The need for an integrative longitudinal framework
Although Uchiyama et al. insightfully draw attention to this, it is important to note the literature they discuss provides little in the way of formally integrating long-term societal change with shorter-term developmental change and genetic modeling. The authors discuss cross-sectional comparisons between culturally distinct geographic areas, longitudinal accounts of developmental change, longitudinal accounts of long-term societal changes in behavior (the Flynn effect), and long-term accounts of changes in heritability (fertility and educational attainment), but without an overarching modeling framework that integrates these effects simultaneously.
For example, in discussing UV exposure, although the authors note that “cultural change can be particularly fast and potent” (sect. 2.1, para. 3), and “the environment of the genome is…a moving reference frame that rapidly evolves in relation to both genes and ecology” (sect. 2.1, para. 3), their empirical discussion is focused on geographical differences in vitamin D in relation to UV, and not changes per se tied to cultural evolution. Other cited research on differences in heritability as a function of cultural variables (e.g., Engzell & Tropf, Reference Engzell and Tropf2019) is similarly cross-sectional. The limitations of making dynamic inferences based on cross-sectional data are well known, and have been raised in the context of the studies discussed (Morris, Reference Morris2020). In short, although these studies work in promising directions, existing literature and the discussion provided by Uchiyama et al. lack a unified modeling treatment, one that integrates two very different timescales of human experience.
Challenges in the modeling of culture, genes, and environment, and potential directions
We see direct parallels between the phenomena discussed by Uchiyama et al. and age-period-cohort (APC) models in the epidemiological literature (Fosse & Winship, Reference Fosse and Winship2019). APC models decompose variance into portions associated with age (development), as well as temporal period (the shared experiences of those living at a particular moment in time, regardless of age or cohort) or cohort (the shared experiences of those developing together during a particular period). Although APC models traditionally do not concern themselves with genetics, they do jointly account for developmental and contextual effects across different timescales.
Extending APC frameworks to include genetic and environmental effects would be relatively straightforward. Multicohort twin and family studies have been conducted; molecular genetic information could also be included in APC designs, such as through polygenic scores or specific alleles (as in Mendelian randomization paradigms). Doing so would provide a formal paradigm for joint modeling of developmental and cultural processes, guiding attention to important theoretical and methodological issues in the study of culture and genetics.
Research on APC models highlights the intertwined nature of developmental and cultural effects and potential challenges in modeling them jointly. Work in this area has demonstrated how effects that appear conceptually distinct can be difficult to distinguish when specified in models (Fosse & Winship, Reference Fosse and Winship2019). Changes in genetic effects modeled in terms of cohorts (e.g., Briley, Harden, & Tucker-Drob, Reference Briley, Harden and Tucker-Drob2015; Rosenquist et al., Reference Rosenquist, Lehrer, O'Malley, Zaslavsky, Smoller and Christakis2015; Sanz-de-Galdeano, Terskaya, & Upegui, Reference Sanz-de-Galdeano, Terskaya and Upegui2020), for example, might equivalently be framed in terms of periods (e.g., war, socioeconomic conditions, or policy eras) which are often implicitly the focus of explanation anyway. Other research suggests that apparently simple sociological concepts might require relatively complex model features to capture when considered simultaneously against the backdrop of development (Fosse & Winship, Reference Fosse and Winship2019).
Many authors have noted that age and period are theoretical proxies for other, unspecified factors of interest, such as specific cultural or environmental agents impacting everyone living at a particular point in time, or neurodevelopmental processes that unfold at specific points in the lifespan. Focusing on these specified causal factors, rather than making assumptions about unspecified factors in the form of age, period, or other (e.g., geographic) proxies, provides numerous theoretical and modeling benefits (Fosse & Winship, Reference Fosse and Winship2019). However, doing so also highlights the importance of how cultural and environmental variables are defined and measured. How are cultural variables different from other measured environmental variables in biometric frameworks, if at all? Is it the scope of an exposure in time or space (the Great Recession, for instance, vs. family cohesiveness)? Transmission across generations? How do you operationalize and validate such measures in a way that provides rigorous tests of a theory?
Although it is true that cultural evolution “can be particularly fast and potent” when compared to genetic evolution, it can also be relatively slow compared to the timeframe typically employed in isolated behavioral genetic studies. Even the effects of relatively discrete, “potent” once-in-a-generation events (e.g., the onset of world war) might require data from multiple generations to discern, and are difficult to capture via serendipity in study design.
In general, these issues, taken together with issues highlighted by Uchiyama et al., point to the need for replication of behavior genetic findings across long time spans, and to do so in a formal, comprehensive longitudinal modeling framework. Doing so allows systematic examination of how parameters of population and molecular behavior genetic models might or might not be invariant across different periods, and to examine potential measured causal or mechanistic variables that might be affecting the non-invariance of those model components. Such designs will be critical in moving past cross-sectional studies to more causally informative, comprehensive accounts of the dynamic interplay between genes, development, and culture as envisioned by Uchiyama et al.
You have
Access
Uchiyama et al. compellingly demonstrate the importance of cultural context in interpreting estimates of genetic effect. As they show, genetic pathways depend on cultural processes that evolve and change, providing a broad, pervasive system of environmental influences transcending individuals and their families.
The need for an integrative longitudinal framework
Although Uchiyama et al. insightfully draw attention to this, it is important to note the literature they discuss provides little in the way of formally integrating long-term societal change with shorter-term developmental change and genetic modeling. The authors discuss cross-sectional comparisons between culturally distinct geographic areas, longitudinal accounts of developmental change, longitudinal accounts of long-term societal changes in behavior (the Flynn effect), and long-term accounts of changes in heritability (fertility and educational attainment), but without an overarching modeling framework that integrates these effects simultaneously.
For example, in discussing UV exposure, although the authors note that “cultural change can be particularly fast and potent” (sect. 2.1, para. 3), and “the environment of the genome is…a moving reference frame that rapidly evolves in relation to both genes and ecology” (sect. 2.1, para. 3), their empirical discussion is focused on geographical differences in vitamin D in relation to UV, and not changes per se tied to cultural evolution. Other cited research on differences in heritability as a function of cultural variables (e.g., Engzell & Tropf, Reference Engzell and Tropf2019) is similarly cross-sectional. The limitations of making dynamic inferences based on cross-sectional data are well known, and have been raised in the context of the studies discussed (Morris, Reference Morris2020). In short, although these studies work in promising directions, existing literature and the discussion provided by Uchiyama et al. lack a unified modeling treatment, one that integrates two very different timescales of human experience.
Challenges in the modeling of culture, genes, and environment, and potential directions
We see direct parallels between the phenomena discussed by Uchiyama et al. and age-period-cohort (APC) models in the epidemiological literature (Fosse & Winship, Reference Fosse and Winship2019). APC models decompose variance into portions associated with age (development), as well as temporal period (the shared experiences of those living at a particular moment in time, regardless of age or cohort) or cohort (the shared experiences of those developing together during a particular period). Although APC models traditionally do not concern themselves with genetics, they do jointly account for developmental and contextual effects across different timescales.
Extending APC frameworks to include genetic and environmental effects would be relatively straightforward. Multicohort twin and family studies have been conducted; molecular genetic information could also be included in APC designs, such as through polygenic scores or specific alleles (as in Mendelian randomization paradigms). Doing so would provide a formal paradigm for joint modeling of developmental and cultural processes, guiding attention to important theoretical and methodological issues in the study of culture and genetics.
Research on APC models highlights the intertwined nature of developmental and cultural effects and potential challenges in modeling them jointly. Work in this area has demonstrated how effects that appear conceptually distinct can be difficult to distinguish when specified in models (Fosse & Winship, Reference Fosse and Winship2019). Changes in genetic effects modeled in terms of cohorts (e.g., Briley, Harden, & Tucker-Drob, Reference Briley, Harden and Tucker-Drob2015; Rosenquist et al., Reference Rosenquist, Lehrer, O'Malley, Zaslavsky, Smoller and Christakis2015; Sanz-de-Galdeano, Terskaya, & Upegui, Reference Sanz-de-Galdeano, Terskaya and Upegui2020), for example, might equivalently be framed in terms of periods (e.g., war, socioeconomic conditions, or policy eras) which are often implicitly the focus of explanation anyway. Other research suggests that apparently simple sociological concepts might require relatively complex model features to capture when considered simultaneously against the backdrop of development (Fosse & Winship, Reference Fosse and Winship2019).
Many authors have noted that age and period are theoretical proxies for other, unspecified factors of interest, such as specific cultural or environmental agents impacting everyone living at a particular point in time, or neurodevelopmental processes that unfold at specific points in the lifespan. Focusing on these specified causal factors, rather than making assumptions about unspecified factors in the form of age, period, or other (e.g., geographic) proxies, provides numerous theoretical and modeling benefits (Fosse & Winship, Reference Fosse and Winship2019). However, doing so also highlights the importance of how cultural and environmental variables are defined and measured. How are cultural variables different from other measured environmental variables in biometric frameworks, if at all? Is it the scope of an exposure in time or space (the Great Recession, for instance, vs. family cohesiveness)? Transmission across generations? How do you operationalize and validate such measures in a way that provides rigorous tests of a theory?
Although it is true that cultural evolution “can be particularly fast and potent” when compared to genetic evolution, it can also be relatively slow compared to the timeframe typically employed in isolated behavioral genetic studies. Even the effects of relatively discrete, “potent” once-in-a-generation events (e.g., the onset of world war) might require data from multiple generations to discern, and are difficult to capture via serendipity in study design.
In general, these issues, taken together with issues highlighted by Uchiyama et al., point to the need for replication of behavior genetic findings across long time spans, and to do so in a formal, comprehensive longitudinal modeling framework. Doing so allows systematic examination of how parameters of population and molecular behavior genetic models might or might not be invariant across different periods, and to examine potential measured causal or mechanistic variables that might be affecting the non-invariance of those model components. Such designs will be critical in moving past cross-sectional studies to more causally informative, comprehensive accounts of the dynamic interplay between genes, development, and culture as envisioned by Uchiyama et al.