We believe cultural group selection (CGS) is an elegant theoretical framework for studying the evolution of complex human behaviour. The target article sets out to present evidence that culturally evolutionary processes have resulted in the selection of large-scale cooperation as a group-level adaptation. However, the empirical evidence presented does, in our opinion, still fail to conclusively demonstrate that CGS explains human cooperation better than the competing hypotheses.

Variation in cooperative behaviour – frequently measured through economic games – across different cultures (Henrich et al. Reference Henrich, Boyd, Bowles, Camerer, Fehr, Gintis, McElreath, Alvard, Barr, Ensminger, Henrich, Hill, Gil-White, Gurven, Marlowe, Patton and Tracer2005; Reference Henrich, McElreath, Barr, Ensminger, Barrett, Bolyanatz, Cardenas, Gurven, Gwako, Henrich, Lesorogol, Marlowe, Tracer and Ziker2006; Reference Henrich, Ensminger, McElreath, Barr, Barrett, Bolyanatz, Cardenas, Gurven, Gwako, Henrich, Lesorogol, Marlowe, Tracer and Ziker2010a) is put forward as evidence that cooperation is a stable, persistent, and group-level trait on which selection at the level of the group could act. However, as we and others have argued elsewhere, high levels of variation in cooperative behaviour are also found within ethnic groups (Lamba & Mace Reference Lamba and Mace2011; Reference Lamba and Mace2012; Nettle et al. Reference Nettle, Colléony and Cockerill2011), suggesting that variation attributed to different cultural norms could in fact reflect individual adaptations to different environments. The F _st data presented in the article supports the idea that countries differ in some beliefs and values, so at least one of the assumptions of CGS is met. However the beliefs and values examined are not associated with actual cooperative behaviours; for example, demonstrating that there is significant country-level variation in attitudes toward abortion has arguably little relevance to the argument at stake. And has there been enough time for countries to be subject to selection on the basis of this trait?

Cultural difference is one of the foundational assumptions of anthropology, so we should expect to find it. We agree that institutions are likely to be of particular interest in this field and the most promising candidates for cultural traits may be those classic subjects of anthropological investigation such as kinship norms, political systems, and religion, which are generally group-level traits, almost by definition. Religion often has the characteristics of a trait designed for group promotion and cohesion (e.g., costly rituals, apostasy punishment, and clear between-group demarcation) making it a strong candidate for spread through intergroup competition. Political systems have to be properties of the group rather than the individual, so group selection should be in operation in these cases (Currie & Mace Reference Currie and Mace2009; Currie et al Reference Currie, Greenhill, Gray, Hasegawa and Mace2010). Normative marriage and dispersal strategies may only be effective if the rest of your local population shares your understanding of the rules; strong frequency-dependent selection could maintain variation between groups, so such traits would also be promising cases to seek evidence for CGS. While language is a classic frequency-dependent group trait, the exact design of particular languages is not generally considered to confer any particular benefit over any other, therefore it is unlikely that CGS is operating on languages.

A key aspect of CGS is the idea that traits that provide an advantage to groups, even if at a cost to the individual, are selected for in a situation of intergroup competition. Our attempts at testing this hypothesis for the evolution of parochial altruism – the combination of in-group altruism and out-group hostility due to intergroup competition and conflict – using naturalistic field experiments in Northern Ireland, found little evidence to support this idea. Individual and neighbourhood exposure to conflict explained a reduction in out-group cooperation, but crucially had either no effect or also reduced cooperation toward the in-group. In addition, no significant differences in cooperation were found between the two cultural groups when taking into account individual socio-economic characteristics, which in turn best predicted the variation in cooperative behaviour (Silva & Mace Reference Silva and Mace2014; Reference Silva and Mace2015).

Richerson et al.'s article aims to present the evidence in favour of CGS, hence its adversarial approach rather than the more scientific approach of falsification. However there is still a large empirical project to be done in order to identify the domains in which CGS may be an important evolutionary force. Context is likely to be fundamental, as it is in any evolutionary scenario. Different traits will have different patterns of transmission and different selective forces acting upon them, and benefits may be operationalized at the level of the individual or of the society, or both. We are at a stage of investigation of this topic where we find one specific assumption may be met in one case and another assumption may be met for a different trait or in a different kind of a group; but a consistent story showing the wide applicability of CGS to the evolution of large-scale cooperative behaviour is still lacking, other than at an anecdotal level.