Richerson et al. present a wide-ranging synthesis of evidence for cultural group selection. An innovative feature of their argument is that they draw attention to the importance of quantifying the apportionment of cultural variation within and between groups. They calculate F _ST values for 29 aspects of culture between neighbouring groups and argue that relatively large cultural F _ST values provide scope for cultural group selection. In addition, they cite Bell et al.'s (2009) finding that average cultural F _ST values (as indexed by responses to the World Values Survey) are greater for neighbouring groups than average genetic F _ST values as evidence that there is greater potential for cultural group selection than genetic group selection. While we agree that this framework for quantifying cultural variation holds considerable potential, we challenge Richerson et al.'s interpretation of the existing empirical data.

The F _ST statistic belongs to a broad family of “fixation indices” – statistics developed by population geneticists to study genetic differentiation between populations (Holsinger & Weir Reference Holsinger and Weir2009). Several studies have demonstrated that human genetic variation is predominantly clinal, with differentiation between populations being strongly predicted by geographic distance across a variety of biological markers (Handley et al. Reference Handley, Manica, Goudet and Balloux2007). Discontinuities do exist (typically associated with geographic obstacles), but many apparent genetic barriers have proven to be artefacts of heterogeneous spatial sampling (Handley et al. Reference Handley, Manica, Goudet and Balloux2007). A clinal pattern of variation is consistent with an “isolation by distance” (IBD) model (Wright Reference Wright1943), in which individuals tend to migrate limited distances relative to the total geographic range of the species. Under IBD, a gradual blending of one population into the next is predicted, rather than group barriers. Nevertheless, two sampling locations can produce significant F _ST values simply due to their geographic separation.

None of the 29 cultural F _ST estimates reported by Richerson et al. or the cultural F _ST estimates reported by Bell et al. (Reference Bell, Richerson and McElreath2009) have been analysed within a spatially explicit framework. This renders these estimates difficult to interpret, since, as in the genetic case, apparent population structure could be an artefact of discontinuous spatial sampling, rather than cultural barriers.

Recently, we published a study that quantified the extent to which geography and group affiliation independently predict cultural differentiation between ethnolinguistic groups (Ross et al. Reference Ross, Greenhill and Atkinson2013). We used geographic coordinates and coded narrative elements of 700 versions of a folktale from 31 European groups, analysing both individual folktales and group-level differentiation using Φ _ST, a fixation index that is closely related to F _ST (Holsinger & Weir Reference Holsinger and Weir2009). We found significant differentiation between groups with an average Φ _ST of 0.091, indicating that, on average, 9.1% of the variation between individual folktales was between groups, which is considerably higher than the genetic differentiation found between comparable European populations (Lao et al. Reference Lao, Lu, Nothnagel, Junge, Freitag-Wolf, Caliebe, Balascakova, Bertranpetit, Bindoff, Comas, Holmlund, Kouvatsi, Macek, Mollet, Parson, Palo, Ploski, Sajantila, Tagliabracci, Gether, Werge, Rivadeneira, Hofman, Uitterlinden, Gieger, Wichmann, Rüther, Schreiber, Becker, Nümberg, Nelson, Krawczak and Kayser2008; Novembre et al. Reference Novembre, Johnson, Bryc, Kutalik, Boyko, Auton, Indap, King, Bergmann, Nelson, Stephens and Bustamante2008). However, incorporating geography into the analysis revealed that at the level of individual folktales, geographic distance explains considerably more variation between folktales than group boundaries (6.6% of variance vs. 3.7%). Such a pattern of predominantly clinal variation is consistent with IBD-like cultural diffusion processes. This means that geographically close individuals/tales from neighbouring groups tend to be more similar than is suggested by the relatively large average cultural Φ _ST, thereby limiting the scope for cultural group selection.

Two recent studies speak to the generality of our findings and the F _ST estimates from Richerson et al. First, Brown et al. (Reference Brown, Savage, Ko, Stoneking, Ko, Loo and Trejaut2014) examined group-level folk song differentiation among nine indigenous ethnolinguistic groups in Taiwan. They found significant cultural differentiation between groups. Intriguingly, however, cultural Φ _ST was found to be an order of magnitude smaller than genetic Φ _ST for the same groups, contra Bell et al.'s (2009) hypothesis that human groups generally show a higher degree of cultural differentiation than genetic differentiation. This result is particularly pertinent to discussions about human evolution, because the folk song data are drawn from small-scale indigenous societies whose lifestyles and group structure better approximate those of our ancestors than the large-scale multiethnic nation states studied by Bell et al. (Reference Bell, Richerson and McElreath2009). Second, Shennan et al. (Reference Shennan, Crema and Kerig2015) examined individual level variation and group-level differentiation in two material culture complexes – pottery and personal ornaments – from 361 sites of 22 putative Neolithic cultural groups in Europe. At the level of individual artefacts, cultural affiliation was an independent predictor of pottery variation, while geography was not; but both cultural affiliation and geography were independent predictors of ornament variation. At the group level, they found significant cultural differentiation, with geography predicting differentiation in ornaments but not pottery.

This result suggests that the relative influence of cultural barriers and geographic effects can vary across different cultural markers in comparable populations, just as is the case for different genetic markers (i.e., autosomal DNA, mitochondrial DNA, and Y-chromosome; Jobling Reference Jobling2012). Caution is therefore warranted in making generalizations about the relationships between groups on the basis of analysis of a limited range of cultural traits.

While we support the rigorous analysis of empirical data to quantify cultural variation, the evidence we present here suggests that cultural differentiation between groups varies considerably across cultural domains and spatial scales and is often best explained in terms of geographic clines – a pattern that suggests IBD-like cultural processes and limited scope for cultural group selection between neighbours. We note that, in the absence of stable individual level trait differences between neighbouring groups, the most important forms of variation for cultural group selection may be group-level traits (Smaldino Reference Smaldino2014), such as the presence or absence of particular religious or political institutions, rather than the values, stories, songs, or material possessions of individuals. Further work that examines individual variation and group-level differentiation using a geographically explicit framework across the full spectrum of aspects of human culture is needed.