INTRODUCTION
Complex, hierarchical social systems, termed multi-level societies, are present in species from many distantly related taxa, such as birds (Papageorgiou et al., 2019), cetaceans (Whitehead et al., 2012), equids (Rubenstein & Hack, 2004), proboscideans (Wittemyer, Douglas-Hamilton, & Getz, 2005), primates (Grueter, Matsuda, Zhang, & Zinner, 2012), and chiropterans (Kerth, Perony, & Schweitzer, 2011). Determining why these types of societies evolve and how they function are key questions in biology. In multi-level societies, stable subgroups (hereafter core units) associate in a hierarchical manner (Grueter, Qi, Li, & Li, 2017). Up to four tiers of non-random association have been documented, with higher tiers numbering hundreds of individuals in some species (Grueter, Matsuda, Zhang, & Zinner, 2012; Snyder-Mackler, Beehner, & Bergman, 2012; Schreier & Swedell, 2012a; Wittemeyer et al., 2005). The factors that determine the number and composition of different social tiers and the ways that ecological and social pressures affect their stability are still poorly understood for most species (Grueter et al., 2017; Farine et al., 2015).
Several hypotheses have been proposed to explain the function of different tiers in multi-level societies. The advantages and disadvantages of group living for animals have been well documented (Krause & Ruxton, 2002) and multi-level societies appear to have evolved because they allow animals to adjust group size more fluidly than is possible in stable groups (Aureli et al., 2008; Grueter et al., 2017). Large aggregations are beneficial, primarily because of the multiple ways that they lower predation risk (i.e., detection, dilution, predator confusion, defence, Hamilton, 1971; Pulliam & Caraco, 1984), while the chief cost of large group size is the increase in food competition that results from many conspecifics together (Terborgh & Janson, 1986). Indeed, there are examples of multi-level societies forming large aggregations at higher tiers when predators are nearby (e.g., Papio hamadryas , Schreier & Swedell, 2012b;Physeter microcephalus , Whitehead et al., 2012), and fissioning to lower tiers when resource availability is reduced (e.g.,Loxodonta africana , Wittemyer et al., 2005; Orcinus orca,Foster et al., 2012; Papio hamadryas , Schreier & Swedell, 2012b;Rhinopithecus roxellana , Qi et al., 2014). An alternative explanation for the evolution of multi-level societies, the bachelor threat hypothesis (Rubenstein, 1986), has garnered support, particularly in zebras and primates (e.g., Rubenstein & Hack, 2004; Pappano, Snyder-Mackler, Bergman, & Beehner, 2012; Xiang et al., 2014). This hypothesis suggests that core units (which are one-male/multi-female (OMUs) in many species) associate to decrease the amount of harassment they receive from extra-unit males (Grueter & van Schaik, 2009; Rubenstein, 1986).
In addition to predation, food competition, and conspecific threat, the degree that animals associate can be influenced by breeding seasonality (Baden, Webster, & Kamilar, 2016), migration (Colbeck et al., 2013), genetic structure (kinship) (Reisinger, Beukes, Hoelzel, & de Bruyn, 2017), and phylogeny (Balasubramaniam et al., 2018). Some of these factors fluctuate over time, often in a predictable, seasonal manner that can lead to adjustments in association patterns as animals try to balance the costs and benefits of group living. Thus, to understand if and how temporal environmental changes influence tier formation in multi-level societies, it is important to study social behaviour over time, both within and between the social tiers (Grueter, Chapais, & Zinner, 2012). Primates are good study subjects to examine the various factors influencing tier formation and stability in multi-level societies because researchers are often able to collect detailed observational data on many identifiable individuals.
Our goal was to determine the degree of temporal variability in inter-unit associations in a recently discovered multi-level society of Rwenzori Angolan colobus monkey (Colobus angolensis ruwenzorii ), and to determine which social and ecological factors may influence this variability. This multi-level society, described for our population at Lake Nabugabo, Uganda, is unique among primates in that it contains not only OMUs, but also core units that are multi-male/multi-female (MMUs), with up to eight socially-integrated, reproductive males (Stead & Teichroeb, 2019). There are at least three tiers of social organization; core units fission and fuse with one another throughout the day but associate preferentially with core units from the same clan and clans share a home range in a band tier of organization. Initial cluster analyses with one year of data revealed two clans in our study band of 12 core units (Stead & Teichroeb, 2019). We do not yet know the selective pressures that influence tier formation or if tiers are stable over time.
In this study, we applied social network analysis (SNA) to core unit associations observed over 21 months at Nabugabo. SNA has proven extremely useful in elucidating the form of multi-level societies (e.g., Papageorgiou et al., 2019; Snyder-Mackler et al., 2012; Zhang, Li, Qi, MacIntosh, & Watanabe, 2012) and determining how within-unit interactions structure them (Matsuda et al., 2012). Further application of SNA to understanding the associations between the higher tiers in multi-level societies offers a view of the dynamics of these complex social systems (Yeager, 1991). We first examined whether clan-level groupings of C. a. ruwenzorii core units stayed the same over time. Unlike many other social orders of mammals, primates tend to form stable groups over relatively long periods of time that are structured by the bonding of the philopatric sex (Altmann et al., 1996; Silk, 2001; 2002; Di Fiore, 2012). These kin-based systems extend beyond the smallest social unit to higher tiers in primate multi-level societies (Colmenares, 2004; Snyder-Mackler, Alberts, & Bergman, 2014; Morrison, Groenenberg, Breuer, Manguette, & Walsh, 2019). Preliminary data thus far suggest that though both males and females disperse from their natal core unit in C. a. ruwenzorii , and males transfer into other core units within the band while most females emigrate out of the band (Stead & Teichroeb, 2019). Thus, male kinship could structure relations between core units, leading to stable clans over time. Second, we examined temporal variation in core unit clustering relative to ecological (seasonality in rainfall and food availability) and social (inter-unit dispersals) conditions. Greater overall food availability should allow larger aggregations to form because food competition is alleviated (e.g., Foster et al., 2012; Schreier & Swedell, 2012b; Wittemyer et al., 2005). This temporal clustering of groups due to resource availability may then allow individuals to assess dispersal opportunities in other groups (Clobert, Le Galliard, Cote, Meylan, & Massot, 2009; Mares, Bateman, English, Clutton-Brock, & Young., 2014), which could potentially lead to a seasonal pattern in dispersals (e.g., Ekernas & Cords, 2007; Yao et al., 2011; Young et al., 2019). In addition, groups that have recently had individuals disperse between them may associate more closely during these periods compared to when no transfers are occurring; likely because it takes time for bonds among former group members to sever and to establish bonds in the new group (Isbell & Van Vuren, 1996).
Accordingly, we hypothesized that both ecological and social variables (i.e., male relationships) would influence clan stability and the degree of clustering among C. a. ruwenzorii core units. We predicted that clans would be stable over time, due to tolerance of close proximity between core units that share male kin or familiar males. Fruits are the only food type positively selected for and these are high in quality compared to the rest of this species’ diet, which consists largely of young leaves (JAT, unpubl. data). We therefore predicted that core unit clustering would be influenced most strongly by the availability of fruits, and that core units would associate more during times of high fruit availability when food competition is minimized. We further predicted that greater clustering of core units at times of peak fruit availability, would allow males to assess the possibilities for immigration in other core units, leading to a seasonal pattern of dispersal between core units. Finally, we predicted that associations between core unit dyads would be more frequent during and immediately after male transfer between them.