We value Pepper & Nettle's (P&N's) target article as an attempt to integrate multiple disciplines in understanding a problem of broad societal importance. However, we do not believe the proposed model has a sound basis in theory or evidence.
First, the model suffers from a fundamental disconnect in its theoretical foundation. P&N's evolutionary model is based on their claim that “principles that were originally used to understand the selective forces leading to the evolution of traits over generations … can be applied to enhance our understanding of how behaviour is shaped by people's environments within their lifetimes” (sect. 3, para. 1). The authors do not cite theory or evidence to support this crucial claim. We believe that this kind of thinking, although widespread in evolutionary psychology, is misguided. The Darwinian reasons a species is the way it is (and is different from other species) pertain to a process over many generations of selection for genes that are reproductively advantageous in the environments the species occupies. This process genetically tailors each species to its environment. There is no equivalent process that differentially tailors each individual within a species to his or her personal environment. Therefore, the same Darwinian models used to explain between-species differences in behaviour and development cannot be applied to corresponding between-individual differences.
Second, the empirical basis of P&N's model is weak. Most of the cited findings linking deprivation with future-discounting and related behaviours are correlational and suffer from a confound that renders them uninformative with respect to the proposed model (see Zietsch Reference Zietsch2016 for a broader discussion of this issue with facultative calibration models). Specifically, genetic variation influences individual differences in childhood socioeconomic conditions and future-discounting (and in turn the “behavioural constellation of deprivation” [BCD]). P&N attempt to address this genetic confound by briefly discussing a single twin study in which they acknowledge that identical twins showed greater similarity than nonidentical twins on delay-discounting tasks (Anokhin et al. Reference Anokhin, Golosheykin, Grant and Heath2011). However, they do not report the variance components estimates: Genetic variation accounted for 30% and 51% of variation in delay discounting at ages 12 and 14, respectively, while none of the variation in delay discounting was attributable to variation in the environment shared by twin pairs (shared environment), which included, for example, the socioeconomic status (SES), neighbourhood crime rate, and home environment in which the twins were raised. This pattern of results was more recently replicated by Anokhin et al. (Reference Anokhin, Grant, Mulligan and Heath2015), with estimates of substantial heritability but no effect of the shared environment. Regarding SES as a dependent variable, massive molecular genetic studies have shown that variation in social deprivation and household income can be partly attributed to common genetic variants (Hill et al. Reference Hill, Hagenaars, Marioni, Harris, Liewald, Davies, Okbay, McIntosh, Gale and Deary2016; Trzaskowski et al. Reference Trzaskowski, Harlaar, Arden, Krapohl, Rimfeld, McMillan, Dale and Plomin2014) and that the same genetic variants underlying SES also underlie educational attainment and intelligence scores (Marioni et al. Reference Marioni, Davies, Hayward, Liewald, Kerr, Campbell, Luciano, Smith, Padmanabhan, Hocking, Hastie, Wright, Porteous, Visscher and Deary2014). An explanation of the correlations between SES and BCD traits that is consistent with all of these findings is that future-discounting parents tend to provide their children with a lower SES childhood environment and with genes predisposing them to future-discounting (and BCD traits) in adulthood. In contrast, P&N's model, in which the socioeconomic environment is a main causal driver of variation in future-discounting and BCD traits, is difficult to reconcile with the behavioural genetic findings that the shared environment cannot explain individual differences in future-discounting.
Muddying the waters with regard to the problematic genetic evidence on future-discounting, P&N invoke research on gene-by-environment (GxE) interactions (sect. 4.5, para. 2). The cited evidence, though, does not reflect the current state of the field. The claim that “children living in poverty are much more heavily constrained by their environments than by any constitutional limits” (sect. 4.5, para. 2) is based on a finding from one underpowered study (Turkheimer et al. Reference Turkheimer, Haley, Waldron, D'Onofrio and Gottesman2003) that has fared poorly in much larger replications, which show instead that the genetic effect on IQ is similarly high in low-SES and high-SES families (e.g., Hanscombe et al. Reference Hanscombe, Trzaskowski, Haworth, Davis, Dale and Plomin2012). P&N's only other citation on the topic is of a candidate-GxE study (i.e., Sweitzer et al. Reference Sweitzer, Halder, Flory, Craig, Gianaros, Ferrell and Manuck2013) of the kind that was already discredited at the time of that publication because of a record of extremely low replicability (Hewitt Reference Hewitt2012). None of this evidence, or any other that we know of, alleviates the aforementioned genetic confound that undermines much of the correlational evidence cited in the target article.
Not all correlational evidence is subject to the genetic confound. For example, P&N report some cross-country associations between extrinsic mortality risk and fast-tracking reproduction that are consistent with their model. However, it is surprising in an article reviewing the effects of mortality risk and future-uncertainty that data were not brought to bear from major periods of high mortality risk, uncertainty, and/or deprivation, such as the Great Depression and World War II. As it turns out, during this time, birthrates sharply decreased and only recovered once prosperity and safety returned some 20 years later (Fishback et al. Reference Fishback, Haines and Kantor2007). Under P&N's reasoning, the onset of deprivation and high risk should have increased birthrates as people discounted the consideration of postponing for a future they may never see, whereas the observed outcomes were the opposite. The 2007–2008 financial crisis also created considerable future-uncertainty, and the same trend of falling and recovering birthrates was again observed (Livingston Reference Livingston2011). A different sort of example can be seen in data on families emigrating from Mexico to the United States – that is, a country in which the homicide rate is three times lower and the gross domestic product is eight times higher. Mexicans living in the relatively safe and prosperous United States have more children than those living in Mexico (Camarota Reference Camarota2005) – opposite to what P&N's model would seem to predict. Further, analysis of native- and foreign-born fertility rates finds that Mexicans who emigrated to the United States as children had lower mean ages at first birth than adult entrants (Glusker Reference Glusker2003).
To properly evaluate P&N's model, we need relevant findings that are sampled in a way that is unbiased with respect to their match to the theory. The target article does not meet this need; P&N admit that they only acknowledge the works that best illustrate the story they want to tell (sect. 1, para. 4). A model will always look like a good fit with the data when only the best-fitting data are presented; in this case, once important disconfirmatory evidence is considered as well – that is, the behavioural genetic findings and other observations we have mentioned here – the model seems to us unlikely at best.
We value Pepper & Nettle's (P&N's) target article as an attempt to integrate multiple disciplines in understanding a problem of broad societal importance. However, we do not believe the proposed model has a sound basis in theory or evidence.
First, the model suffers from a fundamental disconnect in its theoretical foundation. P&N's evolutionary model is based on their claim that “principles that were originally used to understand the selective forces leading to the evolution of traits over generations … can be applied to enhance our understanding of how behaviour is shaped by people's environments within their lifetimes” (sect. 3, para. 1). The authors do not cite theory or evidence to support this crucial claim. We believe that this kind of thinking, although widespread in evolutionary psychology, is misguided. The Darwinian reasons a species is the way it is (and is different from other species) pertain to a process over many generations of selection for genes that are reproductively advantageous in the environments the species occupies. This process genetically tailors each species to its environment. There is no equivalent process that differentially tailors each individual within a species to his or her personal environment. Therefore, the same Darwinian models used to explain between-species differences in behaviour and development cannot be applied to corresponding between-individual differences.
Second, the empirical basis of P&N's model is weak. Most of the cited findings linking deprivation with future-discounting and related behaviours are correlational and suffer from a confound that renders them uninformative with respect to the proposed model (see Zietsch Reference Zietsch2016 for a broader discussion of this issue with facultative calibration models). Specifically, genetic variation influences individual differences in childhood socioeconomic conditions and future-discounting (and in turn the “behavioural constellation of deprivation” [BCD]). P&N attempt to address this genetic confound by briefly discussing a single twin study in which they acknowledge that identical twins showed greater similarity than nonidentical twins on delay-discounting tasks (Anokhin et al. Reference Anokhin, Golosheykin, Grant and Heath2011). However, they do not report the variance components estimates: Genetic variation accounted for 30% and 51% of variation in delay discounting at ages 12 and 14, respectively, while none of the variation in delay discounting was attributable to variation in the environment shared by twin pairs (shared environment), which included, for example, the socioeconomic status (SES), neighbourhood crime rate, and home environment in which the twins were raised. This pattern of results was more recently replicated by Anokhin et al. (Reference Anokhin, Grant, Mulligan and Heath2015), with estimates of substantial heritability but no effect of the shared environment. Regarding SES as a dependent variable, massive molecular genetic studies have shown that variation in social deprivation and household income can be partly attributed to common genetic variants (Hill et al. Reference Hill, Hagenaars, Marioni, Harris, Liewald, Davies, Okbay, McIntosh, Gale and Deary2016; Trzaskowski et al. Reference Trzaskowski, Harlaar, Arden, Krapohl, Rimfeld, McMillan, Dale and Plomin2014) and that the same genetic variants underlying SES also underlie educational attainment and intelligence scores (Marioni et al. Reference Marioni, Davies, Hayward, Liewald, Kerr, Campbell, Luciano, Smith, Padmanabhan, Hocking, Hastie, Wright, Porteous, Visscher and Deary2014). An explanation of the correlations between SES and BCD traits that is consistent with all of these findings is that future-discounting parents tend to provide their children with a lower SES childhood environment and with genes predisposing them to future-discounting (and BCD traits) in adulthood. In contrast, P&N's model, in which the socioeconomic environment is a main causal driver of variation in future-discounting and BCD traits, is difficult to reconcile with the behavioural genetic findings that the shared environment cannot explain individual differences in future-discounting.
Muddying the waters with regard to the problematic genetic evidence on future-discounting, P&N invoke research on gene-by-environment (GxE) interactions (sect. 4.5, para. 2). The cited evidence, though, does not reflect the current state of the field. The claim that “children living in poverty are much more heavily constrained by their environments than by any constitutional limits” (sect. 4.5, para. 2) is based on a finding from one underpowered study (Turkheimer et al. Reference Turkheimer, Haley, Waldron, D'Onofrio and Gottesman2003) that has fared poorly in much larger replications, which show instead that the genetic effect on IQ is similarly high in low-SES and high-SES families (e.g., Hanscombe et al. Reference Hanscombe, Trzaskowski, Haworth, Davis, Dale and Plomin2012). P&N's only other citation on the topic is of a candidate-GxE study (i.e., Sweitzer et al. Reference Sweitzer, Halder, Flory, Craig, Gianaros, Ferrell and Manuck2013) of the kind that was already discredited at the time of that publication because of a record of extremely low replicability (Hewitt Reference Hewitt2012). None of this evidence, or any other that we know of, alleviates the aforementioned genetic confound that undermines much of the correlational evidence cited in the target article.
Not all correlational evidence is subject to the genetic confound. For example, P&N report some cross-country associations between extrinsic mortality risk and fast-tracking reproduction that are consistent with their model. However, it is surprising in an article reviewing the effects of mortality risk and future-uncertainty that data were not brought to bear from major periods of high mortality risk, uncertainty, and/or deprivation, such as the Great Depression and World War II. As it turns out, during this time, birthrates sharply decreased and only recovered once prosperity and safety returned some 20 years later (Fishback et al. Reference Fishback, Haines and Kantor2007). Under P&N's reasoning, the onset of deprivation and high risk should have increased birthrates as people discounted the consideration of postponing for a future they may never see, whereas the observed outcomes were the opposite. The 2007–2008 financial crisis also created considerable future-uncertainty, and the same trend of falling and recovering birthrates was again observed (Livingston Reference Livingston2011). A different sort of example can be seen in data on families emigrating from Mexico to the United States – that is, a country in which the homicide rate is three times lower and the gross domestic product is eight times higher. Mexicans living in the relatively safe and prosperous United States have more children than those living in Mexico (Camarota Reference Camarota2005) – opposite to what P&N's model would seem to predict. Further, analysis of native- and foreign-born fertility rates finds that Mexicans who emigrated to the United States as children had lower mean ages at first birth than adult entrants (Glusker Reference Glusker2003).
To properly evaluate P&N's model, we need relevant findings that are sampled in a way that is unbiased with respect to their match to the theory. The target article does not meet this need; P&N admit that they only acknowledge the works that best illustrate the story they want to tell (sect. 1, para. 4). A model will always look like a good fit with the data when only the best-fitting data are presented; in this case, once important disconfirmatory evidence is considered as well – that is, the behavioural genetic findings and other observations we have mentioned here – the model seems to us unlikely at best.