1.3.1. Hierarchical effects: Timing of puberty

The first tier is energetics – caloric intake, energy expenditures, and related health conditions – which set the baseline for many developmental processes. Energetic stress (e.g., marginal nutritional conditions) and closely related conditions of pathogen stress (e.g., McDade, Reference McDade2003; Urlacher et al., Reference Urlacher, Ellison, Sugiyama, Pontzer, Eick, Liebert and Snodgrass2018) were major co-occurring causes of EM over evolutionary history. These first-tier conditions, which may be directly experienced by the organism through biological pathways (sect. 1.2),Footnote ³ induce tradeoffs that divert energy away from growth and reproduction (Table 1, rows 1–2). Such tradeoffs constrain physical growth and generally result in later pubertal maturation (Ellis, Reference Ellis2004; Ellison, Reference Ellison2003; Urlacher et al., Reference Urlacher, Ellison, Sugiyama, Pontzer, Eick, Liebert and Snodgrass2018).

The second tier becomes evident when bioenergetic resources are adequate to support growth and development (and when the person, even if well-nourished, is not in an “emergency LH stage” [Wingfield et al., Reference Wingfield, Maney, Breuner, Jacobs, Lynn, Ramenofsky and Richardson1998] caused by an acute threat to survival). In that context, ambient cues to EM (second tier) may accelerate pubertal development (as per Belsky et al., Reference Belsky, Steinberg and Draper1991).Footnote ⁴ The two-tiered model, therefore, proposes that environmentally mediated regulation of pubertal timing is hierarchically structured (Table 1, rows 3–4): Contingent firstly on energetics (and related first-tier conditions) and – when these are adequate – secondly on ambient cues to EM (Coall & Chisholm, Reference Coall and Chisholm2003; Ellis, Reference Ellis2004).

An inference of the two-tiered model is that psychosocial effects on pubertal development will be evident in high-resource settings in which the first tier is largely neutral.Footnote ⁵ That appears to be the case in high-income countries, wherein socioeconomically disadvantaged children do not experience slower or later puberty (sect. 2.3.1). In this high-income context, meta-analyses have established that childhood exposures to violence predict earlier pubertal development (d = −0.26) and its biological mediator, accelerated cellular aging (d = −0.43) (Colich, Rosen, Williams, & McLaughlin, Reference Colich, Rosen, Williams and McLaughlin2020; see also Ding, Xu, Kondracki, & Sun, Reference Ding, Xu, Kondracki and Sun2024). The two-tiered model implies that such effects would not be observed in low-resource settings because they would be swamped by energetics (see Kyweluk, Georgiev, Borja, Gettler, & Kuzawa, Reference Kyweluk, Georgiev, Borja, Gettler and Kuzawa2018). Although much research supports the primacy of energetics in regulating timing of puberty (sect. 2.3.1), some populations facing marginal nutritional conditions but high mortality pressures from other sources still display accelerated growth and pubertal development (sect. 2.1).

1.3.2. Countervailing effects: Timing of puberty versus pace of reproduction/offspring number

The two-tiered model proposes that relations between energetic stress, ambient cues to EM, and LH traits change after puberty, shifting from hierarchical regulation of growth and pubertal development (wherein energetics are the primary) to dualistic regulation of adult reproductive parameters (wherein energetics and ambient cues to EM distinctly, jointly, and independently influence the pace of reproduction and offspring number). In addition, the regulatory timeframe expands after puberty. Childhood growth and pubertal development can only be shaped by childhood experiences. By contrast, the two-tiered model assumes that, in addition to regulation by childhood experiences, adult reproductive parameters are calibrated by current conditions, which are informative regarding optimal resource allocations in people of reproductive age.

Most critically, the two-tiered model conceptualizes dualistic effects on adult reproductive parameters as countervailing (i.e., energetic stress and ambient cues to EM are hypothesized to have opposing effects on adult life histories). Responses to parasitic and infectious disease species illustrate such countervailing effects. The two-tiered model distinguishes between having a parasitic/infectious disease (first tier) and the presence of parasitic and infectious disease species in the ambient environment (second tier), as in the distinction between having malaria and inhabiting a malarial ecology. These first- and second-tier conditions have countervailing effects on current fertility: Having a parasitic/infectious disease generally constrains fertility (e.g., Lucas, Reference Lucas2013; Magadi & Agwanda, Reference Magadi and Agwanda2010), whereas living in a pathogenic environment increases community-level child mortality and, through it, supports behavioral and physiological changes in reproductive-age adults that increase fertility (see Boucekkine, Desbordes, & Latzer, Reference Boucekkine, Desbordes and Latzer2009; Bousmah, Reference Bousmah2017; McCord et al., Reference McCord, Conley and Sachs2017).

How does this occur? That is, how do exposures to ambient cues to EM promote faster pace of reproduction and higher offspring number? Mediated through central nervous system processes (sect. 1.2), second-tier conditions (ambient cues) alter key behavioral mediators of adult reproductive function. For example, in research conducted in LMICs, mortality exposures within family, community, and social networks predict higher levels of desired fertility (e.g., Broussard & Weitzman, Reference Broussard and Weitzman2020; Harman, Reference Harman1970; Owoo, Agyei-Mensah, & Onuoha, Reference Owoo, Agyei-Mensah and Onuoha2015; Smith-Greenaway & Lin, Reference Smith-Greenaway and Lin2023), which in turn strongly predicts higher achieved fertility (Cleland, Machiyama, & Casterline, Reference Cleland, Machiyama and Casterline2020; Pritchett, Reference Pritchett1994). Likewise, in LMICs, family- and community-level mortality exposures are associated with reduced likelihood of using contraception (Bhattacharya, Chakraborty, & Kim, Reference Bhattacharya, Chakraborty and Kim2023; Kuate Defo, Reference Kuate Defo, Montgomery and Cohen1998; Rosero-Bixby, Reference Rosero-Bixby, Montgomery and Cohen1998) and, through it, higher fertility (Bradshaw et al., Reference Bradshaw, Perry, Judge, Saraswati, Heyworth and Le Souëf2023). Mortality exposures also cause steeper future discounting (e.g., Cassar, Healy, & Von Kessler, Reference Cassar, Healy and Von Kessler2017; Li et al., Reference Li, Gui, Feng, Wang, Du, Gan and Luo2012; Pepper & Nettle, Reference Pepper and Nettle2013, Reference Pepper and Nettle2017; Ramos, Victor, Seidl-de-Moura, & Daly, Reference Ramos, Victor, Seidl-de-Moura and Daly2013), which in turn may promote faster reproduction (Bulley & Pepper, Reference Bulley and Pepper2017; Chesson et al., Reference Chesson, Leichliter, Zimet, Rosenthal, Bernstein and Fife2006; Pepper & Nettle, Reference Pepper and Nettle2013). Other potential behavioral mediators include: Heightened parental expectations of child mortality (e.g., Delavande, Kohler, & Vergili, Reference Delavande, Kohler and Vergili2024; Syamala, Reference Syamala2001); lowered parental care (Quinlan, Reference Quinlan2007); reduced parental investment in offspring quality (e.g., lower investment in children’s human capital; Becker, Cinnirella, & Woessmann, Reference Becker, Cinnirella and Woessmann2010; Fernihough, Reference Fernihough2017); and shortened breastfeeding duration, increased coital frequency, and decreased contraception use, as documented in women who have experienced own-child mortality (Hossain, Phillips, & LeGrand, Reference Hossain, Phillips and LeGrand2007; Kimani, Reference Kimani2001).

Although first-tier conditions, such as marginal nutrition, constrain adult female reproductive function (e.g., Bentley, Harrigan, & Ellison, Reference Bentley, Harrigan and Ellison1998; Ellison, Reference Ellison2003; Jasienska, Bribiescas, Furberg, Helle, & Núñez-de la Mora, Reference Jasienska, Bribiescas, Furberg, Helle and Núñez-de la Mora2017), the two-tiered model stipulates that natural selection shapes developmental systems to make the best of bad circumstances when faced with ambient cues to EM. Independent of energetics, making the best in this context may bias resource-allocation decisions toward faster pace of reproduction and higher offspring number (sect. 2), as implemented through behavioral mediators. Such resource-allocation decisions are predicted by LH models (sect. 1.2). In total, the two-tiered model proposes that diverse sources of EM – energetic stress and related first-tier conditions, on the one hand; ambient cues signaling premature death or heightened risk of death in one’s local environment, on the other – have dualistic yet countervailing effects on adult reproductive parameters.

Such countervailing effects occur when these different sources of EM (1) have opposing effects on the same LH outcome (as in the preceding example of parasitic and infectious disease species; represented by the two shaded rows in Table 1); and (2) have opposing effects on different LH outcomes (represented by the two bolded rows in Table 1). In this latter case, countervailing effects result in shifts toward both slower and faster LH traits, such as when harsh environmental conditions encompassing energetic stress and ambient cues to EM both constrain pubertal development and promote faster pace of reproduction/higher offspring number (sects. 2.2–2.4). That is why, in some contexts, girls who reach sexual maturity at older ages initiate reproduction sooner, have shorter interbirth intervals, and have similar or higher fertility rates than girls who mature at younger ages.

2.1.1. Early pubertal maturation occurs across diverse energetic and mortality contexts

Across human populations, children mature early under contrasting conditions. Early sexual maturation has been documented not only in populations where food is abundant (regardless of levels of EM), but also in some populations with limited food availability and high EM. This paradox is not easily resolved by a one-tiered model, which predicts a singular, directional link between energy availability, the pace of growth, and sexual development.

Consider the average age at menarche, which has decreased in many developed nations over the last century (Garn, Reference Garn1987; Nichols et al., Reference Nichols, Trentham-Dietz, Hampton, Titus-Ernstoff, Egan, Willett and Newcomb2006; Parent et al., Reference Parent, Teilmann, Juul, Skakkebaek, Toppari and Bourguignon2003), as abundant nutrition and high sugar and fat energy foods have become more prevalent. For example, with some variation across cohorts, among UK-born women, menarcheal age declined from a mean age of 13.5 years in the early twentieth century to 12.3 years in the late twentieth century (Morris, Jones, Schoemaker, Ashworth, & Swerdlow, Reference Morris, Jones, Schoemaker, Ashworth and Swerdlow2011). This decline in menarcheal age has been attributed primarily to improved nutritional and health conditions (Parent et al., Reference Parent, Teilmann, Juul, Skakkebaek, Toppari and Bourguignon2003). Contemporaneously, Savanna Pumé girls (South American hunter-gatherers), who grow up with a high physical activity level in an ecology characterized by distinct seasonal and annual fluctuations in food supply and harsh epidemiological conditions, achieve maturation at similar ages. Pumé girls experience peak gains in weight at age 11.6, the same age as the National Health and Nutrition Examination Survey reference (Kramer, Campbell, Achenbach, & Hackman, Reference Kramer, Campbell, Achenbach and Hackman2022), and reach menarche on average at age 12.9 (SD ± 1.02; Kramer, Reference Kramer2008), an age that is within the normal range of variation relative to other Indigenous South Americans (Hill & Hurtado, Reference Hill and Hurtado1996; Walker et al., Reference Walker, Gurven, Hill, Migliano, Chagnon, De Souza and Yamauchi2006). How do we explain these similarities in the pace of maturation in contrasting energetic environments? One key variable to explain this paradox, unaccounted for in a one-tiered model, is the force of mortality.

2.1.2. Both energetics and juvenile mortality account for variation in ages at menarche and first birth across small-scale societies

A comparative study of energetic conditions, mortality rates, physical growth, and ages at menarche and first birth across 22 small-scale human societies (hunter-gatherers and subsistence horticulturalists; Walker et al., Reference Walker, Gurven, Hill, Migliano, Chagnon, De Souza and Yamauchi2006) clearly rejects a one-tiered model and underscores the necessity of a two-tiered explanatory framework. First, this comparative analysis demonstrated the importance of energetics (tier one): Societies that experienced better energetic conditions (as indicated by larger adult body size) displayed faster childhood growth rates and earlier ages at menarche and first birth in females (Walker et al., Reference Walker, Gurven, Hill, Migliano, Chagnon, De Souza and Yamauchi2006) and had a greater number of surviving offspring (Hochberg et al., Reference Hochberg, Gawlik and Walker2011).Footnote ⁶ Most striking, however, after controlling for adult body size, higher rates of childhood mortality further predicted faster growth and earlier reproductive development (tier two). In multivariate analyses, age at menarche and age at first birth each occurred about 1 year earlier for every 10% decline in survivorship to age 15 (Walker et al., Reference Walker, Gurven, Hill, Migliano, Chagnon, De Souza and Yamauchi2006), indicating a two-tiered model. In total, in many small-scale societies, marginal energetic conditions and high childhood mortality rates have countervailing effects on development, correlating respectively with both later and earlier ages at menarche and first birth. A one-tiered model alone cannot explain these countervailing effects.

Closer observation of the small-scale societies included in Walker et al.’s (Reference Walker, Gurven, Hill, Migliano, Chagnon, De Souza and Yamauchi2006) cross-cultural analysis reveals that not all sources of EM have the same effects on sexual maturation. Among hunter-gatherers and horticulturalists facing harsh energetic conditions, the empirical record on age at menarche can be explained reasonably well by a one-tiered model. For example, the Ju/’hoansi (Botswana/Namibia), the Gainj (Papua New Guinea), and Asai (New Guinea) live in high mortality environments in which malnutrition plays a significant role. Children in these populations grow slowly and girls mature late (ages at menarche: 16.6–18.4 years; Walker et al., Reference Walker, Gurven, Hill, Migliano, Chagnon, De Souza and Yamauchi2006), leading Walker et al. to conclude that when mortality is based in resource deficiency, as often occurs in higher-elevation or drier environments, the pace growth and sexual development are prolonged. This pattern may make the best of bad circumstances: Smaller individuals have more maintenance cost-effective phenotypes that downregulate energy needs (Gurven & Walker, Reference Gurven and Walker2006; Kuzawa, Reference Kuzawa2005), potentially lowering mortality risk in nutritionally poor environments (Stulp & Barrett, Reference Stulp and Barrett2016). The reproductive tradeoff incurred by smaller body size (lower adult body weight) is fewer surviving offspring (Hochberg et al., Reference Hochberg, Gawlik and Walker2011; Walker & Hamilton, Reference Walker and Hamilton2008). This pattern of slow growth, late sexual maturity, small body size, and low fertility in nutritionally restrictive environments underscores the importance of the first tier.

A second tier is needed, however, to account for patterns of growth and development among hunter-gatherers and horticul- turalists facing other mortality pressures (Kramer, Greaves, & Ellison, Reference Kramer, Greaves and Ellison2009). The Savanna Pumé (Venezuela), Hiwi (Venezuela), Yanomamo (Venezuela), Baka (Cameroon), and Negritos (Philippines) all live in high mortality environments. The heightened mortality in these societies is due largely to exposure to parasites and infectious diseases, and to violence in the case of the Hiwi (Hill, Hurtado, & Walker, Reference Hill, Hurtado and Walker2007) and Yanomamo (Macfarlan, Walker, Flinn, & Chagnon, Reference Macfarlan, Walker, Flinn and Chagnon2014). Child survivorship (up to age 15) in these five populations ranges from 33% to 55% (Hackman & Kramer, Reference Hackman and Kramer2023; Walker et al., Reference Walker, Gurven, Hill, Migliano, Chagnon, De Souza and Yamauchi2006). The Savanna Pumé, Hiwi, Baka, and Negritos all display accelerated patterns of childhood growth and maturation – faster and greater linear growth across ages 3–10 followed by earlier ages at menarche than expected for their adult body size (Kramer & Greaves, Reference Kramer and Greaves2010; Walker et al., Reference Walker, Gurven, Hill, Migliano, Chagnon, De Souza and Yamauchi2006) – despite poor environmental conditions. By age 10, Baka and Hiwi girls have reached 70% of their adult size (Walker et al., Reference Walker, Gurven, Hill, Migliano, Chagnon, De Souza and Yamauchi2006), and Savanna Pumé girls have attained 87% of their adult height (Kramer & Greaves, Reference Kramer and Greaves2010). In turn, Hiwi, Savanna Pumé, and Yanomamo girls attain early ages at menarche (12.4–12.9 years; Early & Peters, Reference Early and Peters1990; Kramer, Reference Kramer2008; Walker et al., Reference Walker, Gurven, Hill, Migliano, Chagnon, De Souza and Yamauchi2006) relative to girls in other small-scale societies. Contrary to a one-tiered model, accelerated sexual development in these societies occurs despite high rates of both juvenile and subadult or adult mortality (Kramer et al., Reference Kramer, Greaves and Ellison2009; Walker et al., Reference Walker, Gurven, Hill, Migliano, Chagnon, De Souza and Yamauchi2006). Age at menarche is comparable to that observed in populations in high-income countries with abundant energy availability (see Parent et al., Reference Parent, Teilmann, Juul, Skakkebaek, Toppari and Bourguignon2003).

In total, patterns of accelerated growth and sexual development in a number of hunter-gatherer and horticultural societies experiencing high levels of pathogen stress, and in some cases high levels of violence, are consistent with a two-tiered model. Walker et al. (Reference Walker, Gurven, Hill, Migliano, Chagnon, De Souza and Yamauchi2006) hypothesize that quickly obtaining a larger body size (rapid juvenile growth) is important for combating parasites and infectious diseases because larger body size confers protection via a more robust immune system, reducing the amount of time that individuals spend in the vulnerable juvenile stage. In addition, faster developmental trajectories in this context, including earlier sexual maturation, function to increase the probability of reproduction prior to disability or death while increasing the length of the reproductive span (sect. 1.2). Such acceleration could occur by increasing juvenile growth rates (Savanna Pumé, Hiwi, Baka, and Negritos), terminating growth at smaller adult body size (Hiwi, Baka, Yanomamo, and Negritos), and/or transitioning to reproductive maturity earlier-than-expected for adult body size (Savanna Pumé, Hiwi, Baka, and Negritos) or at an early absolute age relative to adolescents in other small-scale societies (Hiwi, Savanna Pumé, and Yanomamo).

2.1.3. Conclusion

Cross-cultural data from small-scale societies indicate that life histories advance under a variety of energetic and mortality conditions. Across human populations, children mature early under contrasting conditions, with early maturation occurring in some populations with limited energy availability and high EM. Among hunter-gatherers and subsistence-based horticulturalists, a two-tiered model is needed to understand such complex patterns of growth and sexual development. The two-tiered model proposes that energetic stress and ambient cues to EM (e.g., high rates of child mortality in one’s local community) operate simultaneously – often with countervailing effects. Although the biological mechanisms are unclear, different causes and levels of mortality interact with energy availability to regulate growth and development in complex ways that a one-tiered model is challenged to explain. Based on multivariate analyses, Walker et al. (Reference Walker, Gurven, Hill, Migliano, Chagnon, De Souza and Yamauchi2006) concluded that the effects of better energetic conditions and higher juvenile mortality on earlier ages at menarche and first birth were approximately equal, indicating a two-tiered model.

Although the data reviewed in this section are largely consistent with a two-tiered model, we note two important caveats. First, our review focused on a limited number of small-scale societies for which quantitative data were available, and among which populations and consequently sample sizes are often small. Although data from these societies afford a window into developmental and reproductive processes not otherwise documented in more economically developed societies, conclusions regarding multivariate effects of energetic stress and juvenile mortality on reproductive development should be considered tentative.

Second, the two-tiered model focuses on phenotypic plasticity (i.e., development of individual differences resulting from different environmental experiences and exposures within individual lifetimes). However, Walker et al. (Reference Walker, Gurven, Hill, Migliano, Chagnon, De Souza and Yamauchi2006) did not test for phenotypic plasticity. A plausible alternative explanation for their findings is that, when child mortality regimes are driven primarily by parasites/infectious diseases/violence rather than malnutrition, mortality-driven selection pressures favor earlier/faster childhood growth and sexual maturation. This concurs with data indicating (a) that Baka and Negritos LH patterns may be largely genetically determined (Becker et al., Reference Becker, Verdu, Froment, Le Bomin, Pagezy, Bahuchet and Heyer2011; Dávila et al., Reference Dávila, Shea, Omoto, Mercado, Misawa and Baumann2002); and (b) that Savanna Pumé and River Pumé women, who are genetically from the same population, have similar timing of onset of reproduction (and total fertility), despite River Pumé women being plumper and better nourished and sustaining lower levels of infant mortality (Kramer & Greaves, Reference Kramer and Greaves2007). In total, mortality-driven selection could explain why some small-scale societies experience accelerated childhood growth and sexual development despite marginal energetic conditions. As noted by Walker and Hamilton (Reference Walker and Hamilton2008), however, this acceleration contrasts with phenotypically plastic responses to energetic stress, which decelerate growth and sexual maturation (sect. 2.3.1). In total, although the cross-cultural data reviewed in this section reject a one-tiered model and underscore the necessity of a two-tiered framework that accounts for the effects of mortality, these data do not directly support the current two-tiered model because they are agnostic regarding phenotypic plasticity.

To address these limitations, and to advance the main goal of this paper – delineating relations between energetics, mortality exposures, and the development of LH strategies across multiple scales – a diversity of approaches and data sources are needed. Thus, to complement the patterns found among small-scale societies, we next consider larger-scale societies.

2.2.1. Current path analysis

The path analysis was designed to test the hypothesis that low- resource, high-mortality environments have countervailing effects on LH traits, causing (a) later pubertal development and (b) faster pace of reproduction/higher offspring number. Although this hypothesis is causal, the path analysis employed cross-sectional data; causation cannot be inferred. Indicators of resource scarcity were treated as exogenous variables for predicting mortality (life expectancy) and, through it, puberty and fertility outcomes (see Fig. 1).

Figure 1.

Path analytic results showing between-country relations between economic/energetic resources, body mass index (BMI), life expectancy at birth, mean age at menarche, and fertility rate. All path coefficients are standardized. R ² values capture combined direct and indirect effects. *p < 0.05, ***p < 0.001.

We used Šaffa, Kubicka, Hromada, and Kramer’s (Reference Šaffa, Kubicka, Hromada and Kramer2019) publicly available cross-national dataset to test the hypothesis. All variables included in the path analysis are described in Šaffa et al. (Reference Šaffa, Kubicka, Hromada and Kramer2019), including descriptive statistics. The dataset includes 89 countries with mean menarcheal age data (the determining factor for inclusion in the original study). Resource scarcity was indexed by body mass index (BMI; kg/m²), which is commonly used as a measure of energetic condition (e.g., Black et al., Reference Black, McEwen, Smeeth, Popham, Karam and Pluess2023), and by two highly correlated indicators of resource availability: Energy consumption per capita (kcal/person/day) and gross domestic product (GDP) per capita. These two variables were standardized and averaged together (r = 0.71) to form the composite variable economic/energetic resources. We used life expectancy at birth as the index of mortality risk over the lifespan. Age at menarche (age at first menstruation) was employed as the indicator of pubertal timing. Measures of adolescent fertility (births per 1,000 women aged 15–19) and total fertility (births per woman) were standardizing and averaged together (r = 0.82) to form the composite fertility rate variable. Combining these two measures weighted fertility rate toward earlier/faster reproduction (relevant to testing LH models).

Hypotheses, data diagnostic procedures, variable construction methods, data transformations, missingness information, and data analytic methods and decisions were preregistered (https://osf.io/shgcx/). Results of the path analysis, showing standardized path coefficients, are presented in Figure 1. Estimation procedures, zero-order correlations, and maximum likelihood parameter estimates for the final model (including direct and indirect effects) are reported in the supplementary materials. As preregistered, the indirect pathways to age at menarche and fertility rate were added to the model first (to enable hypothesis testing), followed by the addition of the one statistically significant direct path (as indicated by modification indices). No additional paths were statistically significant. Fit statistics indicated an acceptable model fit (CFI = 1.00, RMSEA = 0.00, SRMR = 0.018, TLI = 1.00, chi-square with 3 df = 2.281, p = 0.516). As predicted, there were statistically significant indirect paths (a) from higher economic/energetic resources, through longer life expectancy, to both earlier age at menarche (β = −0.448, z = −7.25, p < 0.001) and lower fertility rate (β = −0.311, z = −6.94, p < 0.001); and (b) from higher BMI, through longer life expectancy, to both earlier age at menarche (β = −0.073, z = −3.53, p < 0.001) and lower fertility rate (β = −0.051, z = −0.051, p < 0.001). These effects (shown in Fig. 1) support the hypothesis that low-resource, high-mortality environments have opposing effects on LH traits.

On the one hand, the results of the path analysis demonstrate clear ecological constraints on sexual development (as per the one-tiered model). Greater resource scarcity (i.e., lower energy consumption per capita, lower GDP per capita, lower BMI) predicted lower life expectancy and, through it, later age at menarche. This between-country correlation converges with a large body of research on individual variation in pubertal timing within LMICs (sect. 2.3.1). On the other hand, the same challenging ecological and mortality conditions that predicted later age at menarche also predicted earlier and higher fertility (i.e., more children at younger ages, as per the two-tiered model). This between-country correlation likewise converges with other research examining this phenomenon at the level of individual differences in fertility within LMICs (sects. 2.3–2.4). In total, as proposed by the two-tiered model, low-resource, high-mortality conditions were associated with countervailing shifts toward both slower and faster LH traits.

2.2.2. Previous path-analytic studies link low-resource, high-mortality conditions to high fertility

To our knowledge, two previous between-country analyses have used path analytic methods to examine mortality as an intervening mechanism in the relation between ecological conditions/resource availability and fertility outcomes (though not pubertal timing). In an analysis of 191 countries, Caudell and Quinlan (Reference Caudell and Quinlan2012) specifically examined pathways leading from lower resource availability (e.g., fewer calories per capita, lower access to clean water and sanitation) to higher rates of adolescent and total fertility. Tests of indirect effects showed that these pathways operated mainly through higher rates of communicable diseases (e.g., HIV/AIDS, malaria) and, in turn, lower life expectancy. Similarly, in an analysis of 150 countries, Pelham (Reference Pelham2021) investigated pathways leading from higher pathogen loads to higher rates of adolescent and total fertility. Tests of indirect effects showed that these pathways operated mainly through higher infant mortality. These indirect paths controlled for past fertility rates and per capita GDP (Pelham, Reference Pelham2021). Taken together, these two previous path analyses – indicating that harsh conditions largely regulate fertility through their effects on morbidity and mortality – converge with the current path analysis (Fig. 1), but are each based on much larger samples of countries.

2.2.3. Changes in mortality regulate changes in fertility over time

Although higher mortality rates are strongly and reliably associated with higher fertility rates at the national level (e.g., Caudell & Quinlan, Reference Caudell and Quinlan2012; Lorentzen, McMillan, & Wacziarg, Reference Lorentzen, McMillan and Wacziarg2008; Low, Hazel, Parker, & Welch, Reference Low, Hazel, Parker and Welch2008; Murray et al., Reference Murray, Callender, Kulikoff, Srinivasan, Abate, Abate and Bililign2018) – an association that is robust even when restricted to LMICs (Bradshaw et al., Reference Bradshaw, Perry, Judge, Saraswati, Heyworth and Le Souëf2023) – correlation does not equal causation. Thus, it is important to examine mortality–fertility relations longitudinally. Herzer, Strulik, and Vollmer (Reference Herzer, Strulik and Vollmer2012) applied panel cointegration techniques to examine the extent to which changes in fertility respond to preceding changes in mortality; over the course of a century (1900–1999), reductions in mortality led to reductions in fertility (as did increases in per capita income) within the 20 countries around the world for which data were available. Other analyses, conducted in a wider range of countries, which also employed panel cointegration methods to study relations between mortality and fertility over long periods of time, reached the same conclusion (Angeles, Reference Angeles2010; Murtin, Reference Murtin2013; Sánchez-Barricarte & García-Espinal, Reference Sánchez-Barricarte and García-Espinal2017), strongly indicating that declines in mortality contribute to declines in fertility.Footnote ⁷ This pattern converges with data from sub-Saharan Africa on changes in fertility over the late twentieth to early twenty-first century: Controlling for variation in women’s educational attainment, reductions in under-5 mortality accounted for 30% (urban) and 35% (rural) of fertility declines across 30 countries (Shapiro & Tenikue, Reference Shapiro and Tenikue2017). Finally, at the individual level, micro-longitudinal data on reproductive change over the demographic transition in the Netherlands, Sweden, and Spain indicate that familial experiences of own-child survival (low mortality) played a central role in shifting reproduction toward longer interbirth intervals and lower total fertility, whereas experiences of own-child mortality had the opposite effect (Reher & Sanz-Gimeno, Reference Reher and Sanz-Gimeno2007; Reher, Sandström, Sanz-Gimeno, & van Poppel, Reference Reher, Sandström, Sanz-Gimeno and van Poppel2017; Van Poppel, Reher, Sanz-Gimeno, Sanchez-Dominguez, & Beekink, Reference Van Poppel, Reher, Sanz-Gimeno, Sanchez-Dominguez and Beekink2012).

2.2.4. Mortality–fertility relations are not confounded by country-specific factors

Of course there are country-specific factors that influence fertility, such as culture and religion, that could affect between-country correlations between mortality and fertility. One way to address this issue is to examine within-country changes in mortality and fertility over time (sect. 2.2.3). Another is to compare different states, districts, or local communities within the same country that vary in mortality rates. Such within-country analyses have been conducted in numerous LMICs and generate the same pattern of results as between-country comparisons (Table 2). Although some effects of child mortality on fertility could be directly explained by changes in reproductive physiology (e.g., reduced lactational amenorrhea), the effects of community-level mortality on fertility exceed replacement effects (sect. 2.4.1) and occur independently of variation between women in socioeconomic conditions and own-child mortality history (Table 2).

2.2.5. Mortality–fertility relations are not confounded by economic development or stage in the demographic transition

Relations between mortality and fertility rates could also be confounded by economic development. Rising per capita income and associated technological changes and demand for human capital could lead specific regions or countries to experience lower levels of both mortality and fertility (e.g., Becker, Reference Becker1992). To address this potential confound, various studies have calculated between- country correlations between mortality (e.g., under-age-5 mortality, life expectancy) and either adolescent or total fertility while controlling for GDP per capita (Angeles, Reference Angeles2010; Bulled & Sosis, Reference Bulled and Sosis2010; Pelham, Reference Pelham2021). Inclusion of this control variable did not reduce the strength of relations between mortality and fertility in any of these analyses. Further, all studies reported in Table 2, which show consistent relations between higher community-level mortality and higher levels of desired and achieved fertility within LMICs, controlled for socioeconomic variation between respondents. These mortality–fertility relations occurred across diverse levels of development, as indicated by the Human Development Index (column 1 of Table 2) and child mortality rates (column 3 of Table 2), indicating that these relations were not an artifact of economic development or stage in the mortality transition.

To evaluate the artifact explanation (i.e., potential confounding by stage in the demographic transition), Canning, Günther, Linnemayr, and Bloom (Reference Canning, Günther, Linnemayr and Bloom2013) analyzed the effects of community level mortality on fertility in sub-Saharan Africa (N = 12,151 communities in 1199 regions). Of the 46 countries included in this analysis, about half were early in the mortality transition (under-5 mortality > 20%) and half were later in the mortality transition (under-5 mortality < 20%). Controlling for education and income, the positive effects of higher community-level (and regional-level) child mortality on total fertility were large and robust within countries in each child mortality category, though effects were somewhat stronger within countries early in the mortality transition. In total, even though higher per capita income contributes to lower fertility (Herzer et al., Reference Herzer, Strulik and Vollmer2012), the link between higher mortality exposure and higher fertility cannot be explained by differences between people in socioeconomic conditions – at either the between-country or within-country level of analysis – or by a country’s stage in the demographic transition.

2.2.6. Mortality–fertility relations are not explained by reverse causation

Relations between mortality and fertility could also reflect reverse causation: Higher fertility potentially reduces child survival (reviewed in Lawson & Borgerhoff Mulder, Reference Lawson and Borgerhoff Mulder2016; McCord et al., Reference McCord, Conley and Sachs2017). To address this issue, McCord et al. (Reference McCord, Conley and Sachs2017) collected timevarying data on parasite transmission and found that pathogen-mediated increases in child mortality contributed to higher fertility. Specifically, in an analysis of 133 countries over a 50-year period (1960–2010), greater temporal variation in malaria ecology (a spatial index of country-level stability of malaria transmission) strongly predicted higher rates of malaria-related deaths and under-age-5 mortality and, in turn, higher total fertility (McCord et al., Reference McCord, Conley and Sachs2017). These analyses controlled for between-country variation in women’s educational attainment and GDP per capita. This same pattern also emerged in a longitudinal analysis of a rural Senegalese community, which showed that increases over time in community-level malaria incidence/case fatality rates predicted increases in current fertility (probability of giving birth) among women aged 15–25, controlling for own-child mortality history and socioeconomic factors (Bousmah, Reference Bousmah2017). This concurs with a larger literature showing that the effects of community-level mortality on fertility operate independently of own-child mortality (Table 2) and, therefore, cannot be explained by reverse causation.

2.2.7. Mortality–fertility relations are robust across countries with the shortest life expectancies

Finally, because of ecological/energetic constraints on reproduction (as per the one-tiered model), higher mortality might only lead to accelerated pace of reproduction and higher offspring number in countries or communities in which living conditions exceed a certain ecological/life expectancy threshold (Low et al., Reference Low, Hazel, Parker and Welch2008). For example, when comparing countries with the shortest life expectancies (<60 years; n = 32), the relation between life expectancy and age at first birth was not statistically significant (Low et al., Reference Low, Hazel, Parker and Welch2008). However, Low et al. did not incorporate relevant control variables into their analysis, such as per capita GDP and mortality rates from HIV/AIDS (during the HIV/AIDS epidemic, HIV-infected women in developing countries experienced both dramatically increased mortality and reduced fertility; e.g., Magadi & Agwanda, Reference Magadi and Agwanda2010). When these variables were controlled for, shorter life expectancy strongly and significantly predicted higher total fertility rates in countries with the lowest life expectancies (<60 years; n = 44) (Bulled & Sosis, Reference Bulled and Sosis2010). Further, a recent analysis divided 195 countries into 5 quintiles based on the Socio-Demographic Index (SDI, which is similar to the United Nations’ Human Development Index but is more sensitive to women’s status in society). Across countries and territories in the lowest SDI quintile (i.e., lowest developmental status), the under-age-5 mortality rate correlated 0.66 with the total female fertility rate and 0.68 with the under-age-25 female fertility rate (Murray et al., Reference Murray, Callender, Kulikoff, Srinivasan, Abate, Abate and Bililign2018: Appendix Table 11). The size of these correlations was very similar across all five SDI quintiles, demonstrating the ecological robustness of the association between higher child mortality and higher female fertility (see also Table 2, which supports this robustness at the community level).

2.2.8. Conclusion

Based on the one-tiered model, a straightforward prediction is that better ecological conditions lead to higher fertility because more energetic resources are available to allocate toward reproduction (Volk, Reference Volk2023). However, this prediction is not supported by between- or within-country comparisons. In fact, the opposite occurs: Resource-poor ecologies and associated reductions in life expectancy correlate with higher fertility (despite later ages at menarche). In the path analysis (Fig. 1), challenging ecological conditions (i.e., lower energy consumption per capita, lower GDP per capita) and lower life expectancy afforded valid indicators of harsh environments, as evidenced by their strong effects on later age at menarche. However, even against this backdrop of slower sexual development, the positive effects of mortality on fertility were robust, clearly demonstrating the countervailing effects of EM on different LH traits.

An extensive research literature in demography has established the effects of higher mortality on higher fertility. Across multiple path analyses (including previously published research), lower resource availability or higher pathogen loads were associated with higher fertility rates, and these effects largely operated through heightened morbidity and mortality (e.g., lower life expectancy, higher infant mortality, higher rates of communicable diseases). Most critically, the effects of mortality on fertility have proven robust against plausible alternative explanations. These effects: (a) were not confounded by country-specific factors; (b) were not reduced by controlling for per capita GDP or other indicators of economic development; (c) were not an artifact of a country’s stage in the demographic transition; (d) clearly emerged when analyzing within-country change in fertility rates over the twentieth century; (e) clearly emerged in within-country analyses comparing different states, regions, or local communities that vary in mortality rates (even after controlling for respondents’ socioeconomic status and own-child mortality history; Table 2); (f) were supported by time varying data on parasite transmission, indicating that pathogen-mediated increases and decreases in child mortality contribute to higher and lower fertility, respectively; and (g) were even present in comparisons between countries with the shortest life expectancies (<60 years), wherein harsh ecological conditions are pervasive.

2.3.1. When do harsh conditions cause later puberty? A methodological note

Before reviewing these lines of research, we present a methodological note. The between-country correlation between harsh ecological conditions and later pubertal development (Fig. 1) replicates in research on adolescent development conducted in lower-resource settings (where many children in poverty face significant energetic constraints) but not in high-resource settings. Accordingly, lower socioeconomic status predicts later timing of puberty in LMICs (e.g., Ethiopia: Belachew et al., Reference Belachew, Hadley, Lindstrom, Getachew, Duchateau and Kolsteren2011; Tanzania: Rebacz, Reference Rebacz2009; Ghana: Adadevoh, Agble, Hobbs, & Elkins, Reference Adadevoh, Agble, Hobbs and Elkins1989; Sudan: Elshiekh & Mohammed, Reference Elshiekh and Mohammed2011; Mozambique: Padez, Reference Padez2003; Nigeria: Abioye-Kuteyi et al., Reference Abioye-Kuteyi, Ojofeitimi, Aina, Kio, Aluko and Mosuro1997; South Africa: Jones, Reference Jones2008; Egypt: Attallah, Reference Attallah1978; Morocco: Montero, Bernis, Loukid, Hilali, & Baali, Reference Montero, Bernis, Loukid, Hilali and Baali1999; Chile (indigenous adolescents): Amigo, Vásquez, Bustos, Ortiz, & Lara, Reference Amigo, Vásquez, Bustos, Ortiz and Lara2012; Colombia: Jansen, Herrán, & Villamor, Reference Jansen, Herrán and Villamor2015; Peru: Aurino, Schott, Penny, & Behrman, Reference Aurino, Schott, Penny and Behrman2018; Mexico: Torres-Mejía et al., Reference Torres-Mejía, Cupul-Uicab, Allen, Galal, Salazar-Martínez and Lazcano-Ponce2005; Haiti: Allman, Reference Allman1982; Brazil: Linhares, Round, & Jones, Reference Linhares, Round and Jones1986; Venezuela: Lopez Contreras, Tovar Escobar, Farid Coupal, Landaeta Jimenez, & Mendez Castellano, Reference Lopez Contreras, Tovar Escobar, Farid Coupal, Landaeta Jimenez and Mendez Castellano1981; Pakistan: Karim, Qaisar, & Hussain, Reference Karim, Qaisar and Hussain2021; Iran: Nasiri et al., Reference Nasiri, Dolatian, Ramezani Tehrani, Alavi Majd, Bagheri and Malekifar2020; India: Aurino et al., Reference Aurino, Schott, Penny and Behrman2018; Tibet: Singh, Singh, & Kshatriya, Reference Singh, Singh and Kshatriya2020; Bangladesh: Foster, Menken, Chowdhury, & Trussell, Reference Foster, Menken, Chowdhury and Trussell1986; Indonesia: Artaria & Henneberg, Reference Artaria and Henneberg2000; Vietnam: Aurino et al., Reference Aurino, Schott, Penny and Behrman2018; Philippines: Adair, Reference Adair2001).Footnote ⁸ This finding converges with a large nutrition literature demonstrating that girls in developing countries who are either malnourished (as determined by height and weight measurements) or have limited access to food (as determined by dietary studies) have later ages at menarche than peers with high or ad libitum access to food (Costa et al., Reference Costa, Darling, Shinde, Tadesse, Sherfi, Mwanyika-Sando and Fawzi2025; Ellis, Reference Ellis2004; Glass, Geerkens, & Martin, Reference Glass, Geerkens and Martin2022).

A contrasting pattern occurs in high-resource settings. In a recent meta-analysis of data from high-income countries, lower socioeconomic status predicted significantly earlier, not later, pubertal timing (Colich et al., Reference Colich, Rosen, Williams and McLaughlin2020).Footnote ⁹ This finding suggests that individuals growing up in poverty (as a group) in high-income countries do not experience enough energetic deprivation to systematically constrain sexual development. For example, in the National Health and Nutrition Examination Survey (United States), food insecurity was not associated with lower caloric intake but did correlate with diets that were energy-dense but of poor quality (Leung, Epel, Ritchie, Crawford, & Laraia, Reference Leung, Epel, Ritchie, Crawford and Laraia2014). This is likely a novel phenomenon from a historical and evolutionary perspective (see Volk, Reference Volk2023) and could confound evaluation of one-tiered versus two-tiered models. To circumvent this issue, the following discussion of environmentally mediated variation in pubertal development and reproductive scheduling is restricted to lower-resource settings – societies in which substantial numbers of children face significant energetic constraints that compromise growth. Excluded from our analysis is research conducted in modern countries in the high-middle or high quintiles of the SDI (Murray et al., Reference Murray, Callender, Kulikoff, Srinivasan, Abate, Abate and Bililign2018).

2.3.2. Historical data link childhood exposures to sibling mortality to earlier ages at marriage and first birth

Störmer and Lummaa (Reference Störmer and Lummaa2014) used seventeenth to nineteenth century data from Germany, Canada and Finland to analyze the effects of sibling death during childhood on timing of reproduction. The Finnish and Canadian populations had high juvenile mortality rates (approximately 40%), though the infant mortality rate in the German population was comparatively low (12%). Across all three populations, children who experienced the death of a sibling while growing up had earlier ages of marriage and first birth (comparable effects also occur in contemporary LMICs; e.g., Broussard & Weitzman, Reference Broussard and Weitzman2020; Gettler et al., Reference Gettler, McDade, Bragg, Feranil and Kuzawa2015; Pebley, Delgado, & Brinemann, Reference Pebley, Delgado and Brinemann1979; Ueyama & Yamauchi, Reference Ueyama and Yamauchi2009). This accelerated onset of reproduction occurred in both males and females (Störmer & Lummaa, Reference Störmer and Lummaa2014). In addition, in the German population, women who experienced the death of a sibling in early childhood (0–5 years) had more out-of-wedlock offspring (Pink, Willführ, Voland, & Puschmann, Reference Pink, Willführ, Voland and Puschmann2020), presumably because they were more likely to begin having sex before marriage ties were formalized.Footnote ¹⁰ Consistent with LH models, accelerated reproduction in the context of ambient cues to EM in early life may be adaptive, as suggested by historical longitudinal data indicating that exposures during childhood to community mortality and sibling death provide reliable cues to premature adult mortality (Van Dijk, Janssens, & Smith, Reference Van Dijk, Janssens and Smith2019).

Historical data from Dominica, however, suggest limits on accelerated reproduction under very harsh conditions. Based on longitudinal data from a rural community in Dominica (1925–2000), periods of moderately high EM (as defined by localpopulation infant mortality rates) predicted earlier age at first birth and higher fertility, relative to periods of either low or very high EM (Quinlan, Reference Quinlan2010). Periods of very high EM (e.g., 72% infant mortality rate in 1950) were, apparently, characterized by levels of energetic deprivation and somatic depletion that were severe enough to suppress adult reproductive function (Quinlan, Reference Quinlan2010). The resulting quadratic effects – linking increasing levels of EM to both accelerated and decelerated reproduction (in an inverted U-shaped curve) – demonstrate the importance of both the first and second tiers: Although women apparently made the best of bad circumstances by accelerating reproduction under relatively high levels of EM (as per the two-tiered model), this occurred within constraints imposed by the first tier.

2.3.3. Longitudinal cohort studies show countervailing effects of energetic stress and mortality-related developmental experiences on LH traits

Both timing of pubertal development and reproductive scheduling were examined in longitudinal cohort studies in South Africa (Cape Area Panel Study), the Philippines (Cebu Longitudinal Health and Nutrition Survey), Brazil (Pelotas Birth Cohort Study), and Peru (Young Lives Study). Across these studies, complex adversity exposures involving marginal childhood nutrition together with other forms of environmental harshness (e.g., poverty, sibling death, exposure to violence) predicted both later pubertal development and earlier ages at first pregnancy or reproduction (South Africa: Anderson, Reference Anderson2015; Brazil: Wells et al., Reference Wells, Cole, Cortina-Borja, Sear, Leon, Marphatia and Menezes2019; Philippines: Gettler et al., Reference Gettler, McDade, Bragg, Feranil and Kuzawa2015; Peru: Aurino et al., Reference Aurino, Schott, Penny and Behrman2018; Favara, Lavado, & Sánchez, Reference Favara, Lavado and Sánchez2020). In total, the complex adversity exposures documented across these different studies had countervailing effects on life histories: Later puberty followed by earlier reproductive scheduling.

The most detailed analysis was conducted in the Philippines cohort. Males who experienced worse nutritional conditions in childhood tended to experience later pubertal development and later onset of sexual activity (Gettler et al., Reference Gettler, McDade, Bragg, Feranil and Kuzawa2015), as predicted by a one-tiered model. In the same sample, however, males who experienced paternal instability or the death of a sibling in childhood become fathers at a younger age (Gettler et al., Reference Gettler, McDade, Bragg, Feranil and Kuzawa2015), consistent with a two-tiered model. Likewise, across waves of the Philippines National Demographic and Health Survey, women living in poverty reported later ages at menarche (presumably because of unfavorable childhood nutrition; see Kyweluk et al., Reference Kyweluk, Georgiev, Borja, Gettler and Kuzawa2018) but earlier ages at sexual debut and much higher fertility rates on average (Lai & Tey, Reference Lai and Tey2014; Tey, Lai, & Ng, Reference Tey, Lai and Ng2019). Taken together, these findings concur with the results of the between- country path analysis (Fig. 1), which showed that resource scarcity and associated increases in mortality were associated with both later ages of menarche and earlier/higher fertility.

2.3.4. Ordinary, day-to-day mortality exposures predict higher levels of desired and achieved fertility: Family, social network, and community-level data

Other research has specifically examined the effects of mortality exposures on variation in fertility preferences, timing, and outcomes. Such exposures have been studied at multiple levels. First, childhood experiences of sibling mortality, which in LMICs strongly predict subsequent own-child mortality (Smith-Greenaway, Weitzman, Lin, & Huss, Reference Smith-Greenaway, Weitzman, Lin and Huss2023), forecast earlier onset of reproduction (sect. 2.3.2). Second, in LMICs, parents tend to respond to the death of an infant or child by having more children (e.g., Gyimah & Fernando, Reference Gyimah and Fernando2004; Kuate Defo, Reference Kuate Defo, Montgomery and Cohen1998; Lindstrom & Kiros, Reference Lindstrom and Kiros2007). Third, beyond family-level effects, community-level mortality appears to regulate fertility in LMICs. Empirical investigations of this phenomenon are summarized in Table 2. Those investigations show that, independent of own-child mortality and socioeconomic factors, residing in an area where other residents experience elevated levels of child mortality uniquely predicts lower contraception use, shorter interbirth intervals, and higher levels of desired and achieved fertility.Footnote ¹¹ Further, independent of socioeconomic conditions, living in a community where others experience elevated levels of lethal violence or premature death correlates with earlier ages at first reproduction, as documented among people in (a) Latin America living in areas with high homicide rates (Braverman-Bronstein et al., Reference Braverman-Bronstein, Vidaña-Pérez, Ortigoza, Baldovino-Chiquillo, Diez-Canseco, Maslowsky and Roux2022; Svallfors, Reference Svallfors2024) and (b) high-income countries residing in areas with short local life expectancies (Copping, Campbell, & Muncer, Reference Copping, Campbell and Muncer2013; Uggla & Mace, Reference Uggla and Mace2016; Wilson & Daly, Reference Wilson and Daly1997). In Colombia, for example, girls who experience consistent community violence (they live in an area with high exposure to armed conflict/violence and then move to another area with high exposure) have more than twice the odds of becoming adolescent mothers than girls who experience decreasing community violence (they live in an area with high armed conflict/violence but then move to a peaceful area) (Sanchez-Cespedes, Reference Sanchez-Cespedes2018).

Such community-level effects, which emerge reliably across diverse settings, underscore the importance of ambient cues to EM in regulating reproduction. Taken together, mortality exposures across different phases of development, from childhood to reproductive age, and in different contexts, from intimate losses in the family to deaths of children in the community to local homicides, predict fertility preferences, timing, and outcomes, as per the two-tiered model.

To better understand these effects in the context of specific exposures to EM, two studies collected nuanced individual-level data on “variation in exposures to ordinary, day-to-day mortality in one’s community and social network” (Smith-Greenaway et al., Reference Smith-Greenaway, Yeatman and Chilungo2022, p. 565). This approach affords a sensitive, personal indicator of EM risk. Both studies were conducted in countries (Malawi and Nepal) experiencing high mortality.

Over a 2-year period, the Malawi study repeatedly assessed indicators of recent mortality exposure in terms of women’s funeral attendance in the prior month (Smith-Greenaway et al., Reference Smith-Greenaway, Yeatman and Chilungo2022). Malawi has one of the lowest life expectancies in the world, and women in this study attended an average of one funeral per month, though with wide variation across participants. The number of funerals attended by individual women prospectively predicted higher rates of pregnancy (confirmed through urine hCG tests), and especially unintended pregnancy, even after controlling for socioeconomic factors.

The Nepal study directly measured rates of infant mortality (child loss) among individuals belonging to women’s social networks (i.e., sociometric network alters) (Sandberg, Reference Sandberg2006). Consistent with the 33% under-5 mortality rate in Nepal at the time of the study (Sandberg, Reference Sandberg2006), the average infant mortality rate among network alters was about 23% (based on 10-year retrospective fertility histories), with wide variation. Higher and more consistent experiences of infant mortality among network alters corresponded to a higher probability of women giving birth within the year, even after controlling for women’s own-child mortality history.

2.3.5. Conclusion

The literature on energetic stress in the context of ambient cues to EM provides a textbook case of the complex – and often countervailing – nature of developmental responses to harsh environments. Despite the negative effects of energetic deprivation on growth and pubertal development, broader phenotypic responses may still make the best of bad circumstances, particularly in response to strong mortality cues experienced in one’s family, community, or social network.

Such responses – in historical studies of childhood mortality exposures in Germany, Canada, and Finland; in longitudinal cohort studies in South Africa, the Philippines, Brazil, and Peru; in research on social network mortality in Malawi and Nepal; and in analyses of community-level mortality in LMICs around the world – involve allocation of resources toward earlier/faster reproduction and higher offspring number, as per the two-tiered model. In the literature reviewed in this section, these allocations involved shifts toward faster demographic LH traits (i.e., earlier ages at first pregnancy and first reproduction, shorter interbirth intervals, higher total fertility) and presumed mediators of these traits (i.e., earlier onset of sexual activity, higher desired fertility, reduced contraception use).

More broadly, these reproductive responses are consistent with a well-documented phenomenon in LMICs: Women who experience low-income and economic marginalization – and thus elevated levels of morbidity and mortality (Banks, Kuper, & Polack, Reference Banks, Kuper and Polack2017; Ellis et al., Reference Ellis, Figueredo, Brumbach and Schlomer2009) – marry at younger ages, use contraception less, and have more children (e.g., Hackman & Hruschka, Reference Hackman and Hruschka2020; Hruschka, Sear, Hackman, & Drake, Reference Hruschka, Sear, Hackman and Drake2019; Schoumaker, Reference Schoumaker2004), despite later pubertal development (sect. 2.3.1). Likewise, the longitudinal cohort studies showed countervailing effects of harsh conditions involving energetic stress and other mortality-related exposures on the development of LH strategies: Later puberty followed by earlier onset of pregnancy or reproduction. Such countervailing effects support the two-tiered model. In total, the diversity of research reviewed in this section clearly contradicts the one-tiered model – even if the first tier ultimately constrains the observed shifts toward earlier/faster reproduction (as apparently occurred in Dominica).

2.4.1. In LMICs, exposures to natural disasters lead to higher fertility, despite reductions in childhood growth

Natural disasters can have myriad negative effects on the lives of people exposed to them (e.g., displacement, economic disruption, loss of social services, food scarcity, disease, and death). When examining the aggregate effects of all kinds of disasters on populations as a whole, including short-term impacts, the overall change in live birth rates may actually be negative (Lee, Batyra, Castro, & Wilde, Reference Lee, Batyra, Castro and Wilde2023). The one-tiered model implies that such negative outcomes result from deteriorating energetic conditions (e.g., prolonged disruption of food supplies). Decreases in fertility following poor harvests in eighteenth-century Finland (Rickard et al., Reference Rickard, Holopainen, Helama, Helle, Russell and Lummaa2010), crop failures in Ireland in the 1840s (Boyle & Gráda, Reference Boyle and Gráda1986), and various droughts in Africa over the last 50 years (Norling, Reference Norling2022) concur with this assumption.

Here we focus on a particular kind of natural experiment that is most causally informative for evaluating the two-tiered model. It involves three criteria. First, a spatially and temporally localized natural disaster results in a major mortality shock (extensive community-level death and destruction). This first criterion excludes long-lasting disasters, such as drought, that unfold gradually over time and space. Second, demographic data on fertility were collected prior to the shock and then again after the shock. Third, fertility outcomes of people living in communities that directly experienced the shock are compared with fertility outcomes of people in neighboring communities that did not experience the shock (or experienced it to lesser degrees). By specifying comparisons between local communities, this third criterion minimizes sociocultural and geographic confounds. Taken together, these criteria implement a difference-in-differences design, which is a widely used method for testing the effects of a policy or event.

The effects of five major natural disasters on fertility (in countries or territories in the lower three quintiles of the SDI) have been examined using this difference-in-differences design. These studies examined responses to the 2004 Indian Ocean tsunami (Nobles, Frankenberg, & Thomas, Reference Nobles, Frankenberg and Thomas2015; see also Jung, Reference Jung2024), the 2001 Gujarat earthquake in India (Nandi, Mazumdar, & Behrman, Reference Nandi, Mazumdar and Behrman2018), the 2010 Haiti earthquake (Behrman & Weitzman, Reference Behrman and Weitzman2016; Harville & Do, Reference Harville and Do2016), the 2005 North-West Frontier earthquake in Pakistan (Finlay, Reference Finlay2009), and Hurricane Mitch in Nicaragua in 1998 (Davis, Reference Davis2017). Each of these natural disasters caused widespread morbidity and mortality. Across all five populations, women living in areas that directly experienced the natural disaster showed significantly greater increases in fertility (e.g., childbirth rates, pregnancy rates) from pre-disaster to post-disaster than did women living in neighboring communities that did not experience the disaster (or experienced it to lesser degrees). These increases in fertility occurred despite deterioration of living conditions (higher energetic stress), as indexed by well-documented reductions in childhood growth that occurred in the wake of each of these five natural disasters (Aceh, Indonesia: Frankenberg, Friedman, Ingwersen, & Thomas, Reference Frankenberg, Friedman, Ingwersen and Thomas2017; Gujarat: Dhamija & Sen, Reference Dhamija and Sen2023; Haiti: Dodlova, Escobar, & Grimm, Reference Dodlova, Escobar and Grimm2023; Pakistan: Andrabi, Daniels, & Das, Reference Andrabi, Daniels and Das2023; Nicaragua: Omitsu & Yamano, Reference Omitsu and Yamano2006). In total, the significant increases in fertility that occurred across all five disaster sites – despite harsh energetic conditions that constrained childhood growth – support the two-tiered model.

Two of the five studies (Finlay, Reference Finlay2009; Nobles et al., Reference Nobles, Frankenberg and Thomas2015) distinguished between replacement effects (when parents who directly lose a child seek to have another child to compensate for their loss) and insurance effects (consistent with LH models, when parents who perceive high risk of extrinsic child mortality increase their family size to hedge their bets against child death). Analysis of fertility responses to both the 2004 Indian Ocean tsunami and the 2005 North-West Frontier earthquake in Pakistan showed evidence of both replacement and insurance effects (Finlay, Reference Finlay2009; Nobles et al., Reference Nobles, Frankenberg and Thomas2015), indicating the people who experienced these natural disasters not only attempted to replace the children they lost, but also shifted toward higher overall fertility. Such insurance effects converge with the literature reviewed above on the links between higher mortality rates in family, community, and social networks and earlier/faster reproduction (sect. 2.3.4; Table 2).

2.4.2. Prolonged exposure to high-intensity armed conflicts promotes higher fertility in low-resource settings

A substantial literature has examined the impact of violent conflicts on fertility outcomes. Localized exposure to armed conflict is generally associated with an initial decline in fertility followed by an increase in fertility in the early postwar period (e.g., Caldwell, Reference Caldwell2004; Heuveline & Poch, Reference Heuveline and Poch2007; Kraehnert, Brück, Di Maio, & Nisticò, Reference Kraehnert, Brück, Di Maio and Nisticò2019; Thiede, Hancock, Kodouda, & Piazza, Reference Thiede, Hancock, Kodouda and Piazza2020). In addition, children exposed to armed conflicts tend to experience reductions in growth (stunting, wasting, underweight) and later ages at menarche in girls (Clarkin, Reference Clarkin2019; Le & Nguyen, Reference Le and Nguyen2023). As reviewed by Black et al. (Reference Black, McEwen, Smeeth, Popham, Karam and Pluess2023), these effects are likely mediated by energetic stress, as per a one-tiered model.

Although outbreaks of armed conflict generally have negative short-term effects on fertility, long-lasting and high-intensity conflicts are associated with higher fertility rates, and this link is specific to low-income countries (Urdal & Che, Reference Urdal and Che2013). Specifically, Urdal and Che (Reference Urdal and Che2013) used timeseries data over 1970–2005 to examine country-level correlations between conflict intensity and fertility. During that time period, there was a strong general decline in fertility rates, but the fertility transition was interrupted in low-income countries exposed to protracted armed conflict. Conflict intensity (measured as number of battle-related deaths) was associated with higher female fertility rates in low-income countries, even after controlling for infant mortality rates (Urdal & Che, Reference Urdal and Che2013). Further, across sub-Saharan Africa, girls who were exposed to warfare during the first decade of life achieved higher fertility later in life (+0.26 children) than peers who were not exposed (controlling for socioeconomic factors; Madsen & Finlay, Reference Madsen and Finlay2019). In total, contrary to a one-tiered model, both childhood and adult exposures to warfare-related mortality predict higher female fertility rates in low-resource settings.

A small body of research has examined the effects of protracted armed conflicts on both childhood growth (e.g., height-for-age, age at menarche) and female fertility within individual countries. In an analysis of spatial and temporal variation in exposure to the Nepalese civil war (1996–2006), Nepal, Halla, and Stillman (Reference Nepal, Halla and Stillman2023) found that children in villages affected by the conflict experienced reductions in growth (lower height-for-age), but that women in those villages experienced higher fertility, relative to children and women in villages not affected by the conflict. This increase in fertility apparently resulted in a quantity–quality tradeoff, as mothers exposed to the Nepalese conflict had more children but at lower weight-for-height (Phadera, Reference Phadera2021). Similar growth and fertility responses (e.g., lower weight-for-age, later ages at menarche, but higher overall fertility rates) have also been documented among people exposed to the decades-long civil conflict in Colombia (Castro Torres & Urdinola, Reference Castro Torres and Urdinola2019; Villamor, Chavarro, & Caro, Reference Villamor, Chavarro and Caro2009), the prolonged civil war in Mali (Torrisi, Reference Torrisi2024; Tranchant et al., Reference Tranchant, Gelli, Bliznashka, Diallo, Sacko, Assima and Masset2019), the protracted Boko Haram insurgency in Nigeria (Ekhator-Mobayode & Asfaw, Reference Ekhator-Mobayode and Asfaw2019; Rotondi & Rocca, Reference Rotondi and Rocca2022), and recurring terrorist activity in Pakistan (Grossman, Khalil, & Ray, Reference Grossman, Khalil and Ray2019; Javeid, Pratt, Li, & Zhao, Reference Javeid, Pratt, Li and Zhao2023). In total, protracted armed conflicts in low-resource settings appear to have countervailing effects on LH traits, as evidenced by slower growth and later puberty but faster pace of reproduction and higher offspring number.

Despite such countervailing effects, the causal link from warfare- and disaster-related mortality to LH outcomes could be confounded by third variables (e.g., population displacements, loss of access to health care and contraceptive services). One research project largely obviated such confounds by using within-family comparisons. Specifically, Lynch et al. (Reference Lynch, Lummaa, Briga, Chapman and Loehr2020) examined a population of evacuees from a region of Finland during World War II. Young women evacuees were split into two groups: Those who volunteered for a woman’s paramilitary organization versus peers and sisters who did not volunteer. Volunteers and non-volunteers were differentially exposed to mortality cues and stress during the war, and were then integrated back into the same population after the war. Comparisons between sisters in the same family who did and did not volunteer showed that volunteers had earlier ages at first birth, shorter interbirth intervals, and more children after the war (Lynch et al., Reference Lynch, Lummaa, Briga, Chapman and Loehr2020). These within-family comparisons controlled for family-wide confounds such as socioeconomic status, father absence, family violence, race, and ethnicity. These results suggest that heightened exposure to wartime mortality and related stressors (e.g., assisting in military hospitals and with preparation of war dead) accelerated reproductive timetables and increased total fertility in volunteers – not just during wartime but prospectively over time.

2.4.3. Conclusion

The natural experiments reviewed in this section both converge with and diverge from the one-tiered model, which posits that limited energetic conditions (at least prior to the demographic transition) induced slower LH strategies (Volk, Reference Volk2023). Significant exposures to both natural disasters and warfare constrain physical growth, delay pubertal development, and, at least in the short term, reduce fertility. That is the first tier. At the same time, however, women in developing countries who are exposed to high levels of community mortality as a result of natural disasters, or to long-lasting and high-intensity armed conflicts (either in childhood or as adults), show accelerated reproduction and higher offspring number. That is the second tier. Such countervailing effects have been documented across a variety of natural experiments, all of which examined fertility responses to mortality shocks that were generally outside of the control of the affected people/communities. This increased fertility involved both replacement and insurance effects, occurred despite deteriorating conditions that constrained growth and delayed puberty among children in the affected communities, and was documented in comparisons between sisters with fully traced reproductive histories who, as young women, were differentially exposed to war. These results converge with the two-tiered model.