Evolution of Multilevel Social Systems in Nonhuman Primates and Humans

A Socioecological Hypothesis for Formation of Multilevel Societies in Colobines

For higher-level social associations such as module-based bands to develop, ecological conditions should not have a limiting effect on grouping (within certain margins). First, resources must permit the formation of very large groups, i.e., an abundant and uniformly distributed resource base is required for the formation of bands (Fimbel et al. 2001; Rodman 1988). The staple foods of many multilevel colobines appear to be fairly abundant, e.g., lichens in Rhinopithecus bieti (Grueter et al. 2009; Kirkpatrick et al. 1998) or nonephemeral young leaves in Nasalis larvatus (Boonratana 1993; Matsuda et al. 2009), so the foraging or ranging costs imposed by assembling in bands are most likely relatively insignificant (Grueter and van Schaik 2010). Band size is controlled and constrained by the availability of resources and patch sizes; bands are larger in more temperate forests where foods seem to occur in larger patches (Mann–Whitney U, z = −3.01, P = 0.0027; Fig. 3; see also Kirkpatrick 1998; Kirkpatrick and Grueter 2010). The degree of terrestriality does not seem to have an effect on actual band size (Grueter unpubl. data). It could be argued that ecological conditions are not merely permissive, but actually drive the nested nature of the multilevel social system of colobines (Matsuda et al. 2010), but a detailed review (Grueter and van Schaik 2010) found no evidence for this. The hypothesis that predation has fuelled the formation of superbands is unlikely to apply; as group size benefit from predation quickly saturates (Hamilton 1971), we woud not expect groups of several hundred members, as commonly seen in snub-nosed monkeys (Grueter and van Schaik 2010).

Open image in new window

We consider the threat of conspecific males (bachelors) that form AMUs of substantial size and are frequently on the outskirts of the group in most multilevel settings (Grueter et al. in press-b; Yeager 1990) to be the most plausible shaping force of multilevel colobine societies. Specifically, we suggest that OMU holders seek one another’s proximity and thereby incur a lower probability of being challenged by bachelor males, either through joint defense or the safety in numbers effect (Bleisch and Xie 1998; Grueter and van Schaik 2010; Rubenstein and Hack 2004). Females would also indirectly benefit from a reduced takeover probability, as this would reduce the risk of infanticide. Increasing the number of coresiding OMUs appears to be functionally comparable to adding more males to a single group. Indeed, primate groups with more males have been shown to be less vulnerable to incursions by nonresident males (Crockett and Janson 2000; Janson and van Schaik 2000; Newton 1986; cf. Borries and Koenig 2000). Rubenstein and Hack (2004) showed that as the chance of being harassed by bachelor males increases, stallions exhibit a higher propensity to associate with other stallions to thwart these extra group hazards in plains zebras in multilevel societies. On the other hand, Fischhoff et al. (2009) have recently shown that stallion–stallion associations do not show long-term cooperative bonds.

That bachelors pose a significant threat to OMU leaders and females in modular colobines is supported by several lines of evidence: 1) AMUs of substantial size are more or less permanently on the outskirts of the reproductive group in most multilevel settings (Grueter et al. in press-b; Hoang 2007; Kirkpatrick 1998; Stanford 1991; Yeager 1990); 2) replacements of resident males are often accompanied by infanticide (Agoramoorthy and Hsu 2005; Ren et al. 2011); 3) the frequency of male aggression in a wild group of Rhinopithecus bieti was elevated when bachelors were present (Grueter et al. in press-b). Additional support for the bachelor threat influencing modularity/aggregation patterns comes from a comparative analysis on Asian colobines that showed that where the expected number of nongroup males is high, units have high home range overlap, show a higher association degree, and have a higher tendency to form bands (Grueter and van Schaik 2010). Researchers have only rarely reported observations that can be interpreted as collective or collaborative aggression of unit males against intruders in modular colobines (Grueter and van Schaik 2010), but some circumstantial evidence has been presented (Zhao and Li 2009). Krzton (2011) observed coordinated patrolling behavior among males of different breeding units in wild Rhinopithecus roxellana, vaguely reminiscent of chimpanzee boundary patrols (Wilson and Wrangham 2003). In conclusion, males in modular societies may balance the costs and benefits of associating with potential allies and competitors, and the benefit of enhanced safety when associating with conspecifics may outweigh the costs and have prompted the formation of bands. An interesting case is the Semnopithecus population at Jodphur/India where there are solitary OMUs as well as AMUs (Koenig and Borries 2001). Even though the OMU males have to defend their females against takeovers from bachelor males, OMUs do not form a higher social level. It could be that the local ecological conditions (semidesert) do not permit modularity in this case.

Once modularity has been established, female transfer among units within the band may create kinship and tolerance/affinity bridges between units, thus favoring the maintenance of a multilevel system in colobines (sensu Chapais 2008). However, data concerning genetic relationships among females in modular colobine societies are yet to be collected. Interunit ties may be further strengthened by infant handling involving females of different units (Zhang et al. 2012).

A Socioecological Explanation for Evolution of Multilevel Societies in Papionins

Modularity in papionins seems to have evolved from ancestral mm–mf groups of considerable size (see earlier). We outline two scenarios, the ecological and the social model, of how large groups might have become substructured. The ecological model assumes that splitting was caused by ecology/environment and social factors came subsequently into play. This model has been generally applied to hamadryas baboons. The alternative, the social model, which we favor here, posits that it was mainly social factors that triggered substructuring, but that ecology was a necessary precondition. An alternative to the ecological model is the time constraint model, and an alternative to the social model is the social brain model, both of which provide other possible explanations for substructuring in papionins.

The Social Model

The following hypothetical scenario can be envisaged for hamadryas baboons (Grueter and Zinner 2004; Zinner et al. 2001). Large and relatively cohesive social groupings appear for ecological reasons, e.g., reduction of predation risk at scarce sleeping sites and possibly rare waterholes acting as a magnet for units (localized resources and predation avoidance hypotheses). Another scenario for how hamadryas groups could have become so large is given by Jolly’s (2009) frontier hypothesis: during northward extension of range of ancestral baboons, neighboring groups were few and distant in those frontier areas, thus favoring male philopatry. The resulting risk of not finding suitable, unrelated males in the natal group may then have promoted the development of very large social groups. In any case, these large social assemblages bear a cost in terms of enhanced aggression potential and begin to partition for social reasons. The danger of being harassed or attacked by an unfamiliar individual might be greater than in a normal sized mm–mf group of savanna baboons. In particular, the presence of a large number of unfamiliar males may pose a threat of sexual coercion (especially when females are in estrus) and possibly infanticide for females. In large groups with a large number of unfamiliar individuals, social stress may reduce fecundity in females. In such a situation, other females may not be effective coalition partners and a female–female social network would not represent the best solution to this problem. Female–female relationships may be weakened, and females fare better when they join a guarding male (bodyguard and mate guarding hypothesis), thereby reducing coercion by unfamiliar group members, especially by males. This would trigger a shift from female bonding to cross-sex bonding (sensu Byrne et al. 1990), which might be an exaggeration of the “friendship” relationships between males and females in other baboon taxa (Palombit 1999, 2009; Smuts 1985). An initially or ancestrally mixed mm–mf group becomes divided into modules (OMUs) with reproductive control for the male. The weakly bonded female network makes it easier for males to establish autonomous units (Barton 2000). Ultimately, it is the demographic change, i.e., an increase in aggregation size due to ecological needs with increasing risk of harassment, which forces females into small stable OMUs. Above a certain size, baboon groups are apt to become substructured. Figure 4 shows that groups (bands) of multilevel papionins (Papio hamadryas, Theropithecus gelada, most likely also P. papio) are larger than groups of mm–mf baboons (P. anubis, P. cynocephalus, P. ursinus; Mann–Whitney U, z = 3.61, P = 0.0003).

Open image in new window

Dunbar (1986, 1988, 1993) envisaged a similar model for the evolution of present day gelada society out of a Papio-like mm–mf group. However, contrary to the hamadryas pattern, in this model it was female bonds that were generated as a means to reduce stress in large gelada groups. Group size increased as an adaptation to open country/plains where these primates faced higher predation risk. Large social groupings or aggregations on a permanent or regular basis (grasslands), which are determined by various ecological needs, such as predation avoidance or optimal habitat use and foraging (localized resources and predation), may result in an insecure social environment. It is then conceivable that large groups split into OMUs because females started to form clusters with closely related females within a larger group, and later a male attached himself to them. Dunbar argues that increased aggression and reduced reproductive output associated with growing group size resulted in females tending to bond together into small matrilineal groups for mutual protection (coalitions) to buffer themselves against the stresses and harassment imposed on them by living in large groups. The alternative, that higher levels of sociality in geladas evolved via a merger of OMUs cannot be discounted, but does not seem to fit well with the phylogenetic relationships. Ohsawa (1979) proposed that lack of antagonistic relations of OMUs with distinct “home areas” permitted amalgamation into fluid bands.

Our portrayal of the evolution of the hamadryas society is derived from Dunbar’s gelada scenario with an important difference: Dunbar highlights the gradual intensification of female–female bonding in geladas, whereas we highlight male–female bonding. This difference reflects the different allocation of social effort within these societies. However, what we need to explain is why strong male–female bonds have not become standard in geladas and why gelada females usually do not select a male as a coalition partner in the face of stress and harassment. It may be that in the beginning the small number of hamadryas females clustering around a male were relatives, but that at a later stage in the evolution of the hamadryas system, males integrated unrelated females into these clusters, breaking up the female network within the unit (Swedell and Plummer 2012). It is also worthy of mention that the gelada system is not exclusively founded on female bonds; some females also form close bonds with individual males (Dunbar 1983c). Nevertheless, the disparate expression of social effort in hamadryas baboons and geladas remains a conundrum requiring further intellectual and empirical scrutiny.

The Phylogenetic Evidence

Chapais (2008, 2010) discussed likely scenarios for the evolution of the human multifamily system. Based on the evolutionary principle of parsimony, he argued that the ancestral hominin species (the last common ancestor of chimpanzees and humans, LCA) was most likely characterized by a chimpanzee-like mm–mf group composition exhibiting a polygynandrous mating system, female dispersal, male philopatry, and male kinship bonds. Several other essays on early australopithecines/hominins advocate a similar system (Foley and Lee 1996; Ghiglieri 1987; Lovejoy 2009; McHenry 1996; Wrangham 1987b). The multifamily system of humans can then be seen as a derived trait, having emerged through a transition from polgynadry to stable breeding bonds at some point after the Pan–Homo split (Fig. 5). This scenario is referred to here as the ancestral male kin group hypothesis (Chapais 2008). It is reminiscent of the separation pathway suggested for modularity in baboons.

Open image in new window

A second putative evolutionary pathway proposes that the ancestral hominin system was gorilla-like, with a unimale–multifemale group composition, and that the human multifamily system arose from the amalgamation of several such independent polygynous units into a multifamily group, as in our coalescence pathway. A similar socioevolutionary scenario has been put forward by Imanishi (1965) and by Foley and Lee (1996), who considered single gorilla-like OMUs to represent the ancestral state and modular groups with OMUs (hence the multifamily system) the final state, with an additional intermediate state of mm-mf groups (for a slight variant of this see also Geary and Flinn 2001). However, a number of lines of evidence, besides the phylogenetic argument, support the ancestral male kin group hypothesis.

The Anatomical Evidence

McHenry’s assertion (1992) that early hominins (australopithecines) were anatomically more similar to chimpanzees than to gorillas in terms of absolute body size and relative brain size supports the ancestral male kin group hypothesis (Chapais 2008). The similarity in body mass suggests that early hominins had a diet comparable to that of chimpanzees, which in turn implies that the behavioral ecology, i.e., group size, group composition, ranging patterns, etc. of early hominins may have been chimpanzee-like. While some argue that “the … behavior of early hominins is … unlikely to represent simple amplifications of those shared by with modern apes” (Lovejoy 2009, p. 74), we believe that studies of extant species coupled with strategic modeling, referential models and cladistic analysis can generate crucial information about behavior of extinct hominins (Whiten et al. 2009).

Further evidence for the ancestral male kin group hypothesis comes from strontium isotope analysis of tooth enamel in hominins. The strontium contained in plants is incorporated into the teeth of mammals ingesting them during the period of enamel mineralization so that the strontium composition of teeth in mature individuals is a likely marker of the geographical location where those individuals grew up (Copeland et al. 2011). The analysis of tooth enamel in 19 South African australopithecines dated around 2 Ma suggests that 50 % of the small individuals (probably females) dispersed from their natal range, whereas 90 % of the large individuals (probably males) did not (Copeland et al. 2011). This is compatible with the male philopatry pattern of chimpanzees but not with the dispersal pattern of gorillas where both sexes disperse.

Finally, data on sexual dimorphism in early hominins are compatible with both a mm–mf social system and an OMU-based modular system. It has long been recognized that pronounced body size dimorphism in pre-Homo fossil hominins is suggestive of high levels of male–male competition (Leigh 1995; Plavcan and van Schaik 1997; Wolpoff 1976). McHenry (1992, 1996) provides data suggesting that body mass dimorphism in early hominins ranged from comparable to a chimpanzee in Australopithecus robustus to quite gorilla-like in A. boisei. Skeletal dimorphism in Australopithecus afarensis is moderate, greater than in Pan and similar to that in Gorilla (Gordon et al. 2008; cf. Reno et al. 2003). White et al. (2009) argued that the primitive Ardipithecus ramidus had minimal body size dimorphism, but postcranial fossil evidence is fragmentary. Sexual dimorphism tends to be highest in OMU-based groups, e.g., gorilla, and in modular societies (Grueter and van Schaik 2009) and intermediate in mm–mf groups, e.g., chimpanzee (Plavcan 2001). The relatively high degree of sexual dimorphism in body size in early hominins could be reconciled with either an OMU-based modular social system or a mm–mf system. However, using dimorphism to infer behavior in early hominins is complicated by the unique combination of minimal canine size dimorphism, indicating a lack of male–male competition, and strong body mass dimorphism, consistent with intense male–male competition (Plavcan 2001; Plavcan and van Schaik 1997).

The Paleoecological Evidence

Several lines of evidence, such as isotopic composition of soil samples and teeth, signify that partly forested environments, particularly woodlands, represented the main habitats for late Miocene hominins (Elton 2008; WoldeGabriel et al. 2001), and therefore possibly for the LCA of the chimpanzee/bonobo–human clade, which has been tentatively dated at 6–8 million years ago (Steiper and Young 2006), but for which there is still no clear taxonomic/fossil candidate. Most Australopithecus and Kenyanthropus (4.2–3.0 mya) and Paranthropus (3.0–2.5 Ma) probably lived in rather open mosaic habitats, with open woodlands, bushlands, riverine forests, and seasonal floodplains (Reed 1997; Zazzo et al. 2000).

It is generally thought that group size became considerably larger when hominins began occupying more open savannah environments (Foley and Lee 1989; see also Moore 1992), because grouping reduces the probability of predation in open environments (Dunbar 1988). As the habitat of these hominins was becoming increasingly dry and open, valuable food items were dispersed. Underground storage organs such as roots, tubers, bulbs, and corms have been suggested as possible key components of the diet of early hominins, especially robust australopithecines (Laden and Wrangham 2005; Peters and O’Brien 1981). Under such environmental conditions, the spatial cohesiveness of groups would be compromised and the group would become prone to substructuring (Aureli et al. 2008; Foley and Gamble 2009) (dispersed resource hypothesis). They would separate during the day to reduce intragroup feeding competition and increase net food intake. This reminds us of the hamadryas strategy (ecological model) where ancestral mm–mf troops are hypothesized to have split into smaller foraging units in response to ecological conditions. The smallest units of the foraging parties may have been were lone individuals (as is sometimes the case among human foragers), labile parties or stable subunits, as in hamadryas. It is unlikely that foraging parties were made up of lone individuals because of the considerable predation threat (Hart and Sussman 2009). Stable subunits are better buffered against predators whereas temporary parties are better buffered against food scarcity/patchiness.

A different ecological force, i.e., localized resources or limited refugia in the increasingly patchy savannah habitat, may have promoted return use of certain areas for sleeping, drinking/feeding or safety, e.g., riverine strips, valleys, sleeping cliffs, water holes etc., and led to band formation. In this scenario, all members of the band would be forced to gravitate toward such locations (Aureli et al. 2008) and develop some sort of “agreement” to share sites (see Stammbach 1987 and Sueur et al. 2011 for hamadryas). These sites would have acted as “information centers” where information about food sources, etc. could be exchanged and cooperative bonds reinforced (Aureli et al. 2008). Frequent use of highly localized resources may ultimately have led to the adoption of home-base sites and “central place foraging” in hominins (Layton and Barton 2004; Marlowe 2006; Moore 1996; Rose and Marshall 1996). Once intermale competition became more relaxed, in line with a reduction in canine size dimorphism, and a wider range of habitats were occupied in the evolutionary transition from Australopithecus to Homo, daily fusing may have become especially beneficial and fully established. The existence of home bases has been inferred from local concentrations of stone tools, raw stone material, and animal bones (Isaac 1978; cf. Binford 1980; Potts 1984).

The Sociosexual Evidence

Ecological parameters probably acted in tandem with sociosexual pressures in the creation of stable families within larger social units, as seen in humans and prehumans. Chapais (2008) argued that the transition from a promiscuous mm–mf society to the humanlike community of mainly monogamous families was not direct, as is often assumed (Fisher 2006; Kaplan et al. 2000, 2009; Lovejoy 1981, 2009), but that an intermediate stage was the multiharem type of structure in which all reproductive units are polygynous, as seen in extant hamadryas baboons and geladas. The evolution of human pair-bonding is thus hypothesized to have taken place in two steps. The first step involved a shift from the sexually promiscuous mm–mf group to the multiharem group: from sexual promiscuity to polygyny. One possibility is that this shift occurred for similar reasons as substructuring in baboons, i.e., through the interplay of the tendency of males to monopolize females and the effect of ecological constraints on the spatial distribution of females. Specifically, male monopolization potential is higher when females forage in small (widely spaced) groups, which is itself caused by the density and spatial dispersion of food (dispersed resource hypothesis and mate guarding hypothesis). Stable breeding bonds in hominins likely resulted from tradeoffs between male sexual competition and female feeding constraints, as has been demonstrated for other mammals. An alternative possibility for the transition from sexual promiscuity to a multiharem structure is the bodyguard hypothesis, which posits that females establish a long-lasting sociosexual (pair) bond with a particular male that can act on behalf of the female as a “bodyguard” or “hired gun” to counter coercion (sexual harassment, infanticide) from outside conspecific males (Blurton Jones et al. 2000; Mesnick 1997; Palombit 1999; Rodseth and Novak 2006; Smuts 1992; van Schaik and Dunbar 1990). Paradoxically, male mate guarding can sometimes take the form of sexual aggression such as rape (Emery Thompson 2009; Smuts 1992). Wrangham (2009) and Wrangham et al. (1999) proposed a credible scenario that sees the male–female pair-bond as a reciprocal arrangements in which males can count on a woman for cooking and providing an evening meal at camp and females rely on male protection from theft of gathered and cooked foods. Females may associate closely not only with a protective male, but also with related females for the purpose of cooperative infant rearing (Hrdy 2009; Swedell and Plummer 2012).

Whatever the exact process involved, several lines of evidence support an intermediate polgynandry-to-polygyny sequence rather than the direct polgynandry-to-monogamy sequence. First, many male primates monopolize more than one female, this resulting in a preponderance of polygynous mating systems and an underrepresentation of monogamous systems (Fuentes 1999). Second, polygyny is the rule in all other multilevel primate societies; modularity based on monogamous pairs has not been documented in nonhuman primates. Presumably this is because the spatial cohesion of females in large mixed-sex groups always makes polygyny feasible (Chapais 2011a). Accordingly, the various models discussed in the preceding text in relation to the evolution of multilevel societies in papionins (ecological model, time constraint model, social model, social brain model) have in common that they predict the fragmentation of large mm–mf groups into polygynous units, not into monogamous ones. Third, the high degree of sexual dimorphism in body size in early hominins argues against monogamy in these species (see earlier). Fourth, polygyny is practiced in the majority of human societies, which suggests that it was an integral part of the human species’ evolutionary past, if not the ancestral hominin mating system (Chapais 2008). Fifth, the view that monogamy evolved directly from polgynandry typically associated with the idea that monogamy evolved in response to an increase in the dependency of children and to selective pressures for the evolution of paternal investment and male provisioning (Fisher 2006; Kaplan et al. 2000, 2009; Lovejoy 1981, 2009). Such a view, however, runs counter to our knowledge of the evolution of monogamy and paternal care in mammals in general (Chapais 2008). Specifically, phylogenetic analyses of mammalian mating and parental care systems indicate that paternal investment evolved after pair-bonding was already established (Brotherton and Komers 2003; van Schaik and Kappeler 2003). The role of paternal care in the evolution of pair bonds has also been questioned on other grounds (Hawkes 2004).

The second step in the evolution of the human mating system was the transition from the multi-OMU structure to the multimonogamous family structure in which most families are monogamous and some are polygynous. It is noteworthy that this sequence markedly reduces the number of scenarios and explanations concerned with the origin of monogamy because it implies that monogamy evolved through a reduction in the number of females monopolized by the average male, in other words, as a result of the evolution of constraints on polygyny (Chapais 2008). One possibility, labeled the provisioning hypothesis, explains the transition in terms of the benefits of male provisioning for mothers and offspring by positing that it became progressively more advantageous for a mother to be the single beneficiary of a male’s provisioning effort. For males to reduce the size of their family units and forego mating opportunities for paternal investment, we must assume a substantial reduction in levels of sexual competition between males. This hypothesis therefore predicts generalized monogamy (little polygyny) in human societies (Chapais 2011a). Another possibility, the leveling hypothesis, states that after the evolution of lethal weapons —a uniquely human feature that equalized the competitive abilities of males— it became much more costly for males to monopolize more than one female. The consistent exclusion of a large fraction of males from the pool of reproductive individuals, as observed in, e.g., hamadryas baboons, would have become prohibitive and would likely have given way to some sort of polygyny–monogamy mix, with most males being monogamously mated and just a few males forming polygynous units. The importance of polygyny in human societies is more compatible with the leveling hypothesis (Chapais 2011a). The majority of human societies allow polygyny (Murdoch 1981), but its frequency depends on subsistence style: monogamy predominates in forager societies, whereas pastoralists and agriculturalists show significant polygyny (Kaplan et al. 2009). So, although polygyny would antedate monogamy evolutionarily, polygyny could reappear secondarily after the adoption of a system based on agriculture and land tenure which generated socioeconomic inequities (Chapais 2011a; Fisher 1989).

Based on the foregoing arguments we propose the following scenario for the evolution of modular societies in hominins. The LCA initially lived in large mm–mf groups with fission–fusion. Substructuring into polygynous units evolved because it minimized feeding competition during the day in a patchy environment, allowed males to adopt a stabilizing reproductive strategy (mate guarding hypothesis), and made it possible for females to retain a protector male (bodyguard hypothesis). Moreover, male monopolization of isolated units was also facilitated in a habitat of dispersed resources (also mate guarding hypothesis). LCA parties/subunits were prone to coalesce around localized drinking/feeding areas and scarce refugia in an open habitat (localized resource hypothesis and predation avoidance hypothesis), factors that promoted band formation. Polygynous subunits subsequently evolved into mostly monogamous units owing to the evolution of constraints on polygyny. Pair-bonding might have later operated as a preadaptation for the evolution of paternal care and the sexual division of labor in response to the evolution of protracted juvenile dependency and high costs of maternity (Chapais 2011a).