The Centrality of Ancestral Grandmothering in Human Evolution

Abstract

When Fisher, Williams, and Hamilton laid the foundations of evolutionary life history theory, they recognized elements of what became a grandmother hypothesis to explain the evolution of human postmenopausal longevity. Subsequent study of modern hunter-gatherers, great apes, and the wider mammalian radiation has revealed strong regularities in development and behavior that show additional unexpected consequences that ancestral grandmothering likely had on human evolution, challenging the hypothesis that ancestral males propelled the evolution of our radiation by hunting to provision mates and offspring. Ancestral grandmothering has become a serious contender to explain not only the large fraction of post-fertile years women live and children’s prolonged maturation yet early weaning; it also promises to help account for the pair bonding that distinguishes humans from our closest living evolutionary cousins, the great apes (and most other mammals), the evolution of our big human brains, and our distinctive preoccupation with reputations, shared intentionality and persistent cultural learning that begins in infancy.

Introduction

Since the 19th century, a favored explanation for the evolution of late maturation, big brains, long lifespans, and pair bonding that characterize humans points to hunting by ancestral males to provision their mates and offspring. As elaborated in the last several decades, the hunting hypothesis proposes that spreading savannas in ancient Africa made hunting a promising way to make a living. Since dependent offspring could interfere with prey pursuit, ancestral mothers did better to pair with a hunting mate; paternal provisioning thus made nuclear families elementary economic and social units with a sexual division of labor (e.g., Washburn and Lancaster 1968; Lancaster and Lancaster 1983; Lancaster et al. 2000; Kaplan et al. 2000; Kaplan and Robson 2002; Kaplan and Gurven 2005). The stone tools and bones of multiple large ungulates that compose the earliest archaeological sites seemed hard evidence that ancestral hunters brought theirkills home to mates and offspring (e.g., Isaac 1978).

Re-examination of the ecological context and assemblage composition of the early sites finds they more likely represent near-kill/ambush/scavenging locations than ancient home bases (O’Connell et al. 1988a, 1988b, 1999, 2002; Hawkes 2016). Quantitative behavioral observations of living people who forage for a living find high daily failure rates for big animal hunting combined with wide distribution of the occasional bonanzas. Unpredictability and collective consumption make successes notable. But those features also make big carcass acquisition unsuited to meeting anyone’s daily consumption needs—let alone those of dependent children (Hawkes 1990, 1991, 1993a; Hawkes et al. 1991, 2001a, 2001b, 2014, 2018; Marlowe 2007, 2010). The salience of hunting reputations (e.g., Bliege Bird et al. 2001; Smith 2004) and the rich ethnographic evidence that men’s alliances dominate community affairs (e.g., Smuts 1992; Rodseth 2012) make male status competition a better candidate than family provisioning to explain why men hunt (Hawkes 1990, 1991, 1992, 1993a,b; Hawkes et al. 1991, 2001a, 2001b, 2014, 2018; Hawkes and Bliege Bird 2002). Male-biased mating sex ratios that accompanied the evolution of our postmenopausal lifespans (Coxworth et al. 2015) point to the probable importance of mate guarding rather than household provisioning in the evolution of our pair bonding habits. Mathematical modeling implicates sexual selection and sexual conflict in the evolution of our exceptional longevity (Chan et al. 2016, 2017).

My aim here is to explain the basis for a grandmother hypothesis about human evolution and then highlight insights about grandmothering mentioned by the architects of evolutionary life history theory as they developed the concepts of reproductive value, inclusive fitness, adaptive sex ratios, and sexual conflict. They ignored their grandmothering insights in pursuit of other questions, but subsequent evidence about primate phylogeny, population age structures, foraging strategies, and social behavior as well as mammalian brain development and social cognition now link those insights to the key role of ancestral grandmothers in human evolution. By highlighting inferences that the founders recorded but did not exploit themselves, I both applaud their prescience and hope to enlist others to help take more advantage of them.

A grandmother hypothesis for the evolution of human life history

It is widely agreed that the evolution of human life history began in the context of ancient climate changes that had started in the Miocene, constricting African forests and spreading savanna habitats. In forests, ancestral populations likely relied on fruits and leaves as great apes continue to do today. Ancestral nursing infants, carried by mother, likely began picking and eating the same foods that she was eating just as ape infants do today (Fletcher 2001; van Noordwijk et al. 2009, 2013; Bădescu et al. 2017; Bray et al. 2018). As more open habitats spread, and seasonal swings steepened, plants that sequester nutrients in geophytes and hard-shelled nuts flourished: opportunities that ancestral populations could exploit. The scenario we hypothesize (Hawkes et al.1997, 1998, 2018; Blurton Jones et al. 1999; O’Connell et al. 1999, 2002; Alvarez 2000; Hawkes 2003, 2006, 2014, 2016; Hawkes and Coxworth 2013; Blurton Jones 2016; Parker et al. 2016) can be summarized as follows.

Economics of foraging for savanna resources

Ancestral adults could colonize more open habitats to take advantage of savanna resources, pursuing and processing more than enough food for their own consumption. But, as with modern humans today, youngsters were not big enough or strong enough to earn high enough return rates on these foods to feed themselves (e.g., Blurton Jones et al. 1989; Hawkes et al. 1995a; Blurton Jones 2016; Crittenden 2016). Mothers in those habitats would have to subsidize offspring while they were nursing and then continue to invest more in each one, lengthening their birth intervals. But unlike mothers’ milk, these foods could be supplied by others big and strong enough to handle them.

Of special importance, the savanna foods of interest do not promote scramble or interference competition where more consumers reduce everyone’s return rates because amounts available for consumption increase with extraction and processing effort. Foragers get mutualistic advantages from gregarious acquisition. Deeply buried geophytes allow increasing rather than diminishing returns for additional effort. Return rates are higher for accumulating piles and bulk processing rather than extracting, processing, and consuming items one at a time. All such economies of scale multiply with cooking (O’Connell et al. 1999; Hawkes and Coxworth 2013; Parker et al. 2016) where start-up costs and maintenance of cooking fires increase marginal benefits for mutual batch processing.

Production of food in lumps in contrast to hand-to-mouth, eat-as-you-go foraging also results in opportunities for others to appropriate shares (Blurton Jones 1987). This was noted by Wrangham et al. (1999), Wrangham (2009) in association with review of evidence for the importance of cooking in human evolution. Wrangham inferred that risks of robbery for batches of resources would be grounds for enlisting a mate as guard. But we have drawn attention instead to the likely appropriators most immediately at hand: dependent juveniles (Hawkes and Coxworth 2013; Parker et al. 2016; Hawkes et al. 2018).

Age and sex preferences, risky versus predictable resources

Although systematic behavioral observations of Ache hunter-gatherers in the forest of eastern Paraguay quantified high failure rates and collective consumption that made men’s preferred resources unsuitable for paternal provisioning (Kaplan et al. 1984; Hawkes 1990, 1991, 1992, 1993a,b; Hill and Kaplan 1993), it was subsequent ethnographic observations of age and sex differences in foraging among Hadza hunter-gatherers just south of the Serengeti in northern Tanzania that prompted this grandmother hypothesis about the evolution of human life history (Blurton Jones et al. 1989; Hawkes et al. 1995a, 1997, 1998, 2018; O’Connell et al. 1999, 2002; Hawkes 2003, 2014, 2016; Hawkes and Coxworth 2013; Blurton Jones 2016). Hadza men hunt big game. With powerful bows and arrows, they were always successful at aggressively scavenging carcasses from carnivores. But their average rate of large carcass acquisition by both means was less than 3.4% per hunter-day (Hawkes et al. 1991, 2001a). The hunter credited with the kill got no special share for his household (Hawkes et al. 2001b, 2014). Meanwhile, older women’s foraging productivity was high (Hawkes et al. 1989); and, although children were active foragers (Blurton Jones et al. 1989), youngsters are too small to be very effective at digging the deeply buried tubers that are year-round staples (Blurton Jones et al. 1989; Hawkes et al. 1995a; Blurton Jones and Marlowe 2002). Children’s weight gains depended on their mothers’ foraging effort until she had a new baby and that relationship disappeared; then weaned children’s gains depended on grandmothers’ foraging (Hawkes et al. 1997, 2001a). Effects of grandmothers on grandchild survival are large in this population (Blurton Jones 2016).

If modern mothers and grandmothers foraging in such savanna habitats leave more descendants with this division of labor, then similar tradeoffs could have driven the ancestral evolution of human age-structures. Unlike humans, great apes become decrepit with age in their mid-30s (Goodall 1986) and rarely out-live their fertility (Hill et al. 2001; Hawkes 2003; Emery Thompson et al. 2007; Hawkes and Smith 2010; Muller and Wrangham 2014; Wood et al. 2017; Emery Thompson and Sabbi 2019). Humans have much lower adult mortality rates. People who survive to adulthood remain healthy and productive decades longer, and women usually live well past menopause (Blurton Jones et al. 2002; Hawkes 2003; Gurven and Kaplan 2007). Our postmenopausal life stage (Levitis et al. 2013) is not shared by other primates (Alberts et al. 2013).

Phylogenetic context of human life history evolution

Our human radiation evolved within the hominid family where genetic evidence now places genus Pan (chimpanzees and bonobos) closer to us than to gorillas, themselves closer to us than to orangutans (Perelman et al. 2011; Prado-Martinez et al. 2013; Groves 2018). While hominids are the largest bodied, largest brained, latest weaning, and longest maturing of all the primates, the non-human hominids have smaller brains, shorter adult lifespans, earlier ages at maturity, and yet later weaning ages with longer birth spacing than we do (e.g., Hawkes et al. 1998; Robson et al. 2006). Similarities among our closest living cousins are grounds for inferring features that likely characterized our most recent common ancestors. As in other primate taxa, females reach menopause—if they live long enough; but unlike us, they rarely survive beyond their cycling years (Emery Thompson et al. 2007; Herndon et al. 2012). Female fertility ends at about the same age in all hominids, making that likely an ancestral feature. Yet vital rates for living people who forage for a living (Howell 1979; Hill and Hurtado 1996; Blurton Jones 2016) show girls that survive to adulthood are very likely to long survive their fertility, just as in agricultural and industrial populations. Distinctively large fractions of human female-adult-years are lived after menopause (Alberts et al. 2013; Levitis et al. 2013).

Grandmothering in the context of systematic variation in mammalian life histories

Charnov noted that among female mammals average adult lifespan and age at first birth vary with body size at the same allometric rate and constructed a model of female life history evolution aimed at explaining the allometries (Charnov 1991, 1993). From mice to elephants, the product of the average age at first birth and adult mortality (the inverse of average adult lifespan) is nearly constant, an invariance that Charnov explains as a trade-off between benefits of growing longer before reproducing and risk of dying first. Charnov (1993) plotted primate data (from Harvey and Clutton-Brock 1985) to show the correlation between age at first birth and female adult lifespan. At the time the other great apes were classified together as pongids with humans the only hominids and Charnov’s figure includes a point for each. Both fit nicely on the regression, with humans—of course—the highest point. Unremarked at the time, this should have been puzzling. Human females, unlike the other primates, do not continue bearing offspring throughout adulthood. As I review below, both Fisher (1930) and Hamilton (1966) had noted that post-fertile women reproduce their genes through grandmothering. If ancestral grandmothers’ contribution to the ancestry of future generations explains why humans fit that mammalian invariant, it should imply another distinctive, measurable consequence within the model. Rates of baby production should be higher (birth intervals shorter), than predicted for non-grandmothering mammals with our age at first birth. Demographic estimates for ethnographic hunter-gatherers are directly consistent with that expectation (Hawkes et al. 1998; Robson et al. 2006; Blurton Jones 2016).

Mathematical modeling of the grandmother hypothesis

Yet, could grandmothering subsidies actually propel the shift from ancestral great ape-like life histories to human-like ones? In mathematical models that assume the allometric relationships identified in Charnov’s model for female mammals and start at equilibrium population age structures like those of living great apes, two-sex models show that the addition of grandmothers’ subsidies does just that (Kim et al. 2012, 2014, 2019; Chan et al. 2016, 2017). Kim’s agent-based models begin with a great ape-like life history and with longevity mutating and so available to selection, simulations maintain that equilibrium over a million (simulated) years with very few females outliving their fertility. When those few can subsidize dependents, mothers bear next offspring sooner. Subsidizing grandmothers leave more descendants. That drives simulations from the great ape-like equilibrium to new equilibrium age structures like those of living hunter-gatherers where about a third of the adult females are past their childbearing years.

Kim’s initial models (Kim et al. 2012, 2014) fixed the end of childbearing at age of 45 years based on the similarity between humans and great apes. The question “why 45?” was explicitly left for subsequent work with the aim to see whether, given that end to female fertility, grandmothering could nevertheless propel the evolution of postmenopausal longevity. Since the end of female fertility varies among the mammals—earlier in monkeys than in apes and extending to even older ages in some mammals (e.g., two decades longer in elephants, Moss 2001; Lee and Moss 2011)—we surmised it was selection, not some constraint of biochemistry that accounts for the persistence of ancestral female fertility termination as longevity increased. Tradeoffs between continued fertility on the one hand or more grandmothering subsidies on the other as suggested by our Hadza data seemed likely. Modeling then took up that trade-off (Chan et al. 2016, 2017; Kim et al. 2019). Two-sex models that allow both longevity and the end of female fertility to evolve show that grandmothers’ subsidies alone increase longevity and also hold the end of female fertility below 50 years.

Mating sex ratio consequences of our postmenopausal longevity

These models of the evolutionary consequences of grandmothering assume that the cost of greater longevity in females is later age at first parturition and longer offspring dependence as in Charnov’s mammal model. However, Kim’s and Chan’s models include both sexes. Increased longevity gives males more paternity chances, at the cost of lower success in competition with shorter lived males. Offspring inherit the mean of their parents’ expected adult lifespans. When grandmothers subsidize dependents, mothers have next offspring sooner, so subsidizing grandmothers have more descendants. As longer-lived grandmothers can subsidize more, the older fraction of the population increases. When the age of female fertility ends is also allowed to evolve, it remains little changed. Since old males continue to be fertile—consistent with mammalian reproductive physiology where spermatogenesis persists through adulthood, the survival of those old males turns the sex ratio in the fertile ages from female- to male-biased.

Coxworth et al. (2015) highlighted these changes in mating sex ratios that come along with our grandmothering life history by running simulations of Kim et al.’s (2014) agent-based grandmothering model, plotting the sex ratio in the fertile ages instead of longevity. Attention to that fertile-age sex ratio and consequences for mating strategies has grown in the past couple of decades (see, e.g., Schact et al. 2017). Widespread usage labels the sex ratio in the fertile ages “the Adult Sex Ratio” (ASR) since adulthood in most animals is spent producing offspring (more on this below). In our lineage, however, a substantial fraction of female adult years-lived is post-fertile. Older fertile males push the sex ratio in the fertile ages from the female bias—typical of mammals generally—to become male biased. The more male-biased these ratios are the fewer conception opportunities per male. When the pool of competitors is large, guarding a current mate can give higher fitness benefits than competing for another.

Focusing only on male mating strategies, Schacht and Bell (2016), Loo et al. (2017a, 2017b) have modeled the effects of varying mating sex ratios on the relative success of three male strategies: multiple mating, dependent care, and mate guarding (Hawkes et al. 1995b). Models converge on a poor showing for dependent care, dominance of multiple mating when sex ratios are female biased, then mate guarding becomes the winning strategy when the bias is toward males—although when the switch occurs depends on the effectiveness of guarding (Loo et al. 2017b). These findings and observations about the effects of mating sex ratios on male strategies across a wide range of species suggest “that human pair bonds evolved with increasing payoffs for mate guarding, which resulted from the evolution of our grandmothering life history” (Coxworth et al. 2015, 11810).

Increased longevity and the evolution of big human brains

The fundamental role of adult mortality (indexed by average adult lifespan) and the correct scaling relationships among life history traits are central to Charnov’s (1993) explanation for the variation in female mammal life histories. Lower adult mortality favors lengthened duration of development, and that is also correlated with expanding brain size (Gonzalez-Lagos et al. 2010). As final brain size expands, brain components increase allometrically, across the eutherian mammals, their proportion of the whole scaling with adult brain volume (Finlay and Darlington 1995; Reep et al. 2007; Workman et al. 2013; Charvet and Finlay 2014). Finlay and colleagues have documented this variation, explaining how it results from the duration of development. Their demonstration that constraints of mammalian neural ontogeny underlie the regularity is a direct challenge to persistent claims that human evolution has been driven by selection for a big brain, particularly an expanded neocortex. To quote Finlay (2019, 318):

The central fact about the nature of the neocortex that has consistently managed to escape attention is that the cortex is precisely the volume it should be for a primate of our overall brain volume … . [Emphasis in original]

Consequences for distinctive features of human cognition are large because the slower neural development that results in our larger brains is combined with earlier weaning (Finlay and Uchiyama 2017; Hawkes and Finlay 2018). From the perspective here, both are consequences of ancestral grandmothering. Ancestral baby brains were wiring in the context of our novel life history. Unlike the independent mothering of the other great apes, where infants are not weaned until they can feed on their own, ancestral human mothers were bearing next offspring while the previous one was still completely dependent. Finlay (2019, 317) tallies costs of the misdirected attention to our “exceptional” neocortex, which:

… has caused both neuroscientists and psychologists to prematurely assign functions distributed widely in the brain to the cortex, to fail to explore subcortical sources of brain evolution, and to neglect genuinely novel features of human infancy and childhood … . [Emphasis added]

Cognitive consequences of our grandmothering life history

Hrdy (1999) pointed out that allomothering allowed human mothers to shorten birth intervals so much that we became more like “litter-bearers” than our singleton-rearing evolutionary cousins. With overlapping dependents, a mother’s fitness depends on how she juggles investments in more than one offspring at a time. Unlike our ape cousins or other anthropoids where it is male infanticide that can be a danger to infants, human mothers sometimes abandon or kill their own offspring (Hrdy 1999, 2016b). With mothers’ distributed attention, human babies are without full maternal engagement as a birthright. Yet infant survival depends entirely on others. Novel survival challenges confronting infants, Hrdy (2009) argued, could explain the notably precocious sociality of human infants, the beginning of our distinctive human appetite for shared intentionality (Tomasello et al. 1993, 2005). Ancestral infants’ own success at engaging support had life or death consequences (Hrdy 2009, 2016a; Hawkes 2012, 2014; Tomasello and Gonzalez-Cabrera 2017).

The grandmother hypothesis links features of savanna foods to advantages for mutualistic foraging and economic interdependence in ancestral populations. That interdependence dominated the world of ancestral babies whose survival turned on social relationships. The dependence continued through the neural and somatosensory development of infancy, but it did not stop then. Wired to prioritize relationships with others, reputations and mutual understanding were crucial lifelong. From this perspective, our appetite for anticipating the preferences of others, seeking their approval, and mutual engagement are aspects of our grandmothering life history, which continue to construct the diversity of our cultural lives (Hawkes 2020).

None of the observations that initially prompted skepticism about the hunting hypothesis or those that stimulated an alternative grandmother hypothesis with its many entailments summarized above were available to Fisher, Williams, and Hamilton as they laid the foundations of evolutionary life history theory. They all used human examples to illustrate more general regularities. As they sought to explore the effects of natural selection, they encompassed variety in the living world far beyond humans, or primates, or even vertebrates. Yet, they had identified the large fraction of productive postmenopausal women in human populations and had seen that to be a major clue to what happened in our own evolution. At the time they wrote, human phylogenetic proximity to the other great apes and the variation in life history and social behavior within our hominid clade were largely unknown; no vital rates were available for human populations depending on wild foods under mortality regimes different from those of agricultural and industrial populations; regularities in life history variation and neural development across the mammalian radiation were still to be discovered. Yet as they sought to understand and explain the process that Williams (1966a) called Mendelian natural selection, they could not ignore the striking fact that all human populations include large proportions of healthy and productive women with zero reproductive value—if that is defined by current and future age-specific fertility.

Fisher’s reproductive value

The concept and evolutionary importance of reproductive value is a central element of Fisher’s foundational role in biology’s modern synthesis. Fisher (1930) established the connection between the consequences of Mendelian inheritance (Fisher 1919) and natural selection to show the combination results in adaptive evolution. Fisher (1927) reviewed the role of vital rates (age specific fertility and age-specific mortality) in determining what happens to age-structured populations over time. Then Fisher (1930, 27) defined reproductive value and its importance this way:

We may ask … about persons of any chosen age … [to] what extent will persons of this age, on the average, contribute to the ancestry of future generations? The question is one of some interest, since the direct action of Natural Selection must be proportional to this contribution. [Emphasis added]

Fisher had rehearsed how population age-structures become stable whenever age-specific fertility and mortality remain the same for a few generations and migration is negligible. Then individuals in any particular age class have a predictable chance of surviving that age class and each age class that follows. Combining those mortality rates with average fertility at each of those ages, the mortality rates of the resulting offspring at each age, and the average fertility at those ages gives the expected direct contribution of members of any age class to descendant gene pools. Fisher (1930, 27) then went on to make the qualification of particular interest here:

There will also, no doubt, be indirect effects in cases in which an animal favours or impedes the survival or reproduction of its relatives; as a suckling mother assists the survival of her child, as in mankind a mother past bearing may greatly promote the reproduction of her children …. [Emphasis added]

Fisher had anticipated the importance of the indirect fitness effects that Hamilton (1964) would later elaborate. But then, in the very same paragraph, Fisher surmised that “[n]evertheless such indirect effects will in very many cases be unimportant compared to the effects of personal reproduction.” By ignoring those indirect effects, he could take advantage of the renewal equation, which attends only to age-specific fertility and mortality. On the following page, he used vital rates from “about 1911” to calculate the “Reproductive value of Australian women” (1930, 28). Ignoring contributions to the ancestry of future generations from women past their child bearing years, he showed women’s reproductive value falling to zero before the age of 50 years when their fertility ends.

Reproductive value and somatic versus current reproductive effort

Williams (1966b) considering how selection would adjust aging rates by trading off somatic and reproductive effort, took up Fisher’s concept of reproductive value, but, like Fisher himself, Williams ignored “indirect effects,” attending only to age-specific fertility and mortality. This quote from Fisher (1930, 43–4) was Williams’s opening epigram:

It would be instructive to know … what circumstances in the life-history and environment would render profitable the diversion of a greater or lesser share of the available resources towards reproduction.

To address that question, Williams divided “the mean amount of future reproductive success for individuals of [a given] age and sex in the population” into the part “immediately at stake” in any current reproductive effort and the rest, which he called residual reproductive value. Since,

… expenditures on reproductive processes must be in functional harmony with each other and worth the costs in relation to the long-range reproductive interest; and the use of resources for somatic processes is favored to the extent that somatic survival, and perhaps growth, are important for future reproduction. (Williams 1966b, 687) [Emphasis added]

Tradeoffs between the present and the future are at the heart of evolutionary explanations for the ubiquity of senescence. Williams (1957) had earlier explained why both senescence itself and variation in the pace of senescent decline are inevitable results of a history of natural selection for a wide array of taxa that include the vertebrates. He had built on Medawar’s (1952) crucial observation that even in an imagined non-senescing taxon, the force of selection would weaken across adulthood because accidents alone would decrease cohort size over time. With fewer members, contributions to future gene pools would fall and so the force of selection would decline with adult age. The rate of decline would depend on how fast the cohort shrinks: the adult mortality rate. Williams noted that rate of decline should also depend on what survivors to older ages could do for their fitness. Indeterminate growers like fish that continue to get larger throughout adulthood, produce more eggs the bigger they get. That raises the fitness benefits of reaching older ages favoring slower senescence and longer adulthoods compared to determinate growers like mammals and birds that do not keep growing in adulthood.

Reproductive value, grandmothering and increased longevity

Williams’s (1957) verbal arguments about the effects of selection on senescence were mathematically evaluated by Hamilton (1966) who began with a thought experiment about humans in which four hypothetical genes each give immunity against some lethal disease for one particular year:

Suppose the first gives immunity for the first year, the second for the fifteenth, the third for the thirtieth, and the fourth for the forty-fifth …. If for further simplicity parental care is ignored and it is assumed that the menopause always comes before age 45, it is at once obvious that the fourth gene is null, whereas all the others do confer some advantage. (Hamilton 1966, 12–13) [Emphasis added]

Hamilton set parental care aside for the simpler treatment allowed by using only age-specific fertility and mortality. He would return to its complications, considering humans specifically but without exploiting his own previously published observations (1964) that parental care itself is just a special case of something much more general. I will come back to that below.

Hamilton’s thought experiment about selection on age-specific immunity was his set-up for using demographic modeling to investigate how the “age at which a gene acts affects its influence on fitness” (1966, 14). His modeling demonstrated that “for organisms that reproduce repeatedly, senescence is to be expected as an inevitable consequence of the working of natural selection” (1966, 26) [emphasis in original]. He confirmed Williams’s inference that selection made senescence inevitable, even among indeterminant growers. Then, turning to evaluate predictions of his theory with empirical examples, Hamilton again began with humans because,

Man is the species for which much the best data is available. … Unfortunately, there are no very good data for contemporary peoples in a pre-agricultural phase, nor even for those with the most primitive forms of agriculture. (1966, 28)

So, he chose rural Chinese farmers in Taiwan around the turn of the 20th century, constructed age-specific survival and age-specific fertility curves, estimated the Malthusian parameter, and calculated what he called reproductive value curves, ignoring indirect contributions to future gene pools. His reproductive value curve for Taiwanese women about 1906 (Hamilton 1966, 32 Figure 3) is very like Fisher’s (1930, 28 Figure 2) for Australian women around 1911. In both, female reproductive value peaks at about the age of 20 years and then falls to zero before the age of 50 years.

Hamilton’s modeling supplied mathematical support for Williams’s (1957, 407) inference that “There should be little or no postreproductive period in the normal lifecycle of any species.” About that inference, Williams himself said:

At first sight it appears that this prediction is not realized. Long post-reproductive periods are known in many domesticated animals and in man himself. In man it may even be longer than the reproductive period. However, these observations lose much of their seeming importance when it is realized that they are largely artifacts of civilization. In very primitive conditions, such as prevailed throughout almost all of man’s evolution, post-reproductive individuals were extremely rare. (1957, 407)

Williams supported that inference with age-at-death estimates for a sample of fossil skeletal specimens, which—as now widely recognized—could neither be accurately aged nor correctly reflect past population age structures (e.g., review in Hawkes and Blurton Jones 2005).

Assuming menopause to be uniquely human, Williams also suggested a second hypothesis about its evolution as a consequence of other distinctly human features. That second hypothesis was cited by Hamilton as he considered the notable discrepancy between the survival curve for Taiwanese women falling to zero only near 90 years, while their fertility ended before the age of 50 years (Hamilton 1966, 29 Figure 1).

It is evident that the rise in mortality in the later reproductive ages of man is by no means asymptotic to the age at which reproduction ends; the indefinite rise comes too gradually and too late. This is particularly evident from the curves given for the Taiwanese women, where the rather definite age of the menopause seems to be conspicuously ignored by the as yet gently rising curve of force of mortality. It is, moreover, a matter of common knowledge that the post-menopausal woman normally remains a useful and healthy member of the community for some time. A woman does sometimes live to twice the age of her menopause …. [The] comparatively healthy life of the post-reproductive women is so long … that it inevitably suggests a special value of the old woman as mother or grandmother during a long ancestral period …. (Hamilton 1966, 37)

Hamilton continued, “As remarked by Williams, an obvious excuse for this discrepancy is to be found in the factor of parental care.” He was citing Williams’s (1957) second surmise about the evolution of menopause which did not mention a long ancestral value of grandmothers. Williams said,

At some time during human evolution it may have become advantageous for a woman of forty-five or fifty to stop dividing her declining faculties between the care of extant offspring and the production of new ones. A termination of increasingly hazardous pregnancies would enable her to devote her whole remaining energy to the care of her living children, and would remove childbirth mortality as a possible cause for failure to raise these children. Menopause, although apparently a cessation of reproduction, may have arisen as a reproductive adaptation to a life-cycle already characterized by senescence, unusual hazards in pregnancy and childbirth, and a long period of juvenile dependence. (Williams 1957, 407–8)

Colleagues and I have characterized this Williams hypothesis as “stopping early,” since it assumes an ancestral condition when female fertility continued to older ages. In contrast, Hamilton’s language suggests that selection favored slower senescence and greater longevity when post-menopausal females contributed to the production of descendants (e.g., Hawkes et al. 1997, 1998; Hawkes and Smith 2010; Hawkes and Coxworth 2013). The point to underline here is the explicit recognition of grandmother effects both by Fisher and Hamilton (and by Medawar 1952, fn 1). Fisher himself had recognized that indirect effects could make substantial contributions to reproductive value. Only by ignoring them (as Fisher himself had done) could Hamilton (1966) and Williams (1966b) focus just on age-specific fertility and mortality.

Inclusive fitness

Hamilton’s (1964) own explanation of the pervasive importance of indirect effects began with this observation:

With very few exceptions, the only parts of the theory of natural selection which have been supported by mathematical models admit no possibility of the evolution of any characters which are on average to the disadvantage of the individuals possessing them. … Sacrifices involved in parental care are a possibility implicit in any model in which the definition of fitness is based, as it should be, on the number of adult offspring. … The selective advantage may be seen to lie through benefits conferred indifferently on a set of relatives each of which has a half chance of carrying the gene in question. (1964, 1)

From that genic perspective, he went on to say

… there is nothing special about the parent-offspring relationship except its close degree and a certain fundamental asymmetry. The full-sib relationship is just as close. … Opportunities for benefitting relatives, remote or not, in the same or an adjacent generation … must be much more common … . (1964, 2)

He then laid out a mathematical argument about selection for both “giving” and “taking” traits that incorporated those relationships, concluding that:

we have discovered a quantity, inclusive fitness, which under the conditions of the model tends to maximize in much the same way that fitness tends to maximize in the simpler classical model … we may consider whether a given character expressed in an individual … is or is not adaptive in the sense of being in the interest of his inclusive fitness. (1964, 8)

Inclusive reproductive value

Hamilton (1964) had demonstrated that classical models miss what his theory building said selection will maximize. Fisher (1930) had recognized that such indirect effects could be important specifically in humans where “a mother past bearing may greatly promote the reproduction of her children.” When Hamilton used Fisher’s (1930) concept of reproductive value two years after publication of his inclusive fitness papers and excluded indirect effects to model the age-specific variation in selection on senescence, he noted that:

it should now be evident that the ratio of the reproductive values is just as important as the coefficient of relationship in determining ideally adaptive social behaviour: the coefficient gives the chance that the offspring carries a replica of a behaviour-causing gene of the parent (Hamilton, l964 … ), while the ratio gives the relative conditional expectation of its reproduction. The inclusive fitness of an individual is maximized by its continually acting in ways that cause increases in its inclusive reproductive value. (1966, 22–3) [Emphasis added].

Recognizing the problem with classic reproductive value, Hamilton might have pursued solutions further. But he subsequently wrote that by the time his 1966 paper was published, “… the theme of senescence no longer excited me much … because my findings on it were not new” (1996, 87). He was already working on what would become his 1967 “Extraordinary sex ratios” paper.

The Fisher condition, sex ratios, sex allocation, and sexual conflict

Hamilton’s (1967) field defining contributions to sex ratio evolution started with analysis of Fisher’s explanation for the common pattern of equal offspring sex ratios, now widely known as “the Fisher Condition.” Hamilton then highlighted implicit assumptions in Fisher’s treatment that would not always hold. I will return briefly to Hamilton’s next steps, but to provide a framework for connecting the topic to human life history evolution I begin discussion of the Fisher condition and sex ratio evolution with Charnov’s (1982) treatment. Charnov (1982, 8) defined the problem of sex allocation this way, “what is the equilibrium sex ratio (proportion of males among offspring) maintained by natural selection?” plus four additional questions about allocation to male or female function in hermaphrodites and sex changers.

… all of the five questions are really one question phrased in different forms. Consider a typical diploid organism, RA Fisher (1930) noted the seemingly trivial fact that with respect to autosomal genes, each zygote gets half its genome from its father, half from its mother. To put it simply: everyone has exactly one father and one mother. However, far from being trivial, this fact holds the key to understanding sex allocation in diploids.

… [It] has two implications. First, an individual’s reproductive success through male function (sperm) is to be measured relative to the male function of other individuals (vice versa for female function). Second, since half the zygote genes come via each pathway, male and female function are in a real sense equivalent means to reproductive success. Consider dioecy and the sex ratio. If many daughters are being produced, then large reproductive gains accrue to the producers of the rare sons. (1982, 8–9)

Charnov went on to discuss the resultant frequency dependent selection:

Maynard Smith (1976) has termed the equilibrium value of a trait an “Evolutionary Stable Strategy (or an ESS).” Suppose we have a population made up of individuals who have some attribute Z; we introduce into this population a rare genotype with alternative attribute $\widehat{z}$ and see whether the $\widehat{z}$ individuals are selected for or against (i.e., does the rare mutant spread?). If for some character of interest … there exists a Z such that all deviants are selected against, Z is termed an ESS. The classic example is selection on the primary sex ratio (Fisher 1930) where the ESS is one-half males at conception in the simplest case.

He then (Charnov 1982, 13 and following) used “the first formal treatment of sex ratio evolution for diploids, from an early paper by Shaw and Mohler (1953) … [to write] the Shaw-Mohler equation for sex ratio (… first derived by MacArthur, 1965),” which gives the sex allocation rule:

Selection favors a mutant gene which alters life history parameters if the percent gain in fitness through one sex function exceeds the percent loss through the other sex function …. It is very often the case that the ESS allocation of resources to male versus female function is that which maximizes the product of the fitness gain through male function … times the fitness gain through female function … Or, more simply selection maximizes m * f. (Charnov 1982, 17) [Emphasis in original]

Robert MacArthur (1965, 390) had treated the same question as follows:

Consider any autosomal gene influencing family size or sex ratio. Half of the genes in the population at this locus came from female parents and half from male parents. Hence, from the grandparent’s viewpoint, the set of all their sons will contribute equally to the set of all their daughters, and precisely one half of the genes at this locus will be expected to have come from the grandparent generation by way of sons and the remaining half by way of daughters.

MacArthur concluded,

… that natural selection will favor that family composition (M[sons], F[daughters]) which maximizes M x F. In this argument M and F … are calculated by counting males and females at the end of the period of parental care, if any, or at the onset of reproduction. … If the genes affect the age at which the mother gives birth (so that generations are no longer synchronized), then the M’s and F’s … must be replaced by the reproductive value of males and females as defined by Fisher. In this case maximizing M x F will involve the simultaneous selection of sex ratio, clutch size, and age at reproduction. This model can be extended to any other factor influencing the reproductive values of the offspring. [Emphasis added]

Before addressing implications of all of this for grandmothering and human evolution, I return to Hamilton (1967). There Hamilton relaxed Fisher’s (1930) assumption of random mating, a step perhaps prompted at least in part by his 1964 paper’s recognition of substantial interaction between neighborhood composition and inclusive fitness selection. In “Extraordinary sex ratios,” he noted that both theory and empirical observations show large effects on offspring sex ratios from mate competition among relatives. For example, parasitoid wasp mothers put their eggs into hosts where the offspring mature and mate with each other. Sons can participate in conceptions with many females, so brothers compete with each other for the same conceptions, and the number of conceptions possible depends entirely on the number of sisters. Then, mothers that bias their offspring sex ratios toward daughters leave more grandchildren.

Instead of further work on sex ratios, Hamilton turned to sexual reproduction itself and why it is so common in spite of its evident substantial costs (e.g., Williams 1975; Maynard Smith 1978). His own report about that shift is more than worth attention—especially in the context of this symposium,

… my own reasoning on sex ratios … had shown how under conditions of extreme inbreeding … reproduction could evolve to be much more demographically efficient. Such efficiency came through a drastic reduction of male production, raising very acutely the problem of why males were ever there in the first place. Both my first strongly upward arching graph of my simulation in my ‘Extraordinary sex ratios’ … and all of that paper’s small inbreeders with their male-deficient sex ratios, seem now to be joining in chorus to force me to attend to the issue … I knew from my reading that parthenogenesis … is present throughout both animal and plant kingdoms. If suitable mutations to parthenogenesis can happen, as these cases prove, and if parthenogenesis is so efficient, why weren’t waves of such self-sufficient females … replacing sexual organisms on all sides? How in the long run has this crazy whim of maleness proved itself the opposite of what it should have been – a fleeting disaster and a long-abandoned experiment? (Hamilton 1996, 354)

The Fisher condition, offspring sex ratios, and mating sex ratios

In the case of our own evolution, well into the mammalian radiation, males may be here to stay. As noted above in the quote from Charnov, the Fisher condition explains why, in most diploids, there are so many males. Even though the number of babies depends on the number of females that can bear them; and even though one male can usually inseminate many females almost concurrently, the fact that half a baby’s autosomal genes come from each parent has enormous consequences. If males are rare, those rare males father all the babies. Since their average reproductive success must be higher than the female average, biasing offspring sex ratio toward males gives more grandchildren. If fertile males become more abundant than fertile females, the average reproductive success for the more plentiful males falls below the female average and mothers get fewer grandchildren through sons. The Fisher condition usually pushes offspring sex ratios toward the ESS of equal investment in sons and daughters.

The same Fisher condition also has consequences for mating strategies (e.g., Parker 1978) with attention growing recently under the label ASRs. The same issue in humans can be framed as “mating markets” or “partner scarcity” (e.g., Schacht et al. 2017). As reviewed above, the grandmother hypothesis links notably male-biased sex ratios in the fertile ages to the evolution of human longevity. In Coxworth et al. (2015), both modeling and empirical data on hunter-gatherer and chimpanzee age structures are consistent about that distinctive male bias in humans. If the ancestral great ape-like condition is assumed similar to chimpanzees, the average proportion female in the fertile ages is 63% for the life tables cited; from four hunter-gatherer life tables, the average female proportion is 38%.

The female bias in chimpanzees is a common pattern in mammals as male mating competition usually includes risky and aggressive behaviors that raise male mortality. Fisher and others following him explained that mortality risk after individuals become independent would not affect equilibrium offspring sex ratios because the fraction that dies raises the average reproductive success for survivors. For example, if adult death rates in males are higher than adult female death rates, probable paternities for a son are conditional on his survival. The female biased sex ratio in adult chimpanzees results from such higher male mortality after weaning, so the higher number of grandchildren a mother might get through a son if he survives is canceled out by the chance he won’t. Along these lines, the Fisher condition usually predicts equal investment in sons and daughters in most mammals, even though ASRs are commonly female-biased.

On the contrary, while mortality is higher at most ages in human males too (e.g., Kruger and Nesse 2006), that higher mortality is overwhelmed by the countervailing source of sex-bias among fertile adults: all those old men. According to the grandmother hypothesis, increased longevity was favored in ancestral populations as the economic productivity of females whose own fertility was ending subsidized the offspring production of those still fertile. Greater longevity through grandmothering resulted in shorter birth intervals so more paternities for fertile males to compete over at the same time it expanded the ranks of competing males. In our lineage adult females depart the fertile ages with approaching menopause. That departure is not mortality, could its effects on adaptive offspring sex ratios cancel in the same way?

This is a puzzle posed by ubiquitous male-biased sex ratios in the fertile ages in our lineage. According to the grandmother hypothesis favored here that bias has persisted for tens of thousands of generations. In spite of likely higher male mortality at most ages, fertile males outnumber fertile females. With our grandmothering life history, the average reproductive success of sons has always been predictably lower than the average for daughters. In Hamilton’s (1966) senescence paper, his attention was elsewhere, but he did surmise that

It is fair to say that as a whole the data … cannot be reconciled with Fisher’s theory of the sex-ratio … A Taiwanese couple could reasonably expect a greater total number of grandchildren if they concentrated on supplying the existing deficiency of females; but they show no inclination to do so. (Hamilton 1966, 33)

Hamilton suggested, “It seems possible that we have here a case where a human cultural factor, the emphasis on maintaining a male line of descent, with its resultant preferential treatment for male children, has balanced the sex-ratio selection at some distance from its natural equilibrium.” That example and other human datasets contributed to Williams’ skepticism about adaptive sex ratio adjustment in diploids with chromosomal sex determination. After reviewing both models and data, Williams (1979, 587) said:

I find it rather mysterious that adaptive control of progeny sex ratio seems not to have evolved. In particular the conformity of human progeny sex ratios to binomial distributions seems to contradict evolutionary theory. Either the physiological advantage of adjusting offspring sex to maternal capabilities, or the demographic advantage of decreasing competition for mates, ought to have produced noteworthy effects. Instead, deviations from random sex determination are trivial at best.

Williams’ expectation about “adjusting offspring sex ratio to maternal capabilities” was reference to the plausibility of Trivers and Willard’s (1973) hypothesis that (given the Fisher condition) selection should favor adjustment of progeny sex ratio depending on maternal condition. They proposed that if maternal condition predicts offspring condition; and if reproductive success has higher variance in males than in females; and if sons gain more from better condition while daughters lose less from poor condition; then selection should favor maternal tendencies to adjust offspring sex ratio with their own condition. Greater variance in male reproductive success usually follows from anisogamy (Parker 2014; Parker and Pizzari 2015), but both theoretical and empirical work has since shown complications for Trivers–Willard expectations (e.g., reviews in Frank 1990; West 2009; Navara 2018). Veller et al. (2016) have argued that Trivers and Willard’s (1973) paper contains two distinct hypotheses, one about offspring sex ratios and the other about sex-biased investment, only the first of which holds under very general conditions.

Evidence for sex-biased investment is sometimes remarkably strong in human populations. Hamilton was right to surmise a cultural preference for sons among those Chinese farmers in Taiwan. Extreme sex-biased treatment of offspring associated with unilineal inheritance of property and privilege in class stratified societies including historical China has been well documented. M Dickeman contributed to identifying some of the extremes and associated their lineage serving outcomes with the Trivers–Willard hypothesis (Dickeman 1975, 1979; Hrdy 1999). The dramatic sex-biased investment that Dickeman persuasively associated with maintaining lineage position occurs in very steeply stratified states. Those emerged in human populations only after the origins of agriculture, within the climate amelioration of the Holocene (Richerson et al. 2001). However, our grandmothering life history is much older, and the importance of male strategies, and their effects on options for females is a deep primate legacy (e.g., Hrdy 1981; Smuts 1992, 1995).

Sexual conflict, longevity, and the product theorem again?

Kim’s (Kim et al. 2012, 2014, 2019) and Chan’s (Chan et al. 2016, 2017) modeling of the grandmother hypothesis included both sexes. Model males and females both face longevity tradeoffs, but the costs and benefits of increased (or decreased) longevity differ for the sexes. That makes sexual conflict inevitable because of the Fisher condition, both mothers and fathers contribute to offspring longevity. So model populations must evolve to a compromise. MacArthur (1965, 390 quoted above) had derived and described what Charnov later called “the product theorem” for overlapping generations and concluded that “This model can be extended to any other factor influencing the reproductive values of the offspring.” Keeping in mind the problems with standard definitions of reproductive value raised above, that would make the product theorem apply to longevity whenever tradeoffs between somatic and current reproductive effort are adjusted by numerous autosomal alleles. Longevity in the population should evolve to maximize the product of expected future gene copies through males and females.

Easy to say, but Alan Grafen’s formal Darwinism project is still dealing with reproductive value as classically defined and treats inclusive fitness separately (Grafen 1999, 2006a, 2006b, 2014, 2020; Crewe et al. 2018; Levin and Grafen 2019). That project continues to demonstrate that proving how selection drives either reproductive value (ignoring “indirect effects”) or inclusive fitness alone poses difficult mathematical problems. Whether or not particular features of the human case can be helpful in combining them, the review here aims to show the need for it has long been recognized for our own lineage.

Chan’s modeling now seems a start at finding out whether grandmothering complications interfere with treating longevity like any other trait under sexual conflict. Chan et al. (2016) explored the ground laid by the agent-based modeling in Kim et al. (2014) with partial differential equations (PDEs) to investigate parameter effects. Chan allowed age at last birth as well as longevity to evolve (as subsequently did Kim et al. 2019). As in Kim et al. (2012, 2014), the model resulted in two equilibria—a great ape-like and a human-like one. The PDE results showed that:

… male competition, arising from a skew in the mating sex ratio toward males, plays a significant role in determining whether the transition from great ape-like longevities to higher longevities is possible and the equilibrium value of the average adult lifespan. (Chan et al. 2016, 145)

Chan et al. (2017) then separately calculated the male and female fitnesses across trait-space to visualize the sexual conflict over longevity. This “allows for a straightforward comparison between fitness landscapes of both sexes and the population compromise that evolves with and without grandmothering” (Chan et al. 2017, 2134). Plotting variation in “the expected number of births by a female” as a function of longevity may seem to reduce things to the classical elements of the renewal equation (age-specific fertility and mortality) thus falling back on excluding indirect effects. It does not because the whole life history evolves. When “number of births” as a function of longevity is modeled with grandmothering, the number of births expected incorporates the subsidies from the post-fertile years that shorten the birth intervals. For males, as longevity increases with grandmothering, the number of competitors expands (more old males) but so does the number of paternity opportunities (shorter birth intervals). Of course, “A natural question is, to which longevity value … does the population converge for given parameter values and trade-off functions of the model? … the answer is simply the longevity value … that maximises the product F(x)M(x)” (Chan et al. 2017, 2137).

Could this be simply a version of the product theorem? The history of ideas reviewed here makes that an arresting possibility. But the modeling assumed equal offspring sex ratios, so the puzzle of how offspring sex ratios can remain so close to equal remains. With male-biased mating sex ratios, producing more daughters should mean more descendants. Yet biases in human offspring sex ratios are as noted by Williams (1979) trivial at best. Further work will be needed. The lesson so far is that the unusual case of a grandmothering species may provide a more general lesson about sex ratios, sexual conflict, and the evolution of life histories.

Concluding discussion

When Fisher (1930) and Hamilton (1966) recognized that human postmenopausal longevity evolved from a history of natural selection, they established venerable foundations for the grandmother hypothesis. Fisher noted that the “indirect” reproductive value of “women past bearing” implied their role in the evolution of our longevity more than three decades before Hamilton (1964) explained the pervasive importance of inclusive fitness. Then, when Hamilton (1966) calculated the effects of natural selection on senescence, he found that classical models that use only direct reproductive value (fertility and mortality) cannot account for so many “useful and healthy” postmenopausal women. He concluded that selection actually maximizes “inclusive reproductive value.”

Combined with Fisher’s (1930) recognition that Mendelian inheritance implies equal contribution from both sexes to (autosomal) genomes, those insights link evolving postmenopausal longevity in our lineage to likely shifts in male mating strategies. As the fraction of post-fertile females expanded, the fraction of old males expanded too, shifting the sex ratios in the fertile ages of ancestral populations to a regular, persistent male bias (Coxworth et al. 2015). Empirical observations of diverse taxa and mathematical modeling both show that when mating sex ratios are male-biased, mate guarding dominates paternity competitions (e.g., review in Schacht et al. 2017). For our lineage, that connects the evolution of a distinctively human pair bonding habit to ancestral grandmothering (Coxworth et al. 2015). Instead of the paternal provisioning proposed in the hunting hypothesis to explain the evolution of human nuclear families, the grandmother hypothesis points to sexual selection and sexual conflict in the evolution of human life history. On Chan’s fitness landscapes, it is sexual conflict over longevity with grandmothering that propels the evolution of model populations from a great ape-like life history to a human-like one.

As reprised above, the grandmother hypothesis now promises to help explain not only postmenopausal longevity, but also our pair bonding habits, big brains, and distinctively cooperative social appetites. As part of our grandmothering life history, ancestral mothers shortened their birth intervals, investing less in each offspring which posed survival challenges to completely dependent ancestral infants and toddlers. The perils of that dependency wired precocious sociality and persistent concerns about relationships early in their slower developing brains. Other people’s actions and responses are central features of our cognitive ecology now because harms and benefits from social relationships have taken priority in developing human somatosensory systems in our human radiation from infancy onward (Hrdy 2009; Finlay and Uchiyama 2017; Hawkes and Finlay 2018; Hawkes 2020). Local understandings about reputations and preferences dominate our lives. Appetites to connect and align understandings with close kin and neighbors helped drive the diversity of languages and cultures in the past (Hawkes 2020). Recent times have witnessed the power of mass communication, now expanded exponentially by social media, to harness our fundamental concerns about reputations to global influencers. Instantaneous reach to distant audiences magnifies the persuasive power of conflicting views of the past, present, and future. According to the hypothesis favored here, credit—or blame—extends all the way back to ancestral grandmothers foraging in the spreading African savannas as climate changed their daily options millions of years ago.

Acknowledgments

I am grateful to Nick Blurton Jones, Matthew Chan, Eric Charnov, Sarah Hrdy, Peter Kim, and Jim O’Connell for their reliable collaboration, productive ideas, and patient advice. And thanks to Virginia Hayssen and Teri Orr for the symposium invitation to pull all this together.

From the symposium “SICB Wide Symposium: Reproduction: The Female Perspective from an Integrative and Comparative Framework” presented at the annual meeting of the Society for Integrative and Comparative Biology January 3–7, 2020 at Austin, Texas.