Culture rather than genes provides greater scopefor the evolution of large-scale human prosocialityAdrian V. Bella,1, Peter J. Richersonb, and Richard McElreathcaGraduate Group in Ecology,bDepartment of Environmental Science and Policy, andcDepartment of Anthropology, University of California, Davis, CA 95616Edited by Richard E. Nisbett, University of Michigan, Ann Arbor, MI, and approved August 31, 2009 (received for review March 25, 2009)Whether competition among large groups played an importantrole in human social evolution is dependent on how variation,whether cultural or genetic, is maintained between groups. Com-parisons between genetic and cultural differentiation betweenneighboring groups show how natural selection on large groups ismore plausible on cultural rather than genetic variation.altruism 兩 cultural FST 兩 group selection 兩 prosocialityHuman societies are unusual among vertebrates. While peo-ple in small-scale societies exhibit much more c ooperationand division of labor than other primates, people in even verylarge societies also show strong tendencies toward altruism.Warfare, food sharing, and taxation are all examples of prosocialpatterns of behavior that are c ommon in human societies butnearly completely absent in other vertebrates. Even when plau-sible analogues can be found in other vertebrates, the scale ofc ostliness of human altruism is extraordinary (1).Ex plaining the levels of human altruism observed ethnograph-ically and experimentally has proven to be difficult. Much of thisaltr uism is directed at strangers, and so is difficult to explain assimple reciprocity, or it benefits entire tribes or nations of onlydist ant genealogical kin, and so is difficult to explain as altruismamong individuals sharing recent common ancestry. Anotherscenario many researchers, since at least Darwin (2), are con-cerned with is competition among residential human groups thatare too large to comprise close genealogical kin (2–6). If g roupsdif fer in the frequency of individuals who are willing to sacrificetheir own labor, time, or safet y in ways that promote thec ompetitive ability of the residential group, then over timegroups with higher frequencies of such ‘‘altruists’’ may tend toreplace groups w ith fewer (7–9).In this paper, we refer to this scenario as ‘‘group-level selec-tion,’’ the evolution of behavior that reduces indiv idual fitnessbut increases the average fitness within large groups of onlydist antly related individuals. By ‘‘distantly related,’’ we meanthat most indiv iduals within the residential group do not sharevery recent common ancestors, and so common descent alonedoes not maintain much genetic variation among residentialgroups. Nevertheless, given the right population structure andlow rates of mixing among groups, individuals within groups maybe much more genetically similar to one another than they areto members of other groups, and therefore they may be closely‘‘related’’ in one important sense of the term (10). If geneticvariation among groups is sufficiently large, evolutionary theorypredicts that self-sacrifice on behalf of large residential groupscan evolve under the same processes that evolve self-sacrifice onbehalf of close kin. This is because all hypotheses about theevolution of altruistic behavior—behavior that reduces theabsolute fitness of the actor but increases the absolute fitness ofrecipients—hinge on processes that change and maintain vari-ation among social groups (11–14).Selection for altruism in such large groups, however, remainsa controversial topic in part because it is not clear that enoughbet ween-group variation existed in human societies to make it anappreciable evolutionary force (15). In very large residentialgroups, migration can quickly erode between-group geneticvariation. Nevertheless, recent work argues that sufficient vari-ation did exist by invoking reproductive leveling (7) [see also (16)].Reproductive leveling reduces the amount of bet ween groupvariation needed for selection to favor group-beneficial butindividually-costly traits. While it is not known how the estimatesof genetic differentiation for small forager g roups reported in (7)relate to Pleistocene foraging groups (see SI), it is intriguing tonote that reproductive leveling itself already has strong hints ofprosocialit y, begging the question of how it could evolve beforealtr uism (17). This illustrates that for genetic selection to favoraltr uism in large residential g roups, theorists need to invokeparticularly strong assumptions.A n alternative scenario is that human propensities to coop-erate arose through selection on cultural rather than geneticvariation (15, 18). Humans developed the capacity for complexculture perhaps beginning 250,000 years ago (19). Since thattime, culturally transmitted traits have come, along side of genes,to have a large influence on human behavior. Ever since theadvanced human capacity for social learn ing began, groups ofindividuals likely began rapid divergence in behavior due tocumulative cultural changes. This behavioral variation betweengroups can persist, given the right kinds of cultural evolutionaryforces (20). Even among our closest living relatives, chimpan-zees, plausibly socially-learned traits show some between groupvariation (21).Selection for culturally-prescribed altruists occurs through thesame process as for genes: groups of altruists leave moredaughter societies (8, 9). However, one advantage that culturalvariation has over genetic is that it does not require violentinter-group competition, nor group extinctions (22, 23). If failedgroups were incorporated routinely into successful ones, c on-for mist transmission and other forms of resocialization of failedgroups can lead to effective cultural selection on g roups eventhough such a pattern will generate rates of migration that keepgenetic FSTvery low between neighbors. Thus selection onculture can be powerful precisely when genetic selection at thegroup level is weakest.What is the scope for group-level selection on cultural vari-ation and how does this compare to the equivalent for genes? Anumber of mechanisms may permit cultural variation to be largerthan genetic variation between groups (15, 20). If these mech-an isms are important, the scope for group-level selection onculture will be much greater than for genes. Here we computeestimates of cultural variation among human g roups and com-pare these