The target article takes a nuanced look at a fundamentally important problem of the genomics era: How the cultural environment and gene–culture coevolution make it difficult to understand the genetic underpinnings of human behaviour.

However, heritability is a murky concept for human traits, particularly behavioural traits. This is the case for many reasons, not least the complicating effects of culture. Here, we consider, first, whether a conclusion of this paper is not that we need to consider culture when we calculate heritability, but rather that attempting to calculate heritability for culturally complex human traits might add little to our understanding of human evolution and behaviour. And second, that heritability, influenced by culture, is conceptually and technically more complicated than the models in the appendix suggest.

In the modelling section (Appendix 7), genotype-by-environment (G × E) interactions are excluded for simplicity. However, as discussed at length elsewhere (Feldman & Lewontin, Reference Feldman and Lewontin1975; Feldman & Ramachandran, Reference Feldman and Ramachandran2018; Lewontin, Reference Lewontin1974), it is difficult or impossible to ignore G × E interactions for most complex human behavioural traits, and there is no simple way to understand the relative importance of genes and environment in the determination of phenotype when the two interact. Furthermore, as shown in foundational papers on cultural evolution (e.g., Cavalli-Sforza & Feldman, Reference Cavalli-Sforza and Feldman1973a, Reference Cavalli-Sforza and Feldman1973b), cultural transmission can distort the signatures of genetic heritability and deepen the impact of G × E interactions, further obscuring the relationship between heritability and family phenotypic resemblances. The importance of cultural inheritance and genotype-by-culture interactions, and the extent to which these affect or even invalidate inferences derived from calculations of heritability in the case of human behavioural traits, is difficult to overstate.

The example given in the modelling section – the trait of melanin production in response to UV exposure – illustrates these issues. The authors point out that this trait has genetic, environmental, and cultural components, but we must also consider that these components interact. Sunscreen absence or presence might have a different effect in different environments but also, for example, the cultural trait of sunscreen use might be most likely to spread to those who have a genetic lack of melanin and high-UV exposure, and populations lacking sunscreen might implement other cultural interventions, such as hats or clothing. This means that the expression of the trait of melanin production is jointly governed by genetics, cultural context, the physical environment, and the numerous complex interactions between those components. As the authors note, the heritability of such a trait is not fully captured by equation (3) without terms accounting for those interactions. Quantitative genetic models were often developed for situations with relatively controlled environments, such as artificial breeding programmes, and G × E interactions could often be safely ignored in these contexts (Falconer & Mackay, Reference Falconer and Mackay1996). For phenotypes influenced by culture, this is rarely the case, requiring justification beyond simplification of the model.

It may be helpful to ask: How can we use the measures of heritability derived in the target article? Heritability is a trait-specific, population-specific measure and is not, alone, globally informative. Narrow-sense heritability is often used, for example, to assess the response of a trait to selection in a given population in a given environment (e.g., Lande, Reference Lande1979). The models in the target article, then, aim to assess the effect of a relevant cultural influence on trait heritability and to improve our understanding of the evolution of these traits. We suggest that there are some technical details that require cautious consideration, implementation, and interpretation to approach this aim. First, the relationship between heritability and the response to selection (and other theoretical aspects of phenotypic evolution) depends on the assumption of a Gaussian phenotypic distribution (see e.g., Karlin, Reference Karlin, Eisen, Goodman, Namkoong and Weir1988; de Villemereuil, Schielzeth, Nakagawa, & Morrissey, Reference de Villemereuil, Schielzeth, Nakagawa and Morrissey2016). This assumption can be violated by this model because the cultural phenotypic contribution is modelled as a bounded uniform distribution, and the sum of the Gaussian genotypic and environmental contributions and the uniform cultural contribution need not be normal. Second, the use of phenotypic evolution models, for example, for retrospective selection studies, relies on the assumption that phenotypic and genotypic variances remain constant over time (Schluter, Reference Schluter1984; but see e.g., Turelli, Reference Turelli1988). Models involving dynamically changing phenotypic variances involve complex departures from standard theory (e.g., Gilpin & Feldman, Reference Gilpin and Feldman2019; Karlin, Reference Karlin, Eisen, Goodman, Namkoong and Weir1988) and the ways in which these variances are free to change in real systems are not straightforward (Arnold, Bürger, Hohenlohe, Ajie, & Jones, Reference Arnold, Bürger, Hohenlohe, Ajie and Jones2008). Adding phenotypic variance fluctuations over time in this model must, thus, also be justified by showing why the assumptions that warrant holding phenotypic variances constant in the absence of culture (e.g., logarithmic metric scales) do not apply to those changes caused by culture.

Finally, it is important to note that the heritable component of a trait with culturally evolving influences is not just genetic, it is also cultural. Therefore, the measure of interest for practical purposes should include the heritable cultural component in the numerator in equation (3), alongside the heritable genetic components. The ability of a trait to respond to selection is determined by how much variation in that trait is amenable to selection – in other words, as pointed out by Danchin and Wagner (Reference Danchin and Wagner2010), the important value to consider is not just the genetic variance but the heritable variance in total. Separating the environmental component of heritability into cultural and ecological components is an important step in the complicated process of modelling cultural influences on heritability. However, cultural traits can alter selection pressures on genetic traits (Feldman & Zhivotovsky, Reference Feldman and Zhivotovsky1992; Laland, Kumm, & Feldman, Reference Laland, Kumm and Feldman1995), influence mating patterns causing hidden population structure (Creanza & Feldman, Reference Creanza and Feldman2014), be transmitted beyond the family unit via oblique or horizontal transmission, and alter the parameters of their own evolution in multiple ways (e.g., Creanza, Fogarty, & Feldman, Reference Creanza, Fogarty and Feldman2012; Fogarty, Creanza, & Feldman, Reference Fogarty, Creanza and Feldman2013, Reference Fogarty, Creanza and Feldman2019) all of which could have profound effects on how we understand, calculate, and use heritability.