Natural selection often favours risk taking and future discounting as facultative responses to cues indicating a relatively high risk of unpredictable, uncontrollable catastrophe (Daly & Wilson Reference Daly and Wilson2005; Promislow & Harvey Reference Promislow and Harvey1990). Pepper & Nettle (P&N) make a powerful case that this insight illuminates many aspects of human development and behavioural variability, which we applaud. However, applying their model is not necessarily straightforward.

In support of their proposed “behavioural constellation of deprivation” (BCD), P&N (sect. 2, para. 4) posit “consistent” SES gradients in behaviours affecting health, one of which is that people of lower SES “are more likely to use illicit drugs and to drink excessive amounts of alcohol.” Research in rich countries certainly supports this generalisation, but studies in Latin America contradict it. Researchers in Argentina (Fantin & de Barbenza Reference Fantin and de Barbenza2007), Brazil (Baus et al. Reference Baus, Kupek and Pires2002; Macinko et al. Reference Macinko, Mullachery, Silver, Jimenez and Neto2015; Malta et al. Reference Malta, Mascarenhas, Porto, Barreto and Morais Neto2014; Muza et al. Reference Muza, Bettiol, Muccillo and Barbieri1997; Pratta & Santos Reference Pratta and Santos2007; Ramis et al. Reference Ramis, Mielke, Habeyche, Oliz, Azevedo and Hallal2012; Silva et al. Reference Silva, Malbergier, de Andrade Stempliuk and de Andrade2006; Souza et al. Reference Souza, Areco and Filho2005; Souza & Martins Reference Souza and Martins1998; Tavares et al. Reference Tavares, Béria and Lima2001), Chile (Florenzano et al. Reference Florenzano, Cáceres, Valdés, Calderón, Santander, Cassasus and Aspillaga2010; Peña et al. Reference Peña, Mäkelä, Valdivia, Helakorpi, Markkula, Margozzini and Koskinen2017; Sepúlveda et al. Reference Sepúlveda, Roa and Muñoz2011), and Mexico (Caballero et al. Reference Caballero, de León, San Martín and Villaseñor1999; Hernández & González Reference Hernández and González2013; Herrera-Vázquez et al. Reference Herrera-Vázquez, Wagner, Velasco-Mondragón, Borges and Lazcano-Ponce2004) have repeatedly found SES to be positively associated with alcohol abuse, drug use, and smoking in adolescents and adults.

A partial reason for this discrepancy is that in poorer countries, the destitute lack the financial means to use and abuse substances, but there may be larger issues regarding how the BCD model can be applied in different contexts. There is meta-analytic evidence of greater variation in the relationship between SES and alcohol problems in low- to middle-income countries than in high-income countries (Grittner et al. Reference Grittner, Kuntsche, Graham and Bloomfield2012), suggesting that what constitutes deprivation may vary, too. In Latin America, adverse colonial legacies, enduring social injustice, and extreme economic and health inequalities (Andrade et al. Reference Andrade, Pellegrini Filho, Solar, Rígoli, de Salazar, Serrate, Ribeiro, Koller, Cruz and Atun2015; Bambas & Casas Reference Bambas and Casas2001) have created situations in which the experience of deprivation is very different from in the rich world, and what constitutes a “contextually appropriate response” may also differ. P&N caution that the “deprivation” in their model refers to “the experience of various hardships” for which SES is only a “proxy,” a caveat that is appropriate psychologically but problematic for measurement and comparisons. Moreover, how SES itself should be measured is controversial (Ensminger et al. Reference Ensminger, Fothergill, Bornstein, Bradley, Bornstein and Bradley2003; Oakes & Rossi Reference Oakes and Rossi2003; Wagstaff &Watanabe Reference Wagstaff and Watanabe2003), necessitating that we evaluate alternative measures before using SES even as a proxy.

In the target article's section 3.2, P&N explain that although responding to extrinsic mortality risk with future discounting is contextually appropriate, doing so can exacerbate one's mortality disadvantage, and such amplification even operates intergenerationally, further disadvantaging the children of the disadvantaged. How is it, then, that many people in the developing world are escaping this vicious cycle? Although the poor may sometimes see little value in educating their children, they often take the opposite view. Why one response rather than the other? Reduced infant mortality and family size, plus female labour-force participation, seem to be key variables, although the causal links are complicated and bidirectional (Gakidou et al. Reference Gakidou, Cowling, Lozano and Murray2010; Goodall & Vorhaus Reference Goodall and Vorhaus2011). Agencies trying to promote the prioritisation of education often target women as the most effective agents of change (Gakidou et al. Reference Gakidou, Cowling, Lozano and Murray2010; Soares et al. Reference Soares, Ribas and Osório2010). Partly, this reflects a recognition that women are relatively likely to spend subsidies on their children and men on themselves, but it may also be the case that women are better prepared than men to adopt the longer view (Campbell Reference Campbell1999; Daly & Wilson Reference Daly and Wilson2005). Changes in child mortality and education (especially for women/girls) seem to be tightly linked, and shifts to longer time horizons can apparently occur quickly where policies support such change.

Evolved psychological mechanisms and processes are adapted to the past and do not necessarily promote fitness in novel environments. Internet pornography is avidly consumed, and motor vehicles evoke less fear than spiders. P&N are well aware of this issue, raising it implicitly in Section 7 and explicitly in Section 8.4, where they note that “[t]he BCD isn't necessarily adaptive and perceptions aren't necessarily accurate.” Nevertheless, some earlier sections of the target article invite misconstrual as claims to the contrary. Indeed, the very phrase “contextually appropriate response” is open to such misconstrual; the claim of “appropriateness” is often warranted only with respect to the direction of responses, not their magnitude. In Section 2.3, for example, P&N quote a young offender who describes his Atlanta neighbourhood as a “war” zone in which “you never know if you gonna live one minute to the next.” P&N continue, “[T]his may seem exaggerated, but . . . ,” implying that it is not – but it is! In 2001, black males in “high-risk urban environments” in the United States had a life expectancy at birth of 66.7 years (Murray et al. Reference Murray, Kulkarni and Ezzati2005), a number only modestly affected by violent deaths and too high to justify, in and of itself, a belief that one has no future. But although the young offender's words exaggerate the dangers in his milieu, his sense of deprivation is fully justified: That 66.7-year life expectancy is lower than that of any other segment of the U.S. population. The crucial deprivation is relative, and it is unsurprising that people should have evolved to care profoundly about relative deprivation, because fitness itself is relative (Daly Reference Daly2016). Statements like the young offender's abound in urban ethnographies, and the extent to which they represent braggadocio, a massive misperception of actual mortality risks, or something else remains unclear. Answering such questions is important, because they bear on the potential efficacy of providing better information.

A common denominator of these cautions is that the psychology of deprivation and risk preferences is not transparent, a problem compounded by sex differences and by the evolutionary novelty of modern environments. Applying P&N's valuable insights to the practical business of alleviating the social costs and self-destructive effects of the BCD will remain conceptually, as well as politically, challenging.