The past few years has seen an increase in the number of publications on the social biogeography of human cognitive ability. Through the judicious application of exploratory path modeling (sequential canonical cascade models; see Figueredo & Gorsuch, 2007), various attempts have been made to predict the average IQ of both national and subnational polities from ambient ecological conditions (Cabeza de Baca & Figueredo, 2014, 2017; Black et al., 2017; Fernandes et al., 2017; Fernandes & Woodley of Menie, 2017; Figueredo, Cabeza de Baca, Fernandes et al., 2017a, b, c; Figueredo, Cabeza de Baca, & Peñaherrera-Aguirre, 2017). These ambient conditions included the following: (1) the physical ecology, such as mean annual temperature and precipitation, as well as weighted combinations of latitude and altitude; (2) the community ecology, such as the geographical prevalence of temperate broadleaf deciduous forests, human parasite burden, and human life history strategy; (3) the social ecology, such as the degrees of social and sexual equality, as well as the prevalence of both within-group and between-group conflict; and (4) the cultural ecology, such as the degrees of strategic differentiation in life history and macroeconomic diversification in employment, production, and exports, and the consequent development of embodied human capital. These were not modeled as competing alternative hypotheses within a single prediction equation, but instead arrayed as a sequential cascade of consequences, each successive level building upon all the previous ones by means of both direct and indirect effects.

All of these evolutionary models are ultimately, if only implicitly, predicated upon the expensive tissue hypothesis (ETH), which states that human brain tissue represents an unusually costly investment for the organism, and that this massive investment must be compensated for by means of other adaptive benefits (or cost-savings) for the organ to have evolved to the size and complexity that it has currently achieved (e.g., Aiello & Wheeler, 1995; Isler & van Schaik, 2006; Fish & Lockwood, 2003; Fonseca-Azevedo & Herculano-Houzel, 2012; Navarrete, van Schaik, & Isler, 2011). If this bioenergetic assessment is correct, then it would follow that different adaptively relevant environments (see Figueredo, Brumbach, Jones, Sefcek, Vásquez, & Jacobs, 2007; Irons, 1998) might select for different tradeoffs between enhanced brain function and other adaptive traits that might locally be more or less important to survival and reproduction, depending upon the ambient ecology. According to the logic of life history (LH) theory, the relative allocation of material and bioenergetic resources to brain development might thus vary substantially based on the relative premium placed by selection on optimizing competing components of fitness that are also demanding of resources (for a review of LH theory, see Ellis et al., 2009).

For example, it has been observed that the worldwide distribution of IQ is partially predictable from the degree of parasite stress experienced by different human populations (r = − 0.76 to r = − 0.82; Eppig, Fincher, & Thornhill, 2010), by virtue of the fact that fending off parasitic infections can be extremely expensive in bioenergetic resources and that this existential priority might locally outweigh the benefits of elevated cognitive ability in terms of survival and reproduction. Other single-factor theories will be reviewed below, but all share the basic logic that the expensive human brain must be somehow “paid for” within the inexorable balance sheets of natural selection, and that individual results may vary as a consequence. These single-factor theories are reviewed presently because they are grounded in evolutionary reasoning and based on ecological facts that each has some validity, as well as some supporting evidence. We make the argument that the significance of these more limited theories are subsumed within our broader biogeographical view of intellectual diversity based on zoogeographical regions, which we will subsequently articulate and test, demonstrating that the macroecological perspective is superior to each of the single-factor theories and in fact to all of them combined.

All of these evolutionary models, whether they are single-factor theories or not, nevertheless rely on a common functionalist perspective that may be worth commenting upon if only briefly. This shared perspective is perhaps best illustrate it by contrasting it with its opposite. In the twentieth century fable, Jonathan Livingstone Seagull (Bach, 1970), the eponymous protagonist rejects the crass materialistic values of his fellow seagulls, as parodied by the “Law of the Flock”:

Life is the unknown and the unknowable, except that we are put into this world to eat, to stay alive as long as we possibly can. (p. 39).

Instead, Jonathan professes wanting to fly for the sake of flying, and this is used as a metaphor for the supposed drive towards spiritual self-actualization (cf., Maslow, 1962; but also see Kenrick, Griskevicius, Neuberg, & Schaller, 2010, for a more evolutionarily informed model).

Perhaps as a result of trends in Western Culture towards postmaterialist values (Inglehart, 1971), intelligence has become widely touted as the summum bonum of human existence. This stance has been taken implicitly by both detractors and proponents of human intelligence research (e.g., Gould, 1996; Murray, 2003). Nevertheless, evolutionary science has to come down on the side of the prosaic seagulls and view cognitive abilities as a means to an end, and a bioenergetically expensive one at that. It is one instrumentality among other alternative instrumentalities, selected with respect to the evolutionary consequences of survival and reproduction. Darwinian theory does not represent spiritual self-actualization as the goal or the ultimate endpoint of the evolutionary process.

As evolutionary scientists, it is not our role to make value judgments, one way or the other. However, we believe it important to point out the intelligence should not be taken a metric of total human worth, but merely one adaptation among many in the service of purely functionalist adaptive goals, or fitness functions (e.g., Kenrick, Griskevicius, Neuberg, & Schaller, 2010).

Single-Factor Theories

Following migration out of Africa into Eurasia occurring over 60,000 years ago, along with later migratory events to South East Asia, Australia, New Zealand, and the Americas, humans became a cosmopolitan species with permanent settlements on six of the seven continents. Adaptation followed migration, with mean population cognitive ability hypothesized to have changed along ecologically induced clines, different aspects of which were emphasized by prominent single-factor theories of intelligence. Cold winters theory was posited to promote encephalization and general intelligence as adaptations to harsher environments, in which the extraction of resources needed for survival and reproduction became increasingly more difficult. Parasite stress theory, while acknowledging tailwinds of cold winters, emphasized the headwinds created by parasitical infections on intellectual development, characterizing Africa and climatically similar regions as imposing significant parasite burdens. Life history theory contextualized the evolved responses to both colder winters and reduced parasite stress within the broader framework of an entire coordinated suite of behavioral adaptations that was not limited to an increase in cognitive ability. In this present section, therefore, each of these theories will be reviewed, along with the associated causal factors each emphasizes.

Cold Winters Theory

According to cold winters theory (Lynn, 1991), the evolution of high cognitive ability derived from post-migration exposure to extreme cold within Eurasian climates. In positing that cold climates created a tailwind augmenting post-migration intellectual evolution, Lynn’s thesis was a within-species application of Jerison’s (1973) observation that migration’s imposition of novel adaptive problems predicted encephalization across species (Kanazawa, 2012), which, in turn, was a specific instance of migration load, or the general observation that migrating organisms experience selective pressures associated with environmental novelty (Perin, 2009; Hertler, 2015). Still, Lynn very specifically emphasized the evolutionary effects of cold, over and above any broad based allusion to novelty. Notwithstanding some morphological adaptation as described by human’s conforming to Allen’s rule and the Bergmann effect (Beall, Jablonski, & Steegmann, 2012), cold selected for future oriented thought, powers of innovation and observation, and related forms of cognitive evolution because it required the creation of shelter and clothing, and the technical challenges that these tasks imposed. These in turn, unavoidably promoted tool creation and tool use, both of which were g-loaded, cognitively demanding tasks.

The observation that average IQ significantly correlates with low temperatures for winter and high temperatures for summer is supportive of cold winters theory (Templer & Arikawa, 2006; Eppig, Fincher, & Thornhill, 2010). This correlation is not as strong as predictions would have it, however, partly due to the disappearance or reversal of the effect as one transitions past temperate zones and into the arctic circle. Nevertheless, cold winters theory gains support from the increasing complexity of Paleolithic tool kits found in more northerly distributed anthropological sites (Frost, 2019). Additional support comes from varying hunting practices. As a general rule, hunting in northern climates centers on larger game and more reliably employs traps and snares, both of which factors presumably increase the cognitive demands of capture (Bailey, 1983).

Parasite Stress Theory

Eppig, Fincher, and Thornhill’s (2010) parasite stress theory is based on the well-grounded proposition that the metabolic costs of fighting infectious disease impose an energetic brake upon brain development. Eppig and colleagues describe four methods of parasite-induced bioenergetic drag: (1) parasites feed on host tissues, (2) limit nutrition intake by inducing diarrhea, (3) coopt host cellular machinery in the service of viral reproduction, and (4) elicit a costly immune reaction within hosts, as a result of all of the above. In support of this prediction, Eppig and colleagues’ analysis of parasite stress and population-level IQs produced significant correlations in the predicted direction. Their overall results establish parasite pressure as a significant predictor of national IQ estimates established by Lynn and Vanhanen (2006), as well as the corrections to some of these estimates suggested by Wicherts and colleagues (2010). Furthermore, Eppig and colleagues found cross-national relationships between parasite stress and intelligence estimates recapitulated independently within five of the six major geographical regions (World Culture Areas) defined by Murdock (1949). Moreover, they cite negative relationships between parasite presence and various proxies for intelligence, such as educational attainment and earnings. Rising incomes following eradication of hookworm in the American South are also discussed in this context.

Eppig and colleagues regard latitude, precipitation, and mean annual temperature as predictors of parasite distribution, which alone is theorized as directly causally connected to intelligence. Eppig and colleagues subordinate climate, an aspect of the physical ecology, to parasite prevalence, an aspect of community ecology. This is only justified to a certain extent, given physical ecology may also operate as an independently relevant causal factor, as posited by cold winters theory. In addition to ignoring climate’s independent causal relevance, parasite stress theory seems not to reconcile Europe’s high IQ estimates with epidemics of smallpox, typhus, plague, diphtheria, and measles serving as potent selective pressures during the later phases of the first epidemiological transition, with its post-Neolithic density-dependent disease burden (Koepke, 2014).

Life History Theory

In addition to exacerbating the direct effects of natural selection acting upon individuals, cold winters are hypothesized to have augmented cognitive evolution indirectly through sexual and social selection. Miller (1991, 1994) emphasized the importance of male-dominated hunting of animals over female-dominated gathering of plant materials, as one progressed from southern to northern latitudes. Female choice correspondingly came to favor delay of gratification, anxiety, and altruism over aggression, extraversion, and dominance. This is consistent with the findings of Gangestad and Buss (1993) that elevated parasite prevalence (more typical of southern latitudes) tips the balance cross-culturally towards tradeoffs favoring greater physical attractiveness in mates (presumably reflecting better pathogen resistance) over other qualities, such as higher levels of parental investment (Gangestad, Haselton, & Buss, 2006; Robson & Kaplan, 2003).

Post-migration social selection also centered on the challenges of acquiring food in northern latitudes. These challenges came in the form of cooperative big game hunting, reliance on which was potentiated by seasonality’s restriction of foraging for plant matter for nearly half the year. Thereafter, as hunting of animal protein became rarer and farming of vegetable crops became more common, northerly climates imposed the added cognitive burden of future oriented thinking and planning. Similarly, Rushton (2000) seized upon cold winters as the evolutionary pressure driving both the slowing of life histories and the augmenting of cognitive evolution. In doing so, Rushton (1999) subsumed the direct effects of colder climates, as well its derivative implications for sexual and social evolution, as can be seen in the passage below:

I hypothesize that the farther north the populations migrated, out of Africa, the more they encountered the cognitively demanding problems of gathering and storing food, gaining shelter, making clothes, and raising children successfully during prolonged winters. Similarly, the winters were socially demanding, putting a premium on cooperation and impulse control. As the original African populations evolved into present-day Europeans and East Asians, they did so in the direction of larger brains, slower rates of maturation, and lower levels of sex hormone with concomitant reductions in sexual potency and aggression and increases in family stability and longevity (p. 60).

Thus, the observed increase in general intelligence with both colder winters and reduced parasite burden is contextualized by both Miller (1991, 1994) and Rushton (1999, 2000) within a broader framework of evolved responses, involving an entire coordinated suite of behavioral adaptations that characterizes slower life history strategies. Life history strategies are what regulate the inevitable tradeoffs in bioenergetic and material resources among different components of fitness, such as between survival and reproduction (Ellis et al., 2009).

In this view, the nutritional and ergonomic stresses imposed by cold winters as well as the bioenergetic sinks created by elevated parasite burden are better understood in the context of the expensive tissue hypothesis, which is supported by research on fish (Kaufman, Hladik, & Pasquet, 2003) and frogs (Liao et al., 2016), as well as nonhuman and human primates (Aiello & Wheeler, 1995; Aiello, 1997). The costs of brain tissue development and maintenance are similarly aggregated into Isler and van Schaik’s (2009) expensive brain hypothesis, wherein the finite energy allocated to the brain becomes unavailable for other important biological functions (such as digestion, locomotion, or reproduction), in a manner consistent with life history theory’s bioenergetic tradeoffs. The logic of these hypotheses rests upon the reality of inescapability high costs associated with operating and developing large brains that afford high intelligence and phenotypically plastic action.

The striking speed of Odontodactylus scyallarus, the mantis shrimp (Zack, Claverie, & Patek, 2009), the tightly coiled tongue of Bolitoglossa dofleini, Central America’s giant palm salamander (Scales, Stinson, & Deban, 2016), and the biomechanics of Utricularia stellaris, a predatory bladderwort plant (Singh, Prabhakar, & Sane, 2011), illustrate a truism in nature, wherein great speed comes from the release of potential energy. The processing speed of the human brain can be classed among these examples, as it is reliant on the storage of potential energy. Neuronal axon potentials unfold with prodigious speed by allowing sudden release of an unnatural gradient of extracellular sodium and intracellular potassium. The constant operation of protein pumps, responsible for creating and maintaining this gradient, is energetically costly, contributing to the brain’s outsized consumption of oxygen and ATP, and its tight correlation with basal metabolic rate (Isler & van Schaik, 2006).

As brains are bioenergetically costly to operate, so they are to develop. Brain mass, for example, correlates with longevity across species of birds (Minias & Podlaszczuk, 2017) and mammals (González‐Lagos, Sol, & Reader, 2010), in part, because large brains are reliant on attenuated age-related increases in mortality (Isler & van Schaik et al., 2006), and associated investment of parental effort in altricial offspring over mating effort productive of larger broods of precocial offspring. Among humans specifically, developing brains vie only with the liver for per unit weight metabolic consumption, and are associated with necessarily protracted gestation and postnatal development (Parker, 1990). As humans exemplify, large brains with invariant patterns of growth and differentiation render mammals highly susceptible to malnutrition (Isler & van Schaik et al., 2006; Deaner, Barton, & van Schaik, 2003).

Thus, just as parasite stress theory has been proposed to partially, if not completely, subsume cold winters theory, life history theory has likewise been proposed to partially, again if not completely, subsume both parasite stress theory and cold winters theory within a more inclusive framework. In the present study, we further proposed that the use of zoogeographic regions as macroecological categories provides an overarching theoretical scaffolding that readily accommodates all three of these single-factor theories and the hypothesized relations among them.

The Present Study: Zoogeographic Regions

Much supporting empirical evidence has been presented in favor of several of these single-factor theories. Based on the principles of strong inference (Platt, 1964), one is therefore tempted to put these alternative hypotheses in competition with each other, as has been done previously (e.g., Eppig, Fincher, & Thornhill, 2010). The results of the social biogeographical analyses cited above, however, would suggest that these ecological parameters are best treated as an interrelated system, rather than as causal factors operating in mutual isolation. This is because most of these ecological parameters manifestly exert causal influences on each other and are thus not completely separable, excepting by statistical legerdemain. For example, cold winters kill parasites.

The core problem in the science of ecology is attempting to account for all the relevant variables, which one cannot ever be sure of having completely catalogued. Fortunately, a different approach can be derived from the field of biogeography and that is the concept of zoogeographical regions, as famously advanced by the nineteenth century naturalist Alfred Russel Wallace, the co-discoverer of the theory of evolution by natural selection. Wallace’s (1876) zoogeographical regions were not intended to capture any single ecological parameter, but to instead capitalize on the complex of covariation among them to carve out the larger ecological systems that were manifested by the correlated distributions of flora and fauna across diverse environments in different parts of the world.

More recently, Holt and colleagues (2013) updated Wallace’s zoogeographic classification by analyzing the distributions of 21,037 species of amphibians, birds, and mammals. This revised classification differed in some important details but was otherwise surprisingly similar in broad outlines to the original nineteenth century version. The new classification included 11 zoogeographic realms, subdivided into 20 smaller regions: the Palearctic, the Sino-Japanese, the Saharo-Arabian, the Afro-Tropical, the Madagascan, the Oriental, the Oceanian, the Australian, the Nearctic, the Panamanian, and the Neo-Tropical realms.

Most recently, this approach has been applied to explaining the biogeography of human diversity in life history strategies (Figueredo, Hertler, & Peñaherrera-Aguirre, 2020). Zoogeographic regions were used to delimit ecologically informed geographical areas within which different life history adaptations could be expected to evolve and develop. First, these results indicated that the zoogeographic regions did a reasonably good job of summarizing the continuous ecological parameters of each biologically defined geographical area, including average annual temperature (R2 = 0.87), average annual precipitation (R2 = 0.39), human parasite burden (R2 = 0.69), and population densities (R2 = 0.21). Second, the results indicated that the zoogeographic regions accounted for the preponderance of the variance (R2 = 0.79) in human life history speeds across the sampled national polities, as measured by a unit-weighted composite of teen pregnancy rate, birth rate, infant mortality, and life expectancy. Even when statistically controlling for the combined effects of the four sampled ecological parameters upon human life history speeds (R2 = 0.39), the zoogeographical regions still accounted for a greater residual proportion of the variance (R2 = 0.41) than all of the other “single-factor” explanations combined. When statistically controlling for the effects of the zoogeographical regions, on the other hand, the combined residual effects of the four sampled ecological parameters were statistically non-significant. This indicated substantial incremental validity (see Bridgeman, McCamley-Jenkins, & Ervin, 2000; Sackett & Lievens, 2008) for the zoogeographical regions with respect to the individual ecological parameters, suggesting that the former carry more contextual, systemic information regarding the local ecologies than the individual analytical dimensions.

In the present study, we apply the explanatory principle of zoogeographical regions to attempt to explain the diversity in cognitive abilities among a sample of extant human populations. We otherwise apply the same methods of measurement and analytical techniques used in the recent Figueredo, Hertler, and Peñaherrera-Aguirre (2020) study on the biogeography of human diversity in life history strategies.



National Polities. This study gathered information on ninety-eight national polities currently found in Africa, Asia, and Europe. It is worth noting that due to the considerable colonization by Eurasian and African human and non-human biota since the fifteenth century AD, neither Australasian nor American polities were included in the present analyses. The current evidence suggests the ecological invasions associated with the Age of Exploration considerably altered the local environments of these regions (Crosby, 1993, 2004; Herter et al., 2018). Furthermore, while some records indicate that Old-World populations sought ecological niches akin to their native environments (Crosby, 2004), subsequent intercontinental migration waves also encountered ecologies dissimilar to Eurasia, often generating a mismatch between the population and the locality. In contrast to the ecological history of the Americas and Australasia, several molecular genetic studies revealed that Asian, African, and European populations have remained genetically similar across their evolutionary history (Basu et al., 2003; Hanihara, 1991; Leslie et al., 2015; Martínez-Cruz et al., 2012; Piazza, Cappello, Olivetti, & Rendine, 1988; Qi et al., 2013). The relative temporal consistency of these populations enables the examination of ecological pressures leading to regional variations in human cognitive ability. Consequently, the present study was limited to national polities featuring the same aboriginal populations that inhabited these regions before the fifteenth century AD.


Physical Ecology. We collected data on the mean annual precipitation and mean annual temperatures from the Climate Charts (2015) data repository. We also gathered information from the Canty Media (2015) website on the mean annual temperature for polities with missing data.

Parasite Burden. Given the reported association between physical ecology, parasite burden, and life history speed (Figueredo et al., 2017a, b, c; Peñaherrera-Aguirre et al., 2019), it was relevant to explore the influence of parasites on regional variation in cognitive ability. For the current purposes, we gathered information on the per capita disability-adjusted age years (DALYs) associated with infectious diseases from the World Health Organization (2015) online database. We also collected data from Murray and Schaller’s (2010) compendium of the historical prevalence of infectious diseases. A log-transformed unit-weighted composite was estimated using these indicators.

Slow Life History Strategy. This study also computed a unit-weighted composite for the estimation of slow life history. This dimension was comprised of four biodemographic indicators: (a) birth rate (Central Intelligence Agency, 2015), a measure of fertility and a proxy of the underlying trade-off between mating and parental effort (MacArthur & Wilson, 1967); (b) adolescent fertility (World Bank, 2019), computed as the number of births per 100,000 women between ages 15 to 19, a metric indicating the degree of early investment in mating and reproductive effort and consequently tied directly to life history dynamics (Stearns, 1992); (c) life expectancy (Central Intelligence Agency, 2015), a metric reflecting the impact of local morbidity and mortality and to some extent the allocation of resources to somatic effort (Charnov, 1991; Figueredo, Vásquez, Brumbach, & Schneider, 2004; Figueredo, Vásquez, Brumbach, & Schneider, 2007; Isler & van Schaik, 2006, 2009; Williams, 1957); and (d) infant mortality (Central Intelligence Agency, 2015), a metric of age-specific environmental harshness (Ellis, Figueredo, Brumbach, & Schlomer, 2009).

National IQ. For present purposes, we collected National IQ (NIQ) data from two primary sources: Becker (2019) and Wicherts, Dolan, and van der Mass (2010). Concerning Lynn and Becker’s NIQ database, this compendium contains information on over 200 contemporary nation-states with more than 130 polities featuring National IQ scores. The database contains both unweighted and weighted averages. The authors calculated the unweighted and unadjusted NIQ(UW) scores with Eq. 1:


IQ (cor.) reflects the raw IQ scores computed post-conversion or extrapolation. Alternatively, NIQ (NW) is equivalent to the means computed from all samples adjusted by their sample sizes. Eq. 2 captures these computations:

$$\left[NIQ(NW)\right]=\left(\frac{\sum_{i=1}^{n}{\left[IQ(cor.)\right]}_{i }* {\left[N(ind.)\right]}_{i }}{\sum_{i=1}^{n}{\left[N(ind.)\right]}_{i }}\right)$$

Lynn and Becker (2019) also calculated National IQ scores, adjusting for both sample size and data quality, NIQ(QNW). First, they computed an intermediate quantity called the QN-Factor, wherein the sample rating is an index of data quality based on the sample’s properties, such as its representativeness, the presence of ethnic and cultural variation within the polity, and age differences, among others. The testing rating is the overall index of data quality due to the partial use of the IQ measure during the test administration, the influence of time deviations, and whether or not the test standardization followed Great Britain (GBR) norms. The method rating controls for errors attributable to raw score corrections (for further details regarding the estimation of these weights, see Lynn and Becker, 2019). Lynn and Becker (2019) calculated QN-Factor with Eq. 3:

$${\left[QN-Factor)\right]}_{ }=\left[N\left(ind.\right)\right]*{(M}_{\left(\left[Sample Rating\right)\right];\left[Testing Rating\right]:[Method Rating])})$$

Then, the authors computed these values of NIQ(QNW) using the Eq. 4:

$$\left[NIQ(QNW)\right]=\left(\frac{\sum_{i=1}^{n}{\left[IQ(cor.)\right]}_{i }* {\left[QN-Factor)\right]}_{i }}{\sum_{i=1}^{n}{\left[QN-Factor)\right]}_{i }}\right)$$

After the publication of The Intelligence of Nations (Lynn & Becker, 2019), Becker released a corrigendum and commentary to the National IQ v1.3.2 and v1.3.3 datasets. The paper addressed the estimation of low NIQ scores (below 60) of nine countries, a significant point of empirical contention. Critics view these results as considerably questionable, given that Physical and Mental Health Statistical Manuals, such as the ICD-10, classify IQ scores ranging from 35 to 70, evidencing moderate to mild mental retardation. Becker, however, clarifies that low national IQ scores are plausible given the correspondence between cognitive development and phylogenetic cultural patterns.

It is worth noting the debate regarding low IQ scores in polities remains unabated. Previously, Wicherts, Dolan, and van der Mass (2010) questioned the low computation of IQ scores by Lynn and Vanhanen (2012), values that generally exceeded those computed for the NIQ database. Wicherts and colleagues (2010) estimated an average IQ of 82 for African countries. For these authors, Lynn and Vanhanen’s underestimation originated from: (1) the use of inaccurate IQ norms or age-based standards, (2) issues due to test adaptation, (3) pragmatic difficulties associated with test administration, (4) biased computations disfavoring African samples, and (5) researchers gathered IQ data from clinical or otherwise unhealthy samples. According to Wicherts and colleagues, their systematic review adequately circumvented these difficulties, allowing them to conclude an average IQ of 82 was not the product of either publication or sampling biases.

Lynn and Meisenberg (2010) questioned the corrections suggested by Wicherts and colleagues (2010), contending that the authors’ systematic review relied on unrepresentative elite samples, overestimating the average IQ scores for African countries. Wicherts, Dolan, Carlson, and van der Maas (2010b, a) recomputed the African IQ scores after removing these elite samples. According to the authors, these adjustments still failed to replicate Lynn and Vanhanen’s estimates (2012). Consequently, due to the current disagreement regarding the optimal estimates for the African polities, we generated a new variable updating NIQ(UW) with Wicherts and colleagues’ corrections, NIQ(UWC).

Becker validated the NIQ computations by correlating the polity’s IQ scores with the performance of various international school assessment studies (SAS; transforming these values into Greenwich IQ scores). Overall, the analyses identified a sizeable and positive correlation between the two IQ estimates. A critic could argue that the observed differences between NIQ and international school assessments are due to the variation in health, nutrition, and education. Becker, however, viewed this claim as unwarranted, given that ecological factors also influence the performance on cognitive tests such as Raven’s progressive matrices. Additional studies are needed to determine the degree to which environmental factors influence IQ and SAS scores proportionally. The author explained that the reported differences between Lynn and Vahannen (2012) and the NIQ database, is attributable to the use of conversion formulas transforming raw IQ scores, as in the case of Raven’s progressive matrices. The author also acknowledged that whereas Lynn and Vanhanen’s calculations relied on British norm tables, the NIQ compendium uses trendline extrapolation based on the association between raw and IQ scores. Besides these caveats, the fact that all three NIQ measures (UW, NW, and QNW) featured positive and statistically significant correlations above 0.80 (Lynn & Becker, 2019), suggests, the cross-regional rank orders for the various national-polities remain relatively unaltered after these adjustments.

In separate indices, Lynn and Becker had also used geographic proximity to impute missing values on NIQ scores for certain unmeasured national polities, but we eschewed this imputation method in the present analyses as it would inevitably bias them in favor of our zoogeographical hypotheses. We did this not to dispute the validity of these imputations, but merely to provide a fairer test of our own hypotheses and provide more conservative estimates.

In the present sample, the bivariate correlations among all the variant indices of National IQ, NIQ(UW), NIQ(UWC), NIQ(NW), and NIQ(QNW), were extremely high and ranged from 0.92 to 0.99 (p < 0.0001). Nevertheless, we decided to concentrate our analyses upon NIQ(QNW), as it appeared to be the most comprehensively adjusted for both sampling error and measurement biases. The present study also conducted a convergent validity test between NIQ(QNW) and the National Harmonized Test Scores (HTS) obtained from the World Bank online database (Patrinos & Angrist, 2018). The HTS are computed from international student achievement assessment programs (such as PISA, PISA + TIMS/PIRLS, SACMEQ, PASEC, LLECE, PILNA, EGRA, and EGRANR). HTS scores can range from 300 to 625 TIMS units. As expected, we found a sizeable and positive correlation coefficient (r = 0.82, p < 0.00001), supporting the convergent validity of the NIQ(QNW) estimates with this independent HTS measure of national-level aggregate cognitive ability. As explained above, this restricted sample did not include Australian, North American, Central American, and South American national polities to circumvent any issues associated with large-scale population migrations after AD 1492.

Statistical Analyses

All statistical analyses were performed using SAS 9.4 and UniMult 2. The unit-weighted factor structures (see Gorsuch, 1983) were estimated using SAS PROC CORR and the analyses of variance were done using SAS PROC GLM (see Cohen & Cohen, 1983) and UM2. Multi-level models (MLMs) were performed using SAS PROC MIXED, specifying single-lagged and heterogeneous spatially autoregressive effects, ARH(1).


Descriptive Statistics

Table 1 displays the sample means, standard deviations, and bivariate (zero-order) correlations among all the continuous variables used in this study. Note that the bivariate correlations among the four indicators of slow life history strategy are very highly correlated and the psychometric implications of this will be presented in the section that follows.

Table 1 Bivariate correlations among slow life history strategy indicators and predictors

The Measurement Model

Infant mortality and life expectancy were statistically adjusted for the logarithmic effects of human parasite burden and exporting the regression residuals. This procedure was performed to control statistically for any geographic variation in morbidity and mortality as a possible result of historically recent advances in medical technology and public health interventions. The concern was that such technologies and interventions were not present in the ancestral environments and to that human populations have had insufficient time to adapt genetically to them by means of natural selection, as per the rationale provided in Figueredo, Hertler, and Peñaherrera-Aguirre (2020).

Table 2 displays the unit-weighted factor loadings (operationalized as part-whole correlations) of each of the indicators from its latent common factor, slow life history. These factor loadings serve as convergent validity coefficients among the four life history traits. The factor structure table shows the results for both the adjusted and non-adjusted forms of the indicators, indicating that the adjustment had negligible effects upon the magnitudes of the coefficients. Nevertheless, we decided to be more conservative by using the LPB-adjusted data in all subsequent analyses, given that the factor loadings were still quite acceptable.

Table 2 Slow life history (SLH) factor unit-weighted loadings: not adjusted for natural logarithm of parasite burden (LPB) in column 1; adjusted for natural logarithm of parasite burden (LPB) in column 2; bivariate correlation between unit-weighted factor scores for LPB-adjusted SLH and non-LPB-adjusted SLH is 0.98*

The Structural Models

The five basic parameters of the physical and community ecology used to examine the predictive validity of the zoogeographic regions were as follows: (1) average annual temperature; (2) average annual precipitation; (3) the natural logarithm of human parasite burden; (4) human population density; and (5) human slow life history strategy. This list of ecological parameters is by no means exhaustive, but it can be used to operationalize several of the principal selective pressures that have been proposed as “single-factor theories” to explain the diversity in cognitive abilities among human populations.

The assigned zoogeographic regions accounted for the following proportions of the variance across national polities: 78.5% (F5,92 = 67.04, p < 0.0001) for average annual temperature; 54.9% (F5,97 = 22.42, p < 0.0001) for average annual precipitation; 68.4% (F5,92 = 39.85, p < 0.0001) for the natural logarithm of human parasite burden; 9.8% (F5,92 = 2.01, p = 0.0851) for the population density; and 63.4% (F5,92 = 31.83, p < 0.0001) for our unit-weighted factor scale of slow life history strategy. These “manipulation checks” indicated that the zoogeographic regions performed reasonably well in partitioning the variance among national polities in four out of five basic dimensions of physical and community ecology but did not adequately predict that of human population density.

For our criterion variable of principal interest, National IQ, as weighted by both sample size and data quality, or NIQ(QNW), the proportion of the variance across national polities accounted for by zoogeographic regions was 71.4% (F5,92 = 45.88, p < 0.0001). Comparing the alternative indices of National IQ, the squared multiple correlations for Zoogeographic Region were quite similar: 76.6% (F5,92 = 60.12.88, p < 0.0001) for NIQ(UW), the unweighted mean NIQ; 63.9% (F5,92 = 32.54, p < 0.0001) for NIQ(UWC), the unweighted mean NIQ corrected for updated African values); and 72.0% (F5,92 = 47.37, p < 0.0001) for NIQ(NW), the mean NIQ weighted by sample size. Figure 1 displays the distribution of these zoogeographic region means for National IQ graphically, using NIQ(QNW), which is weighted by both sample size and data quality.

Fig. 1
figure 1

Distribution of National IQ, NIQ(QNW), across zoogeographical regions: (AT) Afro-Tropical, (MA) Madagascan, (OC) Oceanian, (OR) Oriental, (PA) Palearctic, (SA) Saharo-Arabian, and (SJ) Sino-Japanese

To parse the statistically significant term for zoogeographic regions, we constructed a system of dummy variables (Cohen & Cohen, 1983) to break down the multilevel nominal variable into five separate tests, each with only a single degree of freedom. We used the Afro-Tropical (AT) region as the reference group, as it was the most ancestral for the human species, and created one dummy (binary) variable for each of the others, respectively. All five of these were found positive and statistically significant (a = 68.85, se = 1.39, t92 = 49.55, p < 0.0001; MA: b = 17.97, se = 6.95, t92 = 2.59, p = 0.0113; OR: b = 15.02, se = 2.34, t92 = 6.41, p < 0.0001; PA: b = 25.00, se = 1.74, t92 = 14.35, p < 0.0001; SA: b = 12.62, se = 2.20, t92 = 5.74, p < 0.0001; SJ: b = 37.13, se = 5.01, t92 = 7.41, p < 0.0001), probably reflecting evolutionary responses to the challenges of migration to non-ancestral physical and community ecologies.

To determine the relative incremental validities of the five sampled ecological parameters with respect to zoogeographical regions, we also performed hierarchical general linear models. When entered hierarchically before the five ecological parameters (SLH, Density, LPB, PrecipAvg, and TempAvg), zoogeographic region accounts for 70.6% (F5,87 = 53.77, p < 0.0001) of the variance, with the remaining variables accounting for the remaining 5.8% (F5,87 = 4.16, p = 0.002); this first model indicates the degree to which the zoogeographic regions concisely summarize the more specific information contained in the continuous ecological parameters, thus supporting the generative biogeographic theory. Conversely, when hierarchically entered after the five continuous ecological parameters (SLH, Density, LPB, PrecipAvg, and TempAvg), zoogeographic region accounts for only 5.8% (F5,87 = 4.49, p = 0.001) of the variance, with the remaining variables accounting for the remaining 70.6% (F5,87 = 53.44, p < 0.0001); this second and inverse hierarchical ordering of predictors makes greater sense from a sequential cascade analysis perspective, consistent with the previous work on social biogeography as well as with the summary of results provided in the “Discussion” section of the present paper (also see Appendix 1 for details). The quantitative relation is thus quite symmetrical and also reflects the recent findings of Figueredo, Hertler, and Peñaherrera-Aguirre (2020) with respect to the biogeography of human diversity in life history strategies. The results of the second of these hierarchical general linear models are displayed in Table 3 in greater detail.

Table 3 Hierarchical general linear model for National IQ, NIQ(QNW), weighted for both sample size and quality, as predicted by human slow life history strategy (SLH), human population density (Density), the natural logarithm of human parasite burden (LPB), average annual precipitation (PrecipAvg), and average annual temperature (TempAvg)

We note that the results of this model support at least three of the “single-factor” theories reviewed: (1) cold winters theory, in that average annual temperatures have a statistically significant and negative effect on National IQ; (2) parasite stress theory, in that local parasite prevalences have a statistically significant and negative effect on National IQ; and (3) life history theory, in that slower life history speeds have a statistically significant and positive effect on National IQ. Nevertheless, we also note that each of these hypothesized causal influences adds incrementally to our prediction, even after hierarchical partitioning of variance, and that zoogeographic regions adds a further incremental boost to our predictive power. This once again demonstrates the incremental validity of the zoogeographical regions model over the predictions of the previous “single-factor” theories (for discussions and applications of the concept of incremental validity, see Bridgeman, McCamley-Jenkins, & Ervin, 2000; Sackett & Lievens, 2008).

Finally, to test for any possible effects of any serially autoregressive effects among spatially contiguous data, we also ran a MLM on the National IQs, NIQ(QNW). Using a simple sequential variable representing the ordinal distance “Out-of-Africa” (OOA) for each national polity, we estimated the single-lagged heterogeneous autoregressive serial dependencies, ARH(1), of each successive spatially adjacent national polity within that ordered sequence, and found these to be statistically equal to zero. Nevertheless, the effect of the OOA variable itself was found to be statistically significant and positive (a = 76.25, se = 2.32; b = 0.17, se = 0.04, t96 = 4.13, p < 0.0001), indicating a small but gradual tendency towards increase in IQ with out-migration. This procedure models the possible phylogenetic effects of common human origins and subsequent outward migrations (see Hertler, Figueredo, Peñaherrera Aguirre, Fernandes, & Woodley of Menie, 2018), thus circumventing this potential threat to the validity of correlational analysis. The MLM residuals were then exported for the National IQ data, NIQ(QNW), and used for subsequent general linear modeling. The zoogeographic regions accounted for 63.6% (F5,92 = 32.15, p < 0.0001) of the variance across national polities in the studentized residuals, or OOA-adjusted National IQs, NIQ(QNW). These results suggest that the zoogeographic variations in IQ were not simply the effects of out-migration from Africa, but in subsequent adaptations to the ambient ecological conditions that were encountered locally in the novel environments.


Summary of Results

We have determined that zoogeographic regions explain the preponderance of the variance in National IQ across the sampled polities. Furthermore, we have determined that zoogeographic regions, as macroecological categories, can be used to succinctly summarize much of the cross-national variance in several key ecological parameters: average annual temperature, average annual precipitation, human parasite burden, and human slow life history strategy. Some of these continuous ecological parameters had been used previously in various single-factor theories for the evolution of cognitive ability. We have reviewed claims that had been made by each of these successive theories that each of their preferred “single-factor” explanations entirely subsumed the others, and that this preferred predictor completely mediated the alternative predictors, the final one being the only direct evolutionary cause of National IQ. We have seen that this is not the case, in that none of the major “single-factors” that have been proposed completely mediates all the others. This can be illustrated by examining the following exploratory path model: Fig. 2 graphically represents the results of a sequential canonical cascade analysis (see Appendix 1 for statistical details) including only the relations among average annual temperature, human parasite burden, and human slow life history strategy, to represent the three major “single-factor” theories.

Fig. 2
figure 2

Exploratory path model summarizing the direct and indirect relations among the preferred predictors of the “single-factor” theories reviewed

It is evident from this path diagram that there are statistically significant residual direct effects from each of the prior constructs to each of the successive ones within the sequence. This finding indicates that each of the successive constructs only partially mediates all the previous ones. To be fair, however, we do note that the path coefficients are substantially greater in magnitude along the main sequence hypothesized, as represented by the series of three diagonal arrows in the diagram. The purpose of the path diagram was therefore not to deny or even minimize the high degree of mediation, even if only partial, but instead to show the interconnectedness of the entire system of relationships. We believe that the interconnected of this system supports our proposed use of zoogeographic regions as macroecological categories that summarize these covarying environmental parameters.

Theoretical Considerations

The essence of human intelligence is more or less equivalent to what in nonhuman animals we variously call developmental, behavioral, or phenotypic plasticity, so a brief discussion of the evolutionary dynamics of this phenomenon is warranted (for a more detailed discussion, see Figueredo, Hammond, & McKiernan, 2006). Invariant instincts and routinized behavioral repertoires are believed to be bioenergetically cheaper alternatives to the phenotypic plasticity afforded by human cognitive abilities, for reasons articulated in the expensive brain hypothesis. Phenotypic plasticity allows one genotype to adopt a plurality of phenotypic responses (Davies, Krebs, & West, 2012). Complete canalization of traits are avoided by rudimentary forms of phenotypic plasticity, whether developmentally expressed as temperature-dependent sex-allocation, morphologically expressed among certain beetles and salmon, or seasonally expressed as cyclical responses to photoperiod (Brakefield & Zwaan, 2011). When the average fitness of individuals with plastic strategies exceeds that of individuals with fixed strategies, more elaborate manifestations of phenotypic plasticity can evolve (Czesak, Fox, & Wolf, 2006).

Theoretically, developmental plasticity and behavioral flexibility have been posited to evolve under conditions of increased environmental variability (Figueredo, Hammond, & McKiernan, 2006). More specifically, human general intelligence has been posited to be selected in response to environmental novelty that confronted individuals with evolutionarily unfamiliar problems to solve (Kanazawa, 2012). Based on that evolved plasticity and flexibility, phenotypic accommodation to environmental changes has been posited to be a major driving force in evolution by leading to subsequent genetic accommodation, as shaped by the action of selection on natural variation in the adaptive consequences of the gene-environment interactions triggered (West-Eberhard, 2003). As recently explained by Richerson and Boyd (2020):

Cultural evolutionists have proposed that culture in humans is an evolutionarily active system that can even act as a selective force on genes. Defenders of the Modern Synthesis can be quite intemperate in their rejection of such heresy [46, 51]. However, there are good examples of culture-led gene-culture coevolution in the case of humans [52–54]. For example, Richerson & Boyd’s [55] tribal social instincts hypothesis conjectures that group selection on cultural variation operating via social selection (selective rewards and punishments, [56]) acting on genes shaped our innate social psychology, making us more docile for example. If so, cultural evolution is playing an ultimate role alongside genes in human evolution. Likewise, we think that the human life history coevolved with culture, often driven by cultural innovations [57]. (p. 3).

We therefore believe it reasonable to conclude that the great human migrations out of Africa into entirely new habitats and ambient ecologies must have provided the hypothesized selective pressures for enhanced cognitive abilities (for a discussion of the role of temporal and spatial variability in human life history and cognitive adaptations during the Pleistocene, see Richerson & Boyd, 2020).

Human cognitive abilities may well represent the most baroque expression of phenotypic plasticity among animals. Accordingly, like any form of phenotypic plasticity, the evolution of cognitive ability is subject to a cost–benefit balance sheet, as recognized by single-factor theories as well as by the present zoogeographic reconceptualization of those ecological determinants. The zoogeographic regions used to explain geographic variation in intelligence amount to a macroecological level of organization, subsuming the physical ecological factors emphasized in cold winters theory and the community ecological factors emphasized in parasite stress theory, to the end of more fully delineating both sides of the cost–benefit ledger. With pull and push factors juxtaposed, disparities in cross-regional levels of cognitive ability can be seen to arise from the simultaneous release from the brake of parasite stress and the spur of cold winters. Furthermore, the zoogeographic regions approach is broader in scope than the life history theory of human cognitive ability in that it encompasses the entire ecological system of relationships rather than the set of bioenergetic tradeoffs characteristic of any single species within it.

Of course, neither the single-factor nor the macroecological theories are exhaustive of theories regarding the evolution of human intelligence. Notable among alternatives are the expanded ecology hypothesis (e.g., González-Forero & Gardner, 2018) and the social brain hypothesis (Dunbar, 1998). Of course, neither the single-factor nor the macroecological theories are exhaustive of theories regarding the evolution of human intelligence. Notable among alternatives are the expanded ecology hypothesis (e.g., Powell, Isler, & Barton, 2017; González-Forero & Gardner, 2018) and the social brain hypothesis (Alexander, 1990; Dunbar, 1998). However, these theories both exclusively address the differences in brain development among different species of nonhuman primate as well as between humans and all other species of primate. These theories therefore relate only to between-species differences. Neither theory, as far as we are aware, makes any predictions regarding how humans evolving within different zoogeographical regions might have adapted to different resource allocations to brain development (and hence intelligence). Our theory therefore relates instead to within-species differences. One could conceivably try to extend these other theories (both the expanded ecology hypothesis and the social brain hypothesis) to within-species differences, but it would be a stretch of which we are not sure the original proponents of the theories would necessarily approve.

Limitations of the Study

In view of the fact that this study reviews the zoogeographical distribution of human intelligence, it shares those limitations affecting intelligence assessment generally and its cross-cultural application specifically. First, certain limitations relate to the imperfect relationship between measured (observed score) and true (latent score) intelligence as discussed in the psychometric literature. For example, discrepancies between true and measured cognitive ability can arise from extraneous factors, such as examiner error (Franklin et al., 1982) or restricted examinee motivation (Duckworth et al., 2011). The cross-cultural comparisons of IQ testing data, prerequisite for our analysis of geographical diversity in cognitive abilities, impose additional limitations. For example, cross-cultural comparisons naturally augment any concerns about cultural bias commonly discussed in relation to assessment within ethnically diverse societies.

The existence of the Flynn effect is also of relevance in that the IQ gains it describes are thought to be attributable to changes in culturally relevant variables such as the prevailing educational ethos, which has increasingly emphasized abstract problem solving (fluid intelligence) over and above memorization and knowledge acquisition (crystallized intelligence) in many parts of the developed world. Such limitations are more or less valid depending on the interpretation of results. To the extent that populations are reasonably well-sampled, one can have confidence in the scores, and then doubt only extends to the meaning of those scores. Differences in realized scores across populations from well administered intelligence tests can derive from any combination of environmental or genetic differences. Thus, our use of the cross-national data to measure broader biogeographical differences encounters potential limitations with respect to cultural bias, the Flynn effect, and related nature/nurture issues. Each of these could limit the validity of our interpretations, and the data on which they are founded, but none of these are unique to our application; these are issues which have long been subjected to theory and research within the intelligence testing literature. Doing full justice to these issues is beyond the scope of the present paper due to limitations of journal space, but the interested reader can be referred to an excellent review by Nisbett and colleagues (2012).

Nevertheless, we have seen that the various alternative corrections, adjustments, and weightings used in the Becker (2019) data, based on a variety of different assumptions, have had little impact upon our results. The alternative NIQ indices presented are all very highly correlated with each other (ranging from 0.92 to 0.99, p < 0.0001), and all are well-predicted to a very similar extent (ranging from 63.9 to 76.6% of the variance) by our zoogeographic regions, regardless of which alternative index is used in the analysis. Furthermore, these national averages are based upon a relatively diverse array of different methods of measuring intelligence, which suggests that it is unlikely for any homogeneous biases to pervade the entire data set. We therefore believe that at our present state of knowledge, our results are sufficiently robust for a reasonable degree of confidence in the conclusions.

We also need to clarify that none of our analyses were on g, per se, but instead on aggregate IQ, which includes specialized abilities as well as generalized ones. We believe that using the term “cognitive ability” in this way, if not represented as general cognitive ability (g) is a correct use of this term (see Woodley & Figueredo, 2013; Woodley of Menie et al., 2017). Thus, any possible cross-cultural variability in the relative strengths of the g-loadings of the various IQ tests employed should not compromise our results.

One possible concern in the interpretation of our results is that the different levels of socioeconomic development among different national polities might be influencing National IQ levels over and above the influences of physical, community, and social ecology that we have reported. Indeed, previous work from our laboratory (e.g., Figueredo, Cabeza de Baca, Fernandes, Black, Peñaherrera-Aguirre, Hertler, S.C., et al., 2017) has explicitly modeled National IQ levels as a partial consequence of macroeconomic diversification, gross domestic product, and aggregate human capital. We therefore agree that these are important considerations.

Nevertheless, similar analyses have also been published (Peñaherrera-Aguirre, Hertler, Figueredo, Fernandes, Cabeza de Baca, & Matheson, 2018), using different criterion variables (i.e., National homicide rates instead of National IQ). Peñaherrera-Aguirre and colleagues conducted two Sequential Canonical Analyses (SEQCAs): the first included NIQ as a criterion of economic indicators (e.g., GDP per capita, the Gini coefficient, and the percentage of young males in the labor force), whereas the second SEQCA included NIQ as a predictor of the latter economic indicators. Model comparison (based on the number of nonsignificant path coefficients) concluded that the second model was a more parsimonious path model, which puts NIQ as causally prior to the economic indicators.

A related possible concern in interpreting the results of our analyses is that the noise generated by recent changes in health, nutrition, education, and wealth might have obscured or overwritten the hypothesized evolutionary signal of ancestral ecology. Nevertheless, the relation between current environmental forces and changes in NIQ does not preclude ultimate and evolutionary factors from predicting the reported regional variation in cognitive ability. Rindermann, Woodley of Menie, and Stratford (2012), for example, analyzed haplogroup frequency data for 14 Y chromosomal genes known to be associated with cognitive ability. Controlling for contemporary environmental indicators, such as the Human Development Index (HDI), the authors determined that in addition to socioeconomic development, haplogroup frequency had a unique contribution to the statistical model. In a subsequent study, Becker and Rindermann (2016) estimated genetic distances (also based on Y chromosomal haplogroup frequencies) and explored their relation with cross-national IQ differences (pairwise calculations) in a sample of 101 nations. The authors computed a series of path analyses controlling for the influence of the HDI and the countries' geographical location. The genetic distance vector had a significant and positive association with the cross-national differences in IQ. Employing an alternative methodology, Woodley of Menie and colleagues (2016) extracted a unit-weighted genetic (metagene) factor comprised of SNPs associated with IQ. The study identified a positive and strong association between the metagene factor and NIQ. All of these studies indicate that genetics continue to exert a substantial phenotypic influence in spite of the overlay of environmental effects.

Another similar concern regarding our use of ancestral ecological conditions for modern phenotypic outcomes (such as IQ) is that there have been large scale changes over the last 200 years or so in fertility rates and child mortality, probably as a result of global improvements in public health and nutrition. For example, Appendix 2 includes the details of analyses that show that regional differences in fertility rates and child mortality rates are conserved even in the context of a significant main effect of time, indicating a secular decrease across all regions (see Eastwood & Lipton, 2011; Murtin, 2013; Reher, 2004).

Thus, in spite of these many acknowledged limitations, we believe that the present results serve to advance this field of study by at least another step by broadening the discourse to encompass macroecological perspectives and not limit ourselves to single-variable explanations.