of gorillas are 0.81–1.78 gorillas/km2 calculated from the belt transect method in the Itebero region (Hall et al. 1998) and 0.43–0.47 gorillas/km2 calculated from counting the number of nests in fresh nest sites (up to 2 days old) in the Kahuzi region (Yamagiwa et al. 1993b).
Ecological features of three subspecies of gorillas
Gorilla gorilla gorilla
G. g. graueri
G. g. beringei
Lowland tropical forest
Lowland tropical forest
High montane forest
Altitude (above sea level)
Number of types of plant food (species)
182 (134) a
194 (121) j
129 (79) m
72 (44) o
75 (38) p
230 (129) b
% Fruit in plant food species
71% a, 69% b
Fallback food during fruit scarcity
Marantaceae, Zingiberaceae, aquatic plantsc, d
Cyperaceae, herbs, bark of woody vinesl
Herbs, vines (whole year) o
Herbs, vines (whole year) p
Mean day journey length
1,100~2,600 me, f
500~1,000 mq, r, s
Annual home range (km2)
7~14 g, i
4~11 t, u
Home range overlap
Extensivee, f, g, h
The seasonal fluctuation in food abundance is larger for frugivorous WLGs and ELGs than folivorous MGs, who experience no seasonal changes except during the period of bamboo shoots. Fruit occupies zero to more than half of WLG's diet, according to fruit availability at Lopé and Bai Hokou (Tutin et al. 1997; Remis 1999). The monthly mean number of fruit species per fecal samples changes from almost zero to more than four for both WLGs at Ndoki and for ELGs at Kahuzi (Kuroda et al. 1996; Yamagiwa et al. 1996). By contrast, fruit constitutes <1% of MG's diet during the whole year in the Virungas (Watts 1984; McNeilage 2001). During periods of fruit scarcity, WLGs increase their consumption of foliage and terrestrial herbaceous vegetation (THV) (Kuroda et al. 1996; Tutin and Fernandez 1993; Remis 1997; Remis et al. 2001). They frequently eat some forms of aquatic herbaceous vegetation, which are high in proteins and minerals, in the Bai swamps or clearings (Nishihara 1995; Magliocca and Gautier-Hion 2002). During periods of fruit scarcity, feeding of ELGs on bark of various trees and woody vines tends to increase (Casimir 1975; Yamagiwa et al. 1996b).
Annual home ranges of WLG and ELG groups are larger than those of MG groups. The home ranges of three WLG groups were 7–14 km2 annually at Lossi (Barmejo 1997). From our data on ranging of an ELG group for 8 years at Kahuzi, the mean annual home range is 14.1 km2 (SD=2.0), which is larger than the mean annual home ranges of two MG groups (group 5, n=7, mean=11.5 km2, SD=2.5, t=2.26, P<0.05; and group Nk, n=5, mean=10.2 km2, SD=2.4, t=3.17, P<0.01) monitored at the Virungas (Watts 1998). The ELG study group increased their range every year and the total home range for 8 years was 42.2 km2, which is larger than the total home ranges of MG groups in the Virungas [group 5, 24.3 km2 for 7 years; group Nk, 25.1 km2 for 5 years (Watts 1998)] and WLG groups [21.7 km2 for 10 years at Lopé (Tutin 1996); 23 km2 for over 27 months at Bai Hokou (Remis 1994)]. The seasonal shift in range caused by seasonal fluctuation in food availability may result in a larger annual range of ELGs, as suggested by Casimir and Butenandt (1973). Like MGs, the home range of WLGs and ELGs extensively overlap with those of neighbouring groups (Table 1).
Gorillas inhabiting montane forests show a shorter DJL than gorillas inhabiting lowland forests. First, there was no significant difference in DJL between MGs of Virungas and ELGs of Kahuzi. The mean DJL measured between consecutive night nests of an ELG group at the montane forest of Kahuzi was 716 m (n=109, range 242–2,055 m, SD=308), which was within the range of MG group's DJL at the Virungas (group 5, n=29, mean=1,034 m, range 345–2,868 m, SD=633; group Nk, n=32, mean=472 m, range 112–1,160, SD=257; and group Sz, n=33, mean=681 m, range 182–2,484, SD=493) (Yamagiwa 1986). It was significantly shorter than group 5 (Mann Whitney U-test, Z=2.119, P<0.05), significanly longer than group Nk (Z=4.417, P<0.001), and equivalent to group Sz (Z=1.747, NS).
Second, the mean DJL for WLG groups inhabiting lowland tropical forests was longer than both DJL of MG and ELG groups inhabiting montane forests (Table 1). The mean DJL reported for ELG groups in the lowland forest of Itebero (n=8, mean=1,531 m, range 142–3,439, SD=408, from Yamagiwa and Mwanza 1994) is significantly longer than that of the ELG group at Kahuzi (Z=2.225, P<0.05). The mean DJL for WLG groups at Bai Hokou [n=95, mean=2.6 km, range 0.3–5.3, SD=1 (Goldsmith 1999)] is also far longer than that of the ELG group at Kahuzi (t=19, P<0.0001). The patchy distribution of fruits constitutes the major factor leading to a longer DJL. Both WLGs and ELGs tend to increase their DJL when they shift their diet from foliage to fruit (Yamagiwa and Mwanza 1994; Goldsmith 1999). It may result in an increase in the home range and in frequent inter-unit encounters.
Group size and composition
Group size of three subspecies of gorillas in various habitats. WLG Western lowland gorillas, ELG eastern lowland gorillas, MG mountain gorillas
No. of groups
Median group size
% Multi-male group
Magliocca et al. (1999)
De Mérode et al. (2001)
Yamagiwa et al. (1993b)
Weber and Vedder (1983)
McNeilage et al. (1998)
The maximum group size is not different among the three subspecies (Kruskal Wallis, H=3.32, P=0.19), or between montane and lowland forests (Mann Whitney U-test, Z=−1.36, P=0.17). However, the extraordinarily large group was found under limited environmental conditions. Bermejo (1997) reported a group of 32 gorillas at an abandoned village of Lossi, Congo. Excluding this case, the maximum group size (weaned individuals) of WLGs was <20 at any habitat. We also found that the maximum group size of weaned ELGs in the lowland forest (Itebero) was 17. In the montane forest of Kahuzi and the Virungas, the maximum group size occasionally exceeds 20 (Schaller 1963; Murnyak 1981; Yamagiwa et al. 1993b; Steklis and Gerald-Steklis 2001). Yamagiwa (1983) reported a group of 44 ELGs in Kahuzi, and Steklis and Gerald-Steklis (2001) found a group of 41 MGs in the Virungas including infants. These data suggest that a large group including >20 gorillas may be formed in montane forests or secondary regenerating forests with rich undergrowth.
The striking difference between subspecies is found for group composition. In most of the WLG populations, groups are polygynous (Table 2). Excluding the population of Lopé, in which the proportion of multi-male groups was not reported, Table 2 shows that the proportion of multi-male groups in MGs is significantly higher than in ELGs (df=1, χ2=16.61, P<0.0001) and WLGs (df=1, χ2=2.06, P<0.0001). The proportion of multi-male groups for ELGs and WLGs did not differ (df=1, χ2=1.58, P=0.21). We found ten groups in the lowland forest at Itebero and all groups included only one silverback. The two censuses on the ELG population in the montane forest of Kahuzi found <10% of groups contained two silverbacks (Murnyak 1981; Yamagiwa et al. 1993b). By contrast, the multi-male group accounted for 30–46% of the MG populations of both Bwindi and the Virungas (Yamagiwa 1987a; McNeilage et al. 1998; Robbins 2001). Schaller (1963) reported a group containing four silverbacks, and a group including five silverbacks has been found recently in the Virungas (L. Williamson, personal communication).
Among five habituated groups of ELGs in Kahuzi, two groups shifted from one-male to multi-male composition (Fig. 2). This type of group composition lasted for <3 years in all cases and then shifted to a one-male or all female composition. By contrast, the multi-male composition of the Virunga MG group lasted >10 years (Robbins 2001). The two silverbacks of a ELG group are usually a father and a maturing son in Kahuzi, while brothers or unrelated silverbacks occasionally associate to form a MG group in the Virungas (Yamagiwa 1987b; Robbins 1995). All-female groups are only found in Kahuzi, whereas all-male groups have been reported only in the Virungas. Neither type of unisexual group has ever been reported for WLGs.
Subgrouping rarely occurs in MG groups at Virunga (Harcourt 1978; Fossey 1983), while it occasionally occurs in WLG groups at Bai Hokou (Remis 1997). Although multi-male groups were rare in WLG, most of the subgrouping occurred in multi-male groups during the fruiting season (Goldsmith 1996; Remis 1997). Although no evidence of subgrouping was found in polygynous groups of ELG at Itebero, frequent subgrouping occurred in an ELG multi-male group during the fruiting season at Kahuzi (Yamagiwa et al. 2003). These results suggest that a frugivorous diet may facilitate subgrouping in multi-male groups of WLGs and ELGs.
Based on the long-term records of habituated groups of gorillas, individual movement patterns are comparable between ELGs in Kahuzi and MGs in the Virungas. Among 14 males who had been followed from birth in the habituated groups of ELGs in Kahuzi, all males emigrated from their natal groups before reaching 15 years old (Yamagiwa and Kahekwa 2001). After emigration, males tend to spend a solitary life for months or years and eventually acquire females to form new groups. This process of new group formation is similar to that of the Virunga MGs (Harcourt 1978). However, only four (36%) out of 11 MG maturing silverbacks emigrated before maturity (Robbins 1996). The difference in the proportion of male emigration before maturity is significant between Kahuzi and Virunga (df=1, χ2=9.42, P<0.01). The males remaining in their natal groups after maturity started reproduction at an earlier age than the males emigrating before maturity (Robbins 1995; Sicotte 2001). In three cases, these natal males became the leading males after their putative fathers died. There was one known case of the same process of group succession by a natal ELG male in Kahuzi (Yamagiwa and Kahekwa 2001).
Group takeover by an extra-group male to oust the leading silverback has never been observed in either population. Although a blackback male was observed to join a reproductive group once in the Virungas (Fossey 1983; Robbins 1995), such immigration of maturing males has never been observed in Kahuzi. Maturing silverbacks emigrated from their natal groups with females and it caused group fission in two cases in Kahuzi. One case of group fission, in which two brothers separated from each other, has been reported recently in the Virungas (Robbins 2001).
The patterns of movements of ELG females between groups are basically similar to those of MGs. At Kahuzi, the median age of first emigration (or transfer) among eight subadult and adult females who had been followed from birth to the first emigration was 9 years old. Females who emigrated as immatures with mothers are excluded from this calculation. The median age of eight MG females at the first transfer was approximately 8 years old in the Virungas (Harcourt 1978). At Kahuzi, among 43 ELG females who had immigrated into the four habituated groups, ten females (23%) emigrated again. Six out of these 10 females stayed with the groups for >5 years. A second transfer is also common in MG females in the Virungas (Watts 1991). At least 28 transfers by parous females were observed, and many of these females were multiparous. Some females transferred at least 5 times.
A remarkable difference was found between the two populations in individual movements after the death of the leading male. ELG females continued to associate and travel without any silverback for the prolonged period after the death of a leading male. The maximum length of such all-female group formation was 29 months, and in the other two cases lasted for >1 year. Solitary males and maturing silverbacks from neighbouring groups occasionally visited these all-female groups, and one of them finally joined as a new leading male. By contrast, most of the MG females dispersed to join solitary males or neighbouring groups after the death of the leading male in four cases in the Virungas (Stewart and Harcourt 1987; Watts 1989). The MG females tended to transfer with other females or infants in such cases, and males to whom they transferred killed most of the infants. After disintegration of a group, maturing MG males remained together to form an all-male group (Yamagiwa 1987b; Robbins 1995). Solitary males and maturing males from neighbouring groups joined them and unrelated males associated together for years without participating in reproduction.
Female reproduction and transfers
ELG a (Kahuzi)
Minimum age at first observed copulation
Number producing their first infant in non-natal group
Number producing their first infant in natal group
Number of females emigrating from the natal group after the first birth
Age at first parturition
10.6 years (9.1–12.1)
10.1 years (8.7–12.8)
Interval between surviving births
4.6 years (3.4–6.6)
3.9 years (3.0–7.3)
Interval between surviving births (male infant)
Interval between surviving births (female infant)
Interval between death of infant and the next birth
2.2 years (1.4–2.7)
1.0 years (0.0–3.1)
Sex ratio at birth (number of males/number of females)
The ELG females experienced their first parturition when 10.6 years old on average. This is similar to the MG females in the Virungas (Table 3). The intervals between the surviving births of nine ELG females was 4.6 years, which is slightly longer than those of the MG females. No difference was found in the interval when the former infant was male or female for both ELGs and MGs (ELG, male, n=5, range 3.7–6.6; female, n=4, range 3.6–6.6; MG, male, n=11, range 3.0–7.3; female, n=15, range 3.0–5.1). When the newborn infant died, the next parturition occurred 2.2 years (n=3, range 1.4–2.7) after the death in ELG females. This is longer than the interval calculated for MG females in the Virungas (1.0 years, n=15, range 0.9–3.1), but the sample size is still rather small for a consideration of statistically significant differences.
The sex ratio at birth from data on 64 ELG infants in Kahuzi, who were born between 1978 and 1998, is close to 1:1 (31 males vs. 33 females). The MG sex ratio at birth in the Virungas is also similar (26 male vs. 33 females) (Watts 1991). No significant difference was found in the sex ratio at birth for both Kahuzi and Virunga (df=1, χ2=0.09, NS). The infant mortality is higher for primiparous ELG females (n=21, 33.3%) than parous ELG females (n=25, 20.0%), as observed for MG females in the Virungas (primiparous, n=14, 42.9%; parous, n=45, 17.8%) (Watts 1991), although the difference is not significant (df=1, χ2=0.47, NS for Kahuzi; df=1, χ2=2.45, NS for Virungas). The infant mortality is also higher in the first year than in the subsequent year for both ELGs (n=46, first year, 19.6%; second year, 6.5%) and MGs (n=65, first year, 26.2%; second year, 7.7%), but the difference is not significant (df=1, χ2=2.40, NS for Kahuzi; df=1, χ2=2.53, NS for Virungas). In the Virungas, 37% of infant mortality was due to infanticide (Watts 1991).
The maximum number of surviving offspring in the ELG females that we monitored was three in Kahuzi. The maximum number of births recorded for a single MG female is eight; three of the offspring survived infancy in the Virungas (Watts 1991). Given a reproductive life of females of 25 years and 26% infant mortality with 4.6 years interbirth (viable) interval, an ELG female might produce three to five offspring that survive to adulthood in Kahuzi. A MG female might produce four to six offspring. These data suggest that the population for both ELGs and MGs should be increasing. Actually both populations experienced a slight increase when they were well protected (Yamagiwa et al. 1993b; Aveling and Aveling 1987; L. Williamson, personal communication). However, transfer delays the initial age of parturition in both Kahuzi and Virungas (Harcourt et al. 1981b; Watts 1991; Yamagiwa and Kahekwa 2001). Death of leading males resulted in the prolonged ELG female association without reproduction in Kahuzi and caused MG female dispersal with associated infanticide in the Virungas. These negative factors, mostly consequences of poaching, may reduce the number of surviving offspring and hence the size of future generations in both populations.
Ecological factors influencing gorilla social structure
Socioecological models of female gregariousness and social relationships have explained a variety of social structure in primates (Wrangham 1980; van Schaik 1983; Dunbar 1988; Sterck et al. 1997). For example, Sterck et al. (1997) classified gorillas in dispersal-egalitarian groups, as females regularly transfer between groups, forming neither stable linear hierarchies nor coalitions. Female gregariousness may be determined by opposing pressures from predation and/or infanticide risk and from food distribution. The distribution of males may depend on a combination of female gregariousness and female choice of protector males.
This study indicates a quite similar social structure among three subspecies of gorilla, in spite of considerable differences in their diet. In contrast with the prediction of Harcourt et al. (1981a), the group size of WLGs is not smaller than those of ELGs and MGs. No difference is found in median group size between lowland habitats and montane habitats. Only the maximum group size is different among habitats. A large group including >30 individuals was only found in the secondary regenerating forest (Lossi) and in montane forests with a dense undergrowth of herbaceous vegetation (Kahuzi, Bwindi and Virungas). These data suggest that THV may mitigate within-group feeding competition and a folivorous diet may enable gorillas to form a large group. Socioecological factors other than dietary constraints may influence the group size of gorillas inhabiting forests with dense THV, as observed in other folivorous primates (Janson and Goldsmith 1995).
Frugivory and sparse distribution of fruits may stimulate WLGs and ELGs to disperse during foraging. In fact, subgrouping occasionally occurs in WLGs at Bai Hokou (Remis 1997; Goldsmith 1996) and ELGs at Kahuzi. In such cases, each subgroup usually involved at least one silverback. In contrast, subgrouping was scarcely observed in Mbeli Bai (Parnell 2002) and Itebero (this study) where no group contained more than two silverbacks. These data suggest that female gorillas always need to associate with at least one silverback male irrespective of their diet.
In contrast, subgrouping rarely occurs in MGs, which frequently form a multi-male group (Schaller 1963; Harcourt 1978; Fossey 1983). Social factors, such as infanticide, may prevent MGs from subgrouping and enhance their cohesiveness in multi-male groups (Watts 1989, 1996). MG females are known to usually transfer alone, and to disperse with other females when their leading silverback dies (Harcourt 1978; Stewart and Harcourt 1987; Sicotte 2001). They tend to transfer from small to large groups containing two or more silverbacks, where scramble competition is high but more protection available. Female group size is positively related to the number of males per group (Robbins 1995).
The distinct differences in the proportion of multi-male groups between MGs and other subspecies may support these interpretations. No evidence is yet available on infanticide for WLGs or ELGs. The long-term records on the life history of ELGs in Kahuzi show that ten females carried suckling infants when they transferred between groups, and no infant was killed or wounded in any of the cases (Yamagiwa and Kahekwa 2001). Two or more females occasionally emigrated or immigrated into habituated groups, and females formed all-female groups after the death of their leading silverback (Fig. 2). Due to the absence of infanticide, ELG females may not have an urgent need to seek a protector male after the death of their leading silverback. As ELG females may not compete for reliable protection, they tend to transfer together between groups.
Since multi-male groups may not confer an advantage on females in the form of a reduced risk of infanticide in Kahuzi, ELG females may not choose multi-male groups to join at transfer. Thus, maturing males do not need to remain in their natal groups. ELG males also emigrate with females from natal groups to form new groups. As multi-female transfer occasionally occurs, males can establish large groups in a short period by taking females from their natal groups, by luring females from neighbouring groups, or by taking over a widow group after the death of its leading male. These conditions may prevent ELG and WLG maturing males from remaining to reproduce in their natal groups and possibly result in the rare occurrence of multi-male groups in their habitats.
Several explanations are applicable for the absence of infanticide: (1) kin relations of infants with the males (potential killers) to whom their mothers transferred, (2) sex (female) of infants, as females may be a potential mate for the males in the future, and (3) the lack of solitary males (potential killers) in the population. However, none of them fully explain its absence (Yamagiwa and Kahekwa 2001). It seems likely that infanticide is a common feature of gorillas and has been suppressed as a potential reproductive tactic of males in Kahuzi. Although there are still insufficient observations from which to identify the causal factors of infanticide, it is important to note that the absence of infanticide promotes the emigration of males from their natal groups and prolonged female-association through transfer between groups and after the death of the leading male.
Then, why do WLG and ELG females show high gregariousness and usually associate with males, instead of travelling alone or forming temporary parties? The presence of a predator may prevent them from dispersal during travelling and sleeping on the ground (Stewart and Harcourt 1987; Yamagiwa 2001). ELGs usually nest on the ground across various types of vegetation and seasons. However, after the death of the leading silverback, females prominently increased the tree nest (Yamagiwa 2001). They resume nesting on the ground after a new silverback joins their group. These observations suggest that female gorillas avoid sleeping on the ground without a protector male. Female gorillas may be vulnerable to large terrestrial predators, such as leopards, which occasionally have been reported to attack or prey on WLGs (Fay et al. 1995: Watson 2000). WLG and ELG females may not need to seek the protection of a male against infanticide, but may seek it against a predator. Schaller (1963) reported that leopards had killed MGs until the early 1960s. However, leopards are no longer present in the Virungas and infanticide may be have more influence on the association of MG females.
Dietary differences between MGs and other subspecies (WLGs and ELGs) may have less influence on the production of marked differences in the basic social structure of gorillas. The ecological cost of female transfer may be negligible, and vulnerability to terrestrial predators may result in high gregariousness of females with a protector male. Extensive overlap of the home range with neighbouring groups may facilitate female dispersal and association with unrelated conspecifics. By contrast, infanticide may play an important role in changing the social organization of gorillas. The threat of infanticide may limit the decision of females to transfer when they have infants and promote the formation of multi-male groups as a consequence. The absence of infanticide may enable female association at transfer or after the death of a protector male and may stimulate the emigration of maturing males to form polygynous groups. Such changes in social organization may occur over relatively short periods.
Population structure and social evolution of gorillas
Large similarities in reproductive features between ELG and MG populations possibly suggest that both types of social organization may have been adaptations in their evolutionary history. Both populations are characterized by female dispersal and a polygynous mating system. Although less is known about the movements of WLG females, they seem to transfer between groups (Tutin 1996; Parnell 2002). The presence of solitary males and the one-male group composition suggest the emigration of WLG males from their natal groups. Exclusive mating is the main tactic of adult males for all subspecies.
The prolonged male-female association is promoted by two female strategies: to seek male protection against predators (ELG) and against infanticide (MG). In both ELG and MG, the leading males show long tenure, i.e. even more than 10 years, and group takeover by extra-group males to oust the leading male has never been observed (Fossey 1983; Robbins 1995; Sicotte 2001; Yamagiwa and Kahekwa 2001). However, only the strategies of MG females used to avoid infanticide may promote kin-male association within a group.
A population of gorillas consists of various types of social units (solitary males, all-male groups, all-female groups, one-male/multi-female groups, multi-male/multi-female groups) with extensively overlapping home ranges. Ecological factors, such as group density, home range size and DJL, may influence the rate of inter-unit encounter and thus affect the chance of female transfer (Yamagiwa 1999). Social factors, such as infanticide, may influence female choice of transfer and kin-male association. The different combinations of these ecological and social factors may produce variations in the social organization of gorillas.
The distributions of gorillas discussed here are found within those of chimpanzees, except for the higher montane forest of the Virungas. It is puzzling why they have not extended their distribution like chimpanzees. There are many suitable habitats for gorillas where only chimpanzees are currently surviving, such as forests in Uganda. This study shows that not only dietary constraints but also social factors have prevented them from achieving this. Although data on the population structure of WLGs are still insufficient, this study may contribute to an understanding of the social evolution of the genus Gorilla and to wise planning of strategies for gorilla conservation.
This study was financed by a grant-in-aid for the center-of-excellence research Evolution of Apes and the Origin of Human Beings (project head: O. Takenaka) and by the International Scientific Research Program (No. 12375004 to J. Yamagiwa) sponsored by the Ministry of Education, Science, Sports and Culture, Japan. It was conducted in cooperation with the Centre de Recherches en Sciences Naturelles and Institut Congolais pour Conservation de la Nature. We thank Prof. O. Takenaka, Prof. T. Matsuzawa, Dr S. Bashwira, Dr B. Baluku, Mr M.O. Mankoto, Mr B. Kasereka, Mr L. Mushenzi, and Ms S. Mbake for their administrative help; Dr M. Matsubara for digitising the daily travel routes of the study group on the vegetation map. We are also greatly indebted to Mr K. Kaleme, Mr M. Bitsibu, Mr S. Kamungu and all guides, guards, and field assistants at the Kahuzi-Biega National Park for their technical help and hospitality throughout the fieldwork.