In answer to the question “What is a group?” Smaldino adopts Wilson's (Reference Wilson1975) concept of a trait-group, defining groups according to how the interactions among individuals impact their fitness. This leads to groups that “may consist of a simple dyad or a population of thousands” (sect. 2.1). In addition to concerns raised by others (Okasha Reference Okasha2006; Sterelny Reference Sterelny1996), we see two problems with using the trait-group concept. First, the generality of this definition – encompassing almost any kind of social interaction across vastly different scales – provides little guidance for the task of understanding how the many different kinds of groups found in human societies emerge. Second, because cultural group selection is defined as competition among groups that are defined in cultural terms, it requires us to define the groups in question in cultural terms, not biological ones. Scholars in the social sciences have devoted considerable effort to documenting and analyzing the diverse groups that individuals form to cooperate, coordinate, or act collectively, but links between this work and cultural group selection remain tenuous (Cronk & Leech Reference Cronk and Leech2013). Drawing on insights from the social sciences, we propose two conceptual distinctions that complement Smaldino's ideas about emergent group-level traits.

Our first suggestion is a conceptual distinction between objective and subjective groups. Objective groups are those whose boundaries are shared by all members of the population, meaning the boundaries between members and non-members are the same from one individual to the next. Objective groups correspond most closely to the common understanding of the word “group” and reflect the definitions used by both advocates and critics of early theories of genetic group selection (Maynard Smith Reference Maynard Smith1998). Subjective groups are those whose boundaries differ depending on one's position within the population, meaning that the boundaries between members and non-members will be different from one person to the next. For example, human kinship systems often feature both descent groups – with objective boundaries between lineage members and non-members – and kindreds – with subjective, ego-centric boundaries dividing relatives and non-relatives (Cronk & Gerkey Reference Cronk, Gerkey, Dunbar and Barrett2007). Although the distinction between objective and subjective groups in this example is dichotomous, in other cases it may be more of a continuum, with some groups displaying more or less objective boundaries than others.

An advantage of this approach to defining groups is that it can be used to establish clearer links between theories of multilevel selection and empirical research on individual/group dynamics in the social sciences. In particular, the growing interdisciplinary research program on quantitative social network analysis provides a wide range of conceptual and statistical tools for defining groups along the objective/subjective continuum, including clustering, centrality, and cliques, among others (Jackson Reference Jackson2008; Wasserman & Faust Reference Wasserman and Faust1994). Although some evolutionary scholars have begun to use these tools from network analysis (e.g., Apicella et al. Reference Apicella, Marlowe, Fowler and Christakis2012; Nolin Reference Nolin2011; Pacheco et al. Reference Pacheco, Traulsen, Ohtsuki and Nowak2008), we believe this approach deserves greater attention. This suggestion applies particularly to those who use some version of the trait-group definition, because these “groups” are better understood as networks emerging from and in turn shaping the interactions of individuals. For example, there is consensus among evolutionary scholars that mechanisms of positive assortment are necessary for the evolution of cooperative behavior among individuals, but are different mechanisms (kinship, reciprocal contingency, reputation, adherence to cultural norms) more or less likely to lead to the emergence of objective or subjective group structures?

Our second suggestion is a conceptual distinction between corporate groups and categorical groups. As the name implies, corporate groups are functionally integrated and have clear, body-like organizational structures that coordinate the actions of their members. Categorical groups are composed of individuals who share a bundle of characteristics or traits that distinguish members from non-members, but these boundaries do not rely on the kind of organizational structures found in corporate groups. Our distinction between corporate and categorical groups shares some aspects of Smaldino's distinction between aggregate and emergent groups, particularly the emphasis on diversification and organization in the latter. However, our terminology has two advantages. First, it establishes closer connections to the large body of social science research on competition among corporate groups – including firms (e.g., Arthur Reference Arthur2012), political interest groups (e.g., Baumgartner et al. Reference Baumgartner, Berry, Hojnacki, Kimball and Leech2009), organized religions (e.g., Stark Reference Stark1996), and descent groups (Keesing Reference Keesing1975) – and categorical groups – including those shaped by shared ethnicity (Barth Reference Barth1969), spirituality (e.g., Fuller Reference Fuller2001), and nationality (Anderson Reference Anderson1991). Second, it builds on theories of cultural group selection to develop testable hypotheses about the relationship between group-level traits and the fitness consequences of interactions among individuals. For example, because categorical groups are not organized in any way, cultural group selection acting on such groups should result in traits that help the group's individual members achieve their personal goals rather than any group-level goals. One might refer to this as “soft” cultural group selection. In contrast, cultural group selection acting on corporate groups should result in traits that influence the group's ability to achieve its collective goals, even when this runs counter to its members' efforts to achieve their personal goals. Consider, for example, the fact that institutions frequently impose rules against nepotism to promote the interests of the corporate group over those of its individual members. One might refer to this as “hard” cultural group selection.

These suggestions for conceptual distinctions among different kinds of groups are not meant to be exhaustive. Rather, we believe they reflect important insights gained through the work of generations of social scientists who have explored the implications of Durkheim's famous argument – echoed in Smaldino's article: “society is not the mere sum of individuals, but the system formed by their association represents a specific reality which has its own characteristics” (Reference Durkheim, Solovay, Mueller and Catlin1895/1964, p. 129).