If a socially valued trait was found to be “hardwired” and immutable, this would imply that all individual differences on it are because of the inherent superiority of some over others, and that all efforts toward improvement or equality would be futile. We applaud Madole & Harden's attempt to correct the widespread misunderstanding that several findings of such genetic determinism have already been made, and their clarification that even if genes have causal effects on phenotypes, these effects are typically probabilistic and context dependent with multiple mediating processes. Behavioral genetics is presented as a boon, rather than impediment, to social reform and progression, as it can help identify these mediating processes between genotypes and phenotypes so that we may intervene for the greater common good.
Relevant to this, here we direct attention to the number nine finding by Plomin, DeFries, Knopik, and Neiderhiser (Reference Plomin, DeFries, Knopik and Neiderhiser2016), namely that most phenotypes are not influenced much by the so-called shared environment. This finding is quite helpful when searching for targets of interventions, but also a bit discouraging: It tells us where not to look. Specifically, it tells us not to look in most of the places people are currently looking. The shared environment represents all things except genes that cause pairs of relatives to become more similar to each other. It thus encompasses not just variation in the family environments where children grow up (parenting, nutrition, etc.), but also all variation between families in broader geographical, economic, and cultural settings. If no variance is explained by the shared environment, this either means that it makes no difference to the relevant phenotype, or that these environmental effects are so unpredictable that they are uncorrelated even between close relatives. Either way, they are bad targets for interventions.
Intelligence, mental health, and personality are among the many important phenotypes largely unaffected by the shared environment, and so are sociopolitical traits such as social dominance orientation (SDO; Kleppestø et al., Reference Kleppestø, Czajkowski, Vassend, Røysamb, Eftedal, Sheehy-Skeffington and Thomsen2019) and justice sensitivity (Eftedal et al., Reference Eftedal, Kleppestø, Czajkowski, Sheehy-Skeffington, Røysamb, Vassend and Thomsen2022). Notable exceptions include educational attainment, and political conservatism (Willoughby et al., Reference Willoughby, Giannelis, Ludeke, Klemmensen, Nørgaard, Iacono and McGue2021) and authoritarianism (Eftedal et al., Reference Eftedal, Kleppesto, Czajkowski, Eilertsen, Sheehy-Skeffington, Vassend and Thomsen2020), which often have shared-environmental variance components of at least 20%.
A lack of shared-environmental influences on a trait does not mean that it is necessarily “hardwired” or immutable (e.g., mean IQ has been rising in recent decades). But it does imply that an environmental intervention must be well outside of the normal range of variation to work. For example, if you hope to increase intelligence by encouraging children to read and learn, you would need to provide more than what some parents already give: If it is true that the large amount of variation that exists on parenting, or social expectations, or number of books on the shelf, explains little or no variation in a trait, why, then, should such things suddenly make a difference if we intervene on them? Interventions that only represent nudges on factors that already vary substantially across contexts within a relevant sample are unpromising, regardless of which part of the causal chain from genotype to phenotype they purport to address.
A critical caveat to the conclusion that the shared environment has little influence, however, is the fact that this general pattern of findings primarily stems from samples confined to single cultures, typically white, educated, industrialized, rich, and democratic (WEIRD) ones, at one single point in time. That is, behavioral genetic studies can only conclude that there is no effect of the shared environment within their samples, but these samples mostly fail to measure the substantial variation in environments that we see across human cultures and ecologies in both space and time. For example, while attitudes toward group hegemony, as reflected in SDO, are unaffected by shared-environmental variation within a sample of Norwegian middle-aged twins (Kleppestø et al., Reference Kleppestø, Czajkowski, Vassend, Røysamb, Eftedal, Sheehy-Skeffington and Thomsen2019), there is nevertheless substantial variation across nations and states of the United States in the SDO levels of members of dominant societal groups, which correlate with ecological levels of macrostructural inequality (Kunst, Fischer, Sidanius, & Thomsen, Reference Kunst, Fischer, Sidanius and Thomsen2017). And when looking across time within cultures or countries, the increases we have seen in IQ over recent decades, for instance, are far larger than what genetic shifts alone can plausibly account for (Flynn, Reference Flynn2009). Ever more large-scale social psychological studies continue to pinpoint the ecological correlates of psychological phenomena such as these.
A lesson from behavioral genetics to those looking for effective societal interventions on shared-environmental variables is then to broaden one's horizon, and look at the things that vary across cultural ecologies, rather than the things that vary within them. The shared-environmental differences that actually seem to make a difference are those we see between, for example, Boston and Mumbai, or between 2022 and 1942, rather than those we see between the Smiths from west of town and the Harpers in the east. Behavioral genetics should broaden its empirical scope beyond single-culture WEIRD samples to adequately identify critical effects of the shared environment and thus potential societal targets of environmental intervention.
When such data collection is not feasible, investigating the unique environment of individuals, containing everything that makes relatives differ, also offers lessons for behavioral science. While this unique variance component appears to include mostly idiosyncratic and unsystematic influences (Turkheimer, Reference Turkheimer2000), there are still important things to learn from looking at how non-shared-environmental versus genetic variance components from different phenotypes are correlated. And as Franić et al. (Reference Franić, Dolan, Borsboom, Hudziak, van Beijsterveldt and Boomsma2013) describe, such correlations provide more stringent tests of latent psychological constructs than do regular factor analyses. These techniques revealed, for example, that sensitivities to being the perpetrator, victim, beneficiary, and observer of injustice are undergirded by separate and heritable latent motivations to be morally principled and opportunistic (so that the worse one reacts to injustice to oneself, the less one reacts to injustice to others). Furthermore, the unique environments that increase moral opportunism tend to also increase SDO, while genetic substrates that increase moral opportunism also increase SDO and lower altruism and generalized trust (Eftedal et al., Reference Eftedal, Kleppestø, Czajkowski, Sheehy-Skeffington, Røysamb, Vassend and Thomsen2022).
As non-shared-environmental correlations control for genetic and shared-environmental confounders, they can also help narrow our search for pairs of variables that are connected causally. For instance, individual educational attainment is associated with lower authoritarianism even when controlling for genetic and shared-environmental confounding (Eftedal et al., Reference Eftedal, Kleppesto, Czajkowski, Eilertsen, Sheehy-Skeffington, Vassend and Thomsen2020). Surely such knowledge about the underlying nature of latent psychological constructs and their potential causal connections are useful stepping stones in the search for good interventions.
You have
Access
If a socially valued trait was found to be “hardwired” and immutable, this would imply that all individual differences on it are because of the inherent superiority of some over others, and that all efforts toward improvement or equality would be futile. We applaud Madole & Harden's attempt to correct the widespread misunderstanding that several findings of such genetic determinism have already been made, and their clarification that even if genes have causal effects on phenotypes, these effects are typically probabilistic and context dependent with multiple mediating processes. Behavioral genetics is presented as a boon, rather than impediment, to social reform and progression, as it can help identify these mediating processes between genotypes and phenotypes so that we may intervene for the greater common good.
Relevant to this, here we direct attention to the number nine finding by Plomin, DeFries, Knopik, and Neiderhiser (Reference Plomin, DeFries, Knopik and Neiderhiser2016), namely that most phenotypes are not influenced much by the so-called shared environment. This finding is quite helpful when searching for targets of interventions, but also a bit discouraging: It tells us where not to look. Specifically, it tells us not to look in most of the places people are currently looking. The shared environment represents all things except genes that cause pairs of relatives to become more similar to each other. It thus encompasses not just variation in the family environments where children grow up (parenting, nutrition, etc.), but also all variation between families in broader geographical, economic, and cultural settings. If no variance is explained by the shared environment, this either means that it makes no difference to the relevant phenotype, or that these environmental effects are so unpredictable that they are uncorrelated even between close relatives. Either way, they are bad targets for interventions.
Intelligence, mental health, and personality are among the many important phenotypes largely unaffected by the shared environment, and so are sociopolitical traits such as social dominance orientation (SDO; Kleppestø et al., Reference Kleppestø, Czajkowski, Vassend, Røysamb, Eftedal, Sheehy-Skeffington and Thomsen2019) and justice sensitivity (Eftedal et al., Reference Eftedal, Kleppestø, Czajkowski, Sheehy-Skeffington, Røysamb, Vassend and Thomsen2022). Notable exceptions include educational attainment, and political conservatism (Willoughby et al., Reference Willoughby, Giannelis, Ludeke, Klemmensen, Nørgaard, Iacono and McGue2021) and authoritarianism (Eftedal et al., Reference Eftedal, Kleppesto, Czajkowski, Eilertsen, Sheehy-Skeffington, Vassend and Thomsen2020), which often have shared-environmental variance components of at least 20%.
A lack of shared-environmental influences on a trait does not mean that it is necessarily “hardwired” or immutable (e.g., mean IQ has been rising in recent decades). But it does imply that an environmental intervention must be well outside of the normal range of variation to work. For example, if you hope to increase intelligence by encouraging children to read and learn, you would need to provide more than what some parents already give: If it is true that the large amount of variation that exists on parenting, or social expectations, or number of books on the shelf, explains little or no variation in a trait, why, then, should such things suddenly make a difference if we intervene on them? Interventions that only represent nudges on factors that already vary substantially across contexts within a relevant sample are unpromising, regardless of which part of the causal chain from genotype to phenotype they purport to address.
A critical caveat to the conclusion that the shared environment has little influence, however, is the fact that this general pattern of findings primarily stems from samples confined to single cultures, typically white, educated, industrialized, rich, and democratic (WEIRD) ones, at one single point in time. That is, behavioral genetic studies can only conclude that there is no effect of the shared environment within their samples, but these samples mostly fail to measure the substantial variation in environments that we see across human cultures and ecologies in both space and time. For example, while attitudes toward group hegemony, as reflected in SDO, are unaffected by shared-environmental variation within a sample of Norwegian middle-aged twins (Kleppestø et al., Reference Kleppestø, Czajkowski, Vassend, Røysamb, Eftedal, Sheehy-Skeffington and Thomsen2019), there is nevertheless substantial variation across nations and states of the United States in the SDO levels of members of dominant societal groups, which correlate with ecological levels of macrostructural inequality (Kunst, Fischer, Sidanius, & Thomsen, Reference Kunst, Fischer, Sidanius and Thomsen2017). And when looking across time within cultures or countries, the increases we have seen in IQ over recent decades, for instance, are far larger than what genetic shifts alone can plausibly account for (Flynn, Reference Flynn2009). Ever more large-scale social psychological studies continue to pinpoint the ecological correlates of psychological phenomena such as these.
A lesson from behavioral genetics to those looking for effective societal interventions on shared-environmental variables is then to broaden one's horizon, and look at the things that vary across cultural ecologies, rather than the things that vary within them. The shared-environmental differences that actually seem to make a difference are those we see between, for example, Boston and Mumbai, or between 2022 and 1942, rather than those we see between the Smiths from west of town and the Harpers in the east. Behavioral genetics should broaden its empirical scope beyond single-culture WEIRD samples to adequately identify critical effects of the shared environment and thus potential societal targets of environmental intervention.
When such data collection is not feasible, investigating the unique environment of individuals, containing everything that makes relatives differ, also offers lessons for behavioral science. While this unique variance component appears to include mostly idiosyncratic and unsystematic influences (Turkheimer, Reference Turkheimer2000), there are still important things to learn from looking at how non-shared-environmental versus genetic variance components from different phenotypes are correlated. And as Franić et al. (Reference Franić, Dolan, Borsboom, Hudziak, van Beijsterveldt and Boomsma2013) describe, such correlations provide more stringent tests of latent psychological constructs than do regular factor analyses. These techniques revealed, for example, that sensitivities to being the perpetrator, victim, beneficiary, and observer of injustice are undergirded by separate and heritable latent motivations to be morally principled and opportunistic (so that the worse one reacts to injustice to oneself, the less one reacts to injustice to others). Furthermore, the unique environments that increase moral opportunism tend to also increase SDO, while genetic substrates that increase moral opportunism also increase SDO and lower altruism and generalized trust (Eftedal et al., Reference Eftedal, Kleppestø, Czajkowski, Sheehy-Skeffington, Røysamb, Vassend and Thomsen2022).
As non-shared-environmental correlations control for genetic and shared-environmental confounders, they can also help narrow our search for pairs of variables that are connected causally. For instance, individual educational attainment is associated with lower authoritarianism even when controlling for genetic and shared-environmental confounding (Eftedal et al., Reference Eftedal, Kleppesto, Czajkowski, Eilertsen, Sheehy-Skeffington, Vassend and Thomsen2020). Surely such knowledge about the underlying nature of latent psychological constructs and their potential causal connections are useful stepping stones in the search for good interventions.
Financial support
The research was funded by grants 0602-01839B from Independent Research Fund Denmark, 231157/F10 from the Norwegian Research Council, and 101040978 from the European Research Council (all to L.T.) and by center grants DGF-144 from the Danish National Research Foundation and 288083 from the Norwegian Research Council.
Competing interest
None.