THE SOCIAL BRAIN HYPOTHESIS proposes that a major factor driving the evolution of large brain size and enhanced cognitive capacity in primates is social complexity, for which the maximum size of organized groups is used as a proxy (Dunbar 1993). In line with our fission-fusion pattern of social organization (see Layton & O'Hara ch. 5 this volume; Lehmann et al. ch. 4 this volume), human groups occur in a variety of characteristic sizes rather than being smoothly distributed along a continuum from small to large (Caporael & Baron 1997; Dunbar 1998; Zhou et al. 2005). Attention to the different sorts of ‘adhesives’ that hold groups together can provide insight into why the apparently strong constraints limiting the size of some kinds of groups do not apply to groups held together with different sorts of ties. Freed of many of the environmental constraints that limit community size for hunter-gatherer societies living in harsh environments (see Layton & O'Hara ch. 5 this volume), some groups can grow to huge sizes, while others stay small. The architecture of cohesion helps explain why.

Social psychologists distinguish between the dyadic ties that connect two individuals (interpersonal relationships), and the connection between an individual and a group (group identification or social identity). They also distinguish between the strong bonds of socio-emotional cohesion and the more flexible adhesives of exchange or task cohesion (Evans & Dion 1991; Stokes 1983). Combining these two variables yields four adhesives for forming groups: strong and weak dyadic ties and primary and secondary group identification. After explaining in more detail the multiple sources of cohesion as understood by social psychologists, I will provide a fuller picture of the group prototypes these give rise to. The mix of strong and weak adhesives and the greater scalability offered by the (p.270) member-group bond in particular has facilitated the assembly of large scale societies that are orders of magnitude larger than ancestral human groups, without creating unmanageably large demands on the social brain.

The focus of this chapter is on how groups of three or more people who are not necessarily close kin are structured and held together. In small-scale hunter-gatherer societies and in ancestral bands and tribes, most members of a band are kin, whether close or more distant. Family members also form an important part of the personal networks of modern humans (see Roberts ch. 6 this volume), as do the pair bonds (see Dunbar ch. 8 this volume) that are embedded in and generative of families. The family ties of kin relations and mates are clearly fundamental to human communities.

Social psychologists studying groups, however, have largely ignored kin groups and kin relations, and mating and marriage are not generally studied in relation to larger groups. Research into groups has either composed groups from scratch in order to manipulate variables such as task, group size or leadership structure, or it has focused on the existing nonkin groups that are ubiquitous in communities, the workplace, sports, social networks and the military. While social network studies reveal more of the private social sphere (Roberts ch. 6 this volume; Wallette ch. 7 this volume), studies of existing groups in modern industrial societies tend to ‘find’ groups in more public and accessible settings.

It is abundantly clear that genetic relatedness promotes cooperative relations and—in many mammalian species—acts of altruism that benefit both direct offspring and other kin (e.g. Mateo & Holmes 1997). However, genetic relatedness falls off sharply as one moves from siblings to first cousins to more distant relations, and thus the glue of inclusive fitness is not by itself sufficient to construct massively large and complex societies. Read (ch. 10 this volume) notes a limit of around 500–800 individuals for social organizations based on an extended kinship system. A next step for categorization beyond relatedness is universal kin classification (see Barnard ch. 12 this volume) by which pairs of people are classified as related based on, for example, sharing the same name. This generalization falls into the realm of non-kin-based social adhesives that is the focus of the present chapter.

Briefly sketched, the four group prototypes chosen to illustrate different architectures of cohesion are cliques, coalitions, comrades and colleagues. Cliques are dense clusters of individuals linked by strong, interconnected dyadic bonds. Coalitions, which also depend on dyadic (p.271) ties, are strategic alignments that improve the relative power positions of the individuals involved (de Waal & Harcourt 1992). Comrades and colleagues are examples of groups that do not depend on dyadic ties. Comrades are united by collective action, joint membership and shared commitment to the group. Groups of colleagues are drawn together by common interests and shared social identity (for example, academics interested in the social brain). Dyadic relationships that develop among comrades and colleagues can be strong and enduring, but they are not fundamental building blocks for such groups.

These four prototypes should not be considered an exhaustive list of non-kin-based groups. Rather they constitute an informal typology designed to illuminate structural differences in cohesion that affect the cognitive loads on group members and the consequent group size limits. In other words, it is a typology created for the niche defined by this book. Existing typologies and category systems in group and organizational studies (e.g. Arrow & McGrath 1995; Hackman 1990) are less useful for this purpose as they tend to distinguish groups based either on the specific task they perform (e.g. top management groups or customer service teams) or on the different models used by managers and supervisors in creating work groups.

The rest of the chapter is organized as follows. First I discuss some different aspects of self and identity, social attraction and group cohesion. The relative importance of different forms of attraction and cohesion in different groups helps explain how they can be cohesive without having close interpersonal bonds among members. After establishing this groundwork, I consider the four exemplar group types in more depth, and discuss how the last two types in particular provide a mechanism for cohesion in very large organizations and societies.

The Many Aspects of Self and Identity

The self has been broadly defined by social psychologists as including the totality of thoughts about one's self (Trafimow et al. 1991). It is generally understood to be a complex psychological structure that includes a multiplicity of ‘selves’, some of which are more explicitly social than others. Distinctions are made between the private, public and collective self (Triandis 1989); the independent and interdependent self (Markus & Kitayama 1991); and the personal, relational and collective self (Brewer & Gardner 1996). As Barnard (ch. 12 this volume) might put it, some of (p.272) our selves stop at the skin, and some do not. One approach to studying the self is to have people complete the sentence ‘I am’ multiple times and then classify the statements as belonging to different aspects of self. Statements such as ‘I am tall’ reference the personal or independent self; ‘I am Susie's mom’ refers to the relational self, and ‘I am American’, ‘I am a Methodist’ or ‘I am a wide receiver for the Broncos’ are aspects of the collective self. As these examples suggest, the relational self is most relevant to dyadic interpersonal relationships, while the collective self involves membership in and identification with a larger group.

Beyond these distinctions, research on self-concepts and self-esteem provides abundant evidence that the self is a special and privileged psychological structure that we are prone to enhance and that evokes strong favouritism. For example, even fleeting connection with the self makes objects more special and valued. A host of experiments have demonstrated a robust ‘endowment’ effect: the tendency to value ‘my cup’ more highly than other cups, even if has only been ‘my’ cup briefly, and to demand a higher price to sell it than to purchase an otherwise identical cup (Kahneman et al. 1991). This effect has clear implications for any relationship or group that is incorporated into the relational and collective self. The ‘stickiness’ of the self and our corresponding attachment to anything connected with the self provides a powerful adhesive that helps relationships and groups stay together.

Social identity theory focuses primarily on the collective self in the context of inter-group relations (Hogg et al. 2004). Social identity is a person's knowledge of belonging to a social group, including demographic categories such as European or women; it is enhanced by contrast with ‘outgroups’, whether these comprise a general category (e.g. men) or a more specific group (e.g. a rival football club). When people think of themselves as group members their feelings and actions are biased in favour of the group in the same way that the self is generally accorded special privilege in our thoughts, feelings and actions.

Forms of social Attraction

Attraction among fellow members of a group has long been recognized as a source of cohesion. It can take several forms. Interpersonal attraction is positive feeling between individuals independent of whether they are in a relationship or a group. It is promoted by inter-individual similarity (Hogg & Hardie 1991) and other bases of interpersonal attraction, such (p.273) as physical attractiveness. Depersonalized attraction (Hogg & Hains 1996) describes attraction between individuals based on shared category membership. It is similar to the much older notion of solidarity among people who feel they are in similar circumstances. Hogg and colleagues (Hogg & Hains 1996; Hogg & Hardie 1991) have demonstrated that depersonalized attraction may be distinguished from interpersonal attraction both in cause and effects. It is also stronger the more a person is seen as prototypical of the category or group, as depersonalized attraction is derivative: the primary attraction is to the group, and the more prototypical a member is, the stronger the association between that person and the attractive group.

Group members may feel and act upon loyalty to one another based on depersonalized attraction regardless of whether they particularly like one another personally. Whereas dyadic bonds are promoted and cemented by interpersonal attraction, groups based on membership rely instead on depersonalized attraction.

A third form of social attraction, inter-member attraction, denotes positive feelings based on satisfying role relations within a group context (Brewer & Gardner 1996). It is grounded in interaction.

Group Cohesion

A broad and often ambiguous concept (Hogg 1987; Mudrack 1989), ‘cohesion’ originally referred both to a quality of group structure such as compactness (Festinger et al. 1950) and to the totality of forces that act on members to keep them in the group (Festinger 1950). This confusing use of the term as both an attribute and a process continues to resist resolution (Brawley et al. 1987; Drescher et al. 1985; Keyton & Springston 1990; Stokes 1983). However, despite the lack of a consensus definition and it has become clear that cohesion has many sources and is also a multidimensional construct. For our purposes the most important distinction is that between socio-emotional cohesion and task cohesion (Tziner 1982).

Socio-emotional cohesion is typically measured by asking group members how much they like one another, enjoy socializing and are friends. It has obvious connections with interpersonal attraction, and addresses needs for intimacy and belonging (Baumeister & Leary 1995). Task cohesion refers instead to attraction to the group's activity and effective integration of members in pursuing the group task (Carron et al. 1985). A third type of cohesion rarely mentioned in the literature is (p.274) exchange-based cohesion (Thibaut & Kelley 1959), which is based on satisfaction with an exchange relation.

Cliques: Clusters of Close Interpersonal Ties

Drawing on these distinctions, cliques evoke the relational self and are held together by dyadic bonds, interpersonal attraction, and socio-emotional cohesion. Knowing what keeps a group together also reveals what will threaten the group's integrity. In a friendship clique, disruptions of dyadic ties are traumatic not just for the two people involved, but for the whole group. Heider's balance theory applies strongly in cliques: if two friends embedded in the same cluster become openly hostile this can be a significant problem for the group (Cartwright & Harary 1956). The need for close monitoring of any disruption in the network of dyadic ties and the rippling effect of troubles in any one dyad or triad make the maintenance of cliques socially demanding, even beyond the direct costs of maintaining the dyadic bonds in which one is directly involved. The proliferation of dyads and triads that must be monitored and managed means that costs escalate as the clique gets larger. Small group size (4–5 members) helps keep these social cognitive costs reasonable and also makes it easier for the group to coordinate joint activities.

In the literature on prototypical human group sizes, the size of the smallest core of human personal networks—the support clique—typically ranges from 3 to 7, with the mean varying in part based on how data is gathered (McPherson & Smith-Lovin 1987; Marsden 1987; Stiller & Dunbar 2007). However, such ‘cliques’ are not necessarily cliques in the network or group sense, as an individual may have others in their support clique who have no relationship with one another. In the network sense, the support clique is an ego-centric star, and need not be a fully connected cluster. When two dyads are not tied together into a triad, the challenge of maintaining balance in all three of the dyads is less significant. Instead, stresses created by increased demands from one dyadic partner may reduce time available to attend to the other relationships. When the dyads are fully clustered, a crisis for one member can be handled cooperatively by the group and the support load flexibly distributed. This reduces the cumulative risk posed by multiple dyadic obligations and protects against the damage to a bond that might otherwise result when one cannot respond effectively to a sudden need by a partner.

(p.275) Coalitions: Exchange-Based Strategic Alignments

The ethological literature on coalition formation in primates (e.g. Harcourt & de Waal 1992) distinguishes between momentary coalitions that emerge for a single interaction and more enduring alliances among partners who routinely come to each other's aid or defence. Studies of coalitions in the social sciences (e.g. Komorita & Kravitz 1983; Mannix 1993; Murnighan 1985) typically do not make this distinction, and use the term ‘coalition’ to refer to strategic groupings regardless of stability.

Coalitions rely on exchange-based cohesion and, for those that endure over time, on dyadic obligations of reciprocal favours. Structurally, a coalition is based on the perception of shared advantages to grouping together, and to persist all members must remain convinced that the coalition benefits them more than some alternate political grouping. The emphasis on the individual self (and not the relationship or the collective) makes this a weak adhesive. In contrast to inwardly focused cliques, coalitions are focused on the external political context. Individuals solicit support from others to help counteract the power of some other individual or group. They need not like one another or feel depersonalized attraction based on shared category membership, for coalitions can bring together ‘strange bedfellows’ who otherwise have little in common.

Group theories based on social exchange (e.g. Moreland & Lévine 1982; Thibaut & Kelley 1959) fit the coalition model. They assume that members are regularly scanning the environment and assessing other potential groups that might offer a better deal. If they find one, they will leave or threaten to leave to get a better deal from the group. In an ecology of shifting coalitions (e.g. the US Congress or a parliament), any membership change has political consequences, especially when allies who abandon one coalition join another. In this case dyadic ties of obligation can exacerbate the threat to the group rather than keep it together. A departing member may actively recruit others to leave with them, especially if he or she has a few strong allies within the coalition. Maintaining a positive power position thus requires constant monitoring not only of one's own dyadic alliances within the coalition but also the strength of political alignment of all other members with one another. For advanced players, monitoring the ties of coalition members with others outside the current group is also wise, as the value of any particular coalition is always relative. The social cognitive challenges of this Machiavellian monitoring are very high.

(p.276) Despite these challenges, coalitions can be relatively large if the behavioural demands on members are modest (‘vote with me on this; let others know by your presence that you support me’). Members can minimize the demands of monitoring the political situation by keeping track of a small subset of partners within their shifting coalitions, and by delegating comprehensive monitoring to dedicated roles such as the ‘whips’ of US congressional politics who keep track of the commitments and intentions of all relevant players. A coalition of hundreds in a deliberative body, however, is a more fragile grouping than a coalition among just a few people who can more easily monitor one another.

Comrades: bonded Collective Action Groups

The defining characteristics of comrades are collective action and strong bonding to a group. The prototype of comrades is a war party, in which all members are bonded collectively (rather than dyadically) to one another in service of a joint goal. This bond with the group is essential to ensure loyalty and commitment in trying circumstances, as in military action the small group and not the individual soldier is the fundamental unit of effectiveness (Marshall 1947). Although powerful dyadic relationships can and do form among comrades, they are not essential as the primary attraction is depersonalized, the primary form of cohesion task cohesion, and the primary identity the collective self. A Marine feels a bond with another Marine and can immediately coordinate with him or her in joint action even if they are working together for the first time.

Comrades illustrate the difference between an acting group and a standing group or pool from which acting units are assembled or recruited. This is similar to the distinction between the interacting bands of a hunter-gatherer society and the larger communities of members from which the bands are composed (see Layton & O'Hara ch. 5 this volume). For comrades, the standing group can be very large (as with the Marines) even though the effective acting unit of soldiers for a particular mission is small. Among ancestral humans, the pool of comrades would probably be a sex—and age-stratified group of male adults eligible for missions such as hunting large dangerous game or conducting raids against other groups.

The norms that govern bands of comrades are strong and rigorously instilled through training and initiation. In contrast to coalitions, members do not come and go according to shifting political winds; instead, (p.277) membership in the pool is permanent and defection a serious violation. The psychological machinery of induction helps to ensure loyalty via a powerful collective identity and attachment to the group, and hence the need for constant monitoring of intentions is minimized.

The size of a pool of comrades can be scaled up through a stratified structure of nested groups of different size. In the military, a squad is nested within a platoon, platoons are collected into companies, which is turn make up battalions, and so on up to an army composed of tens of thousands. In day-to-day action an individual's strongest attachment is to their smaller unit, in which interpersonal bonds can help the unit cohere. The strength of socio-emotional connection should also be especially strong with the cohort with which one was initiated. However, the ultimate loyalty is to the larger-scale unit to which each cohort belongs.

In the organization of comrades with cells and larger units, a new layer of social is complexity is evident. Belonging is generalized so that smaller groups in turn ‘belong’ to larger units. In this way groups are themselves linked and subordinated in a hierarchical structure. While inter-group rivalries can still be a source of tension and prove problematic in joint actions, awareness of shared membership in a larger unit to which both units belong can temper inter-group conflicts. In the social psychology literature, this phenomenon is studied under the rubric of categorization and re-categorization (Gaertner et al. 2000).

Colleagues: Shared Interests and Social Distinctiveness

Colleagues, like comrades, are connected via a shared social identity and depersonalized attraction. However, groups of colleagues differ from comrades as collectives and in the form of social organization in two important ways. First, colleagues with specialized skills or training are often dispersed across many primary action groups in which they play a specialized role. Thus what colleagues share is not membership in fully nested groups but boundary-spanning ties based on common training and similar roles. For example, accountants share a common training and interests, and may get together regularly at seminars and talk together on listserves, but many will be the only accountant in their immediate work group or organization.

Clustering of specialized colleagues in primary work groups (like comrades) is a phenomenon that probably developed gradually as the (p.278) scale of human societies increased. Small bands of hunter-gatherers might each have a shaman, for example, and each shaman might have one or two apprentices. Gatherings of shamans, however, would occur only at seasonal gatherings of the larger tribe, when shamans could get together and share information and techniques, and perform ceremonies together. In towns and cities, however, people with the same specialized occupation can cluster together in guilds and live and work concentrated in the same district.

In the transition from smaller units to larger societies, the ‘invisible colleges’ of colleagues can provide connections across bands and also across tribes, as specialists can recognize one another's common distinctiveness even when there is a language barrier to overcome. Shared training and identity can also help to promote the development of norms and rules that transcend tribal boundaries. For example, European armies observed common rules of protocol regarding the surrender and ransoming of officers. During the battle of Waterloo British infantry who took temporary cover among freshly wounded Frenchmen engaged in what amounted to shop talk, exchanging professional opinions about how the battle might end (Keegan 1976, 204). The mutual recognition of colleagues made cordial exchanges possible even between official enemies.

Groups and Social Complexity

A full account of social complexity needs to incorporate the ways in which groups can bind people together and coordinate their actions on a large scale without overwhelming their social and cognitive capacities. Groups such as bands and colleagues rely on the cross-level ties of membership as a source of cohesion beyond dyadic relationships, which are costly to maintain and monitor. The generalization of attachments beyond dyads to larger units provides a more flexible array of building blocks for creating large complex societies. In such societies, all four of the group types described here (plus others not detailed) operate, serving different functions. The ability of people to simultaneously belong to cliques, coalitions, and groups of comrades and colleagues adds social complexity. At the same time, group norms and roles—including the special role of leadership—can simplify the social demands of everyday life. One of the essential insights in the study of complexity is that complex structures are not necessarily generated by highly complicated processes (p.279) (Holland 1995). A small number of building blocks can generate highly complex architecture. And a large number of agents, following simple rules that make their actions dependent on the actions of others, can generate patterns of action and structure that are staggeringly complex.