Humans wean fast, mature slow, and live decades past fertility — no other ape does all three
Start with the numbers, because the pattern is more extreme than the folklore about “human childhood being long” usually conveys, and because every explanation in this article is trying to account for the same three facts at once rather than any one of them alone.
Weaning first. Kaplan, Hill, Lancaster and Hurtado’s comparative synthesis of hunter-gatherer and chimpanzee life histories puts human weaning age at approximately 2.5 years against approximately 5 years for wild chimpanzees, drawing on demographic and growth data from the Ache, Hiwi, !Kung and Hadza against a synthetic chimpanzee life table pooling records from Bossou, Gombe, Kibale, Mahale and Taï [3]. Hawkes, O’Connell, Blurton Jones, Alvarez and Charnov’s independent estimate is close: about 2.8 years in humans against 3 to 6 years across the great apes [1]. Humans stop nursing while a chimpanzee infant still has one to three years of dependence on its mother’s milk ahead of it.
Age at first reproduction runs the other direction. Kaplan and colleagues’ Table 1 gives a forager mean age at first birth of 19.7 years, built from four groups studied while still largely independent of markets and modern medicine: the Ache at 19.5, the Hiwi at 20.5, the !Kung at 19.2, and the Hadza reported separately in the same dataset [3]. Their chimpanzee mean, from the same pooled sites, is 14.3 years. In the paper’s own words, “female chimpanzees give birth for the first time about five years earlier than do hunter-gatherer women” [3]. A chimpanzee is already a mother at an age when a forager woman’s own growth has barely finished.
And then the long tail. Kaplan and colleagues report that hunter-gatherer women “spend more than a third of their adult life in a postreproductive phase,” while “very few chimpanzee females survive to the postreproductive phase” at all [3]. Hawkes and Coxworth’s later review states the forager figure more precisely: among hunter-gatherer girls who survive to adulthood, about three-quarters live past age 45, and at any given moment a third or more of the adult women in a forager population are past childbearing age [5]. Gurven and Kaplan’s cross-cultural mortality synthesis, built from twenty-two hunter-gatherer, forager-horticulturalist and acculturated-forager populations rather than the four in the original comparison, finds a strikingly consistent shape underneath this: mortality falls sharply through infancy and childhood, then holds essentially flat until about age 40, then rises in roughly Gompertz fashion, with a modal age at adult death of about seven decades across the sample [12]. That figure, not the life expectancy statistics that infant mortality drags down, is the honest baseline for “how long a person who reaches adulthood is built to live,” and it is markedly higher than the chimpanzee comparison Kaplan and colleagues report: fewer than ten percent of chimpanzees survive to age 40, against more than fifteen percent of hunter-gatherers who survive to 70 [3].
Put the three numbers next to each other and the puzzle sharpens rather than resolving. A species that weans early is, on the usual mammalian logic, moving resources toward producing more offspring faster. A species that delays first reproduction by five years relative to a close comparator is, on the same logic, moving resources the other way, into growth and skill before any output. A species in which a large share of the reproductively spent population survives for decades afterward is carrying dependents — grown or not — who cannot be explained by ordinary offspring investment, because offspring investment stops paying by the time the investor can no longer reproduce. The three facts do not obviously belong to a single motive. One classic model puts a number on how much of the delay in maturity is simply growth: the allometric relationship between body weight and growth rate that Kaplan and colleagues cite from the wider mammalian literature is
where
Grandmothers who dig tubers: the Hadza data behind the grandmother hypothesis
The grandmother hypothesis did not start as a theory of menopause. It started as an anomaly in a foraging-effort dataset.
Hawkes, O’Connell and Blurton Jones spent field seasons with Hadza foragers in Tanzania recording, individual by individual, how much food each person acquired and how children’s weight changed over time. Their 1997 paper reports that children’s weight gains tracked their own mothers’ foraging effort — except when a mother had a new nursing infant, in which case the older, already-weaned child’s weight gain instead tracked its grandmother’s foraging effort [2]. A weaned child in a Hadza camp is not fed primarily by its own labor. When its mother’s attention shifts to a new baby, someone else, usually a grandmother, becomes the relevant provider, and the child’s growth outcome depends on that person’s foraging output rather than on the mother’s or the child’s own.
The mechanism the Hadza data pointed to was specific rather than generic: deeply buried tubers, a staple resource in the environments Hawkes and colleagues studied, that require strength and practiced digging skill to extract efficiently. Young children cannot process them; older women, past their own peak strength but with decades of practiced skill and no infant of their own to carry, turn out to be highly efficient at exactly this task. Kaplan and colleagues’ independently collected energy-acquisition data are consistent with unusually high output from this age-sex class in at least one of the three foraging populations they measured in detail: “Hadza children and postreproductive women appear to acquire substantially more food than do their Ache and Hiwi counterparts” [3], though the same paper flags that the absolute Hadza consumption estimates implied by that production figure look implausibly high against known body-weight ratios, a caveat about measurement rather than a reversal of the pattern.
Hawkes, O’Connell, Blurton Jones, Alvarez and Charnov turned this ethnographic observation into a formal life-history argument the following year, using Eric Charnov’s dimensionless framework for comparing life-history variables across mammal species. Their comparison of fertility-rate parameters is precise: the annual birth-rate parameter they call
The 1998 paper was a verbal and comparative argument, not a population-genetic model, and its authors knew the difference mattered: a plausible mechanism is not the same thing as a demonstration that the mechanism is sufficient, on its own, to move a chimpanzee-like lifespan to a human one. Peter Kim built the first agent-based test and found, in Hawkes and Coxworth’s later account of that unpublished work, that “very weak grandmothering was enough to drive the ape-like equilibrium to a human-like one” once the starting population included even a small proportion of postfertile women able to subsidize weaned grandchildren [5]. Kim, Coxworth and Hawkes then published the fully specified version. Starting from expected adult lifespans of 16 to 27 years — the ancestral, chimpanzee-like condition — and holding female fertility fixed to end at 45, their probabilistic model let care from postmenopausal grandmothers subsidize the survival of weaned dependents too young to feed themselves, with strict constraints: a grandmother could support only one dependent grandchild at a time, and infants under two were never eligible for that subsidy at all [4]. Across their scenarios, the population converged on final geometric mean adult lifespans of roughly 49.4 years, crossing the modern human target of 43 years within 24,000 to 56,000 years depending on the specific cost assumptions made about male longevity [4]. The share of caring adult females who were eligible grandmothers rose from under one percent at the ancestral equilibrium to 43 percent at the new one [4] — a large compositional shift in the adult female population, produced entirely by making postfertile survival occasionally payoff-positive.
That result is more fragile than a single successful simulation makes it look, and Hawkes and Coxworth say so themselves. Frederick Kachel and colleagues’ earlier two-sex agent-based model found that even when grandmother effects were built in strongly enough to guarantee grandchild survival, “those effects had little impact on longevity,” because the model gave males no cost for living longer, so the fitness advantage that should have come from a longer female lifespan was diluted across a male population that already lived long for free [5]. Kim’s later model works, on Hawkes and Coxworth’s account, specifically because it removes that assumption — but removing it is a modeling choice, not an empirical measurement, and the sensitivity of the result to exactly this choice is a legitimate point against treating the viability question as settled.
Menopause is nearly unique among mammals — until toothed whales complicate the story
The grandmother hypothesis and its rivals are all trying to explain the same rare trait: a long span of post-reproductive life is not the mammalian default. Most female mammals, including most primates, either reproduce until they die or have such short potential lifespans past their fertile years that the pattern is not evolutionarily interesting. If post-reproductive life is adaptive rather than a side effect of modern medicine, comparative biology predicts it should show up, independently, wherever the specific ecological and kinship conditions that favor it recur — and that is exactly the test five species of toothed whale unexpectedly supply.
Ellis and colleagues’ 2024 analysis compiled Bayesian mortality models and age-structured reproductive data across a phylogenetically broad set of 32 female toothed whale species, together with reproductive-lifespan estimates from ovarian corpora counts in 18 of them, to test whether menopause in whales evolves by species living unusually long relative to their body size (“live-long”) or by species truncating their reproductive lifespan while total lifespan stays typical for their size (“stop-early”) [11]. Menopause turns out to have evolved independently at least four separate times among toothed whales, in short-finned pilot whales, false killer whales, killer whales, narwhals and belugas [11]. The comparative result favors the live-long pathway specifically: species with menopause live on the order of 40 years longer than a same-sized species without it would be expected to, while showing no shortening of reproductive lifespan relative to expectation [11]. The paper also reports a “relative grandoffspring years” measure — a proxy for how much temporal overlap a female gets with her daughters’ offspring — of 1.04 in menopausal species against 0.36 in non-menopausal ones [11], the same directional signature the human grandmothering literature is built around, arrived at independently in an entirely different mammalian order.
Resident killer whales carry the direct behavioral evidence. Brent, Franks, Foster, Balcomb, Cant and Croft’s analysis of the long-studied northern and southern resident populations found that postreproductive females are disproportionately likely to lead pods during collective movement, and that this leadership role becomes especially pronounced in years when salmon, their primary prey, are scarce [9]. The authors describe this as the first evidence that the transfer of ecological knowledge by older, non-reproducing females can itself be a force of selection acting on survival after reproduction ends [9]. Nattrass, Croft, Ellis and colleagues followed with a direct survival analysis of the same population records, spanning 1973 to 2016 in the north and 1976 to 2016 in the south across 378 whales with known maternal grandmothers, totalling 4,578 whale-years of observation [10]. They found that a grandoffspring’s mortality hazard rose roughly 4.5-fold in the two years following its grandmother’s death relative to having a living grandmother, and that this hazard rose further, to roughly 6.7-fold, when the grandmother who died had herself already been postreproductive [10]. Postreproductive grandmothers, by their estimate, conferred about 1.5 times the survival benefit of still-reproducing grandmothers, independent of how scarce salmon happened to be that year [10] — a direct quantitative echo of Hawkes’s argument that stopping reproduction, not merely surviving, is what makes a grandmother’s help most valuable.
None of this proves the human mechanism is grandmothering rather than something else; killer whales do not dig tubers, and the ecological logic connecting menopause to matrilineal social structure in a highly social, long-lived, food-limited predator is its own argument, not a transplant of the human one. What the whale evidence does responsibly establish is that post-reproductive lifespan is not a fluke restricted to one primate lineage with unusual medicine and unusual culture. It has now evolved, independently, at least five separate times under conditions that share a specific structural feature — kin-directed help persisting after fertility ends, inside a social group where a female’s older relatives remain contactable — and that convergence is evidence for the general class of “grandmothering-style” explanation even where it says nothing definitive about any one hypothesis’s mechanism.
The reproductive-conflict alternative complicates even that generalization, because it predicts the same rarity from a completely different direction. Cant and Johnstone’s model treats menopause not as a subsidy anyone is providing but as the resolution of competition between two overlapping generations of breeding females [6]. In their framework, when younger females disperse into a group and breed alongside an older, related female, the relatedness asymmetry between the two — the older female shares fewer genes, in expectation, with the younger female’s offspring than the younger female does with her own — gives the younger female’s reproduction a decisive competitive advantage; the model computes the older female’s relative payoff from continuing to breed versus helping as roughly a quarter of the younger female’s equivalent payoff [6]. Applied to humans, where wives typically move to live near their husbands’ kin rather than the reverse, this predicts that a woman’s fertility should decline sharply around the age her sons’ wives begin reproducing — and Cant and Johnstone report that human ovarian follicle loss accelerates sharply around age 38, matching that prediction, while cross-species comparison shows humans have essentially zero average reproductive overlap between generations of related females, against 0.71 in Japanese macaques, 0.52 in orangutans and 0.39 in chimpanzees [6]. Hawkes and Coxworth’s rebuttal targets the model’s load-bearing assumption directly: the claim that ancestral human groups were reliably patrilocal, with wives dispersing into husbands’ natal groups, has, in their reading, “now been repeatedly falsified by evidence indicating high mobility and regularly changing group composition” among contemporary hunter-gatherers [5]. Cant and Johnstone’s mechanism needs the dispersal pattern to be reliable in one direction; Hawkes and Coxworth say the ethnographic record does not show that reliability. Neither side has produced a decisive resolution of that specific disagreement.
Embodied capital: the rival that treats childhood as an investment, not a gift
Kaplan, Hill, Lancaster and Hurtado’s embodied-capital theory does not deny that grandmothers help. It denies that grandmothering is the right place to locate the causal engine of the whole life-history package, and it starts from a different observation: adult men in the same three forager populations Kaplan and colleagues measured directly — the Ache, the Hiwi and the Hadza — produce far more food than any other age-sex category, with production continuing to climb through the twenties, thirties and into the forties before declining, and with a substantial surplus of production over consumption sustained through much of middle adulthood [3]. That surplus, on their account, is the actual subsidy funding the long human childhood: by age 15, children in their forager sample “had consumed over 25% of their expected lifetime energy consumption but had acquired less than 5% of their lifetime energy acquisition” [3], a gap closed by parental — and specifically, on their reading, largely paternal — provisioning rather than by grandmothers’ foraging.
The theory’s causal chain runs the opposite direction from the grandmother hypothesis’s. Kaplan and colleagues argue that a shift toward calorie-dense, difficult-to-acquire food resources — large game requiring hunting skill, tubers requiring extraction skill — created returns to learning that could only be captured by a long juvenile period spent acquiring that skill rather than immediately foraging for oneself; that long juvenile period could only be afforded if someone else was subsidizing the juvenile’s energy needs in the meantime; and once that subsidy structure existed, it created selection for a longer adult lifespan generally, because a longer productive life meant a longer window over which the investment in skill acquisition paid off, and because a longer life let both sexes remain productive providers rather than dependents for a larger share of their lifespan [3]. On this account, long postreproductive life in women is not itself the selected trait; it is downstream of the same selection pressure that lengthened the entire adult lifespan of both sexes, women’s included, once the human dietary and learning niche made a long life profitable to have. The theory’s authors are explicit that their own proposal is “the first to do so with a specific model of natural selection that unifies the evolution of life history, brain and intelligence, diet, and age profiles of food production and consumption” [3] — a claim of broader scope than Hawkes and colleagues make for the grandmother hypothesis, and one of the reasons the two theories talk past each other as often as they engage directly.
Jocelyn Peccei’s critique cuts a different way again, targeting an ambiguity inside the grandmother hypothesis’s own history rather than defending embodied capital specifically. She distinguishes what she calls the “old” grandmother hypothesis — the claim that menopause itself, the early curtailment of fertility, is the adaptation, engineered to free a woman for grandmothering before her body would otherwise stop reproducing — from the “new” version Hawkes and colleagues actually argue in 1998, in which grandmothering is what drove the evolution of a longer total lifespan, and early menopause is simply left over from an ancestral ovarian-aging schedule that a lengthening body did not also extend [7]. Peccei’s own preferred account, sometimes called the extended-mothering hypothesis, holds that selection continued to favor investment in a woman’s own already-born children for longer than the “old” hypothesis assumed, and that the postreproductive period is better understood as an extension of mothering’s payoff period than as a grandmothering-specific adaptation; she also proposes, as a complementary rather than competing mechanism, that reduced paternal investment as men’s mates aged — through male mortality and through some men’s tendency to leave aging partners for younger ones — added an additional selective cost to continued childbearing at older ages. Both of Peccei’s proposals share a feature with embodied-capital theory that the grandmother hypothesis does not: they locate the selective force in a woman’s relationship to her own children rather than in her relationship to her daughter’s children, which is a genuinely different claim about whose fitness the trait evolved to serve, not a rhetorical variant of the same idea.
None of the three positions is short of counter-evidence against the others. The embodied-capital account has to explain why grandmother foraging effort tracks weaned grandchildren’s growth specifically in the Hadza data, rather than being incidental; the grandmother hypothesis has to explain why male provisioning rates are, in Kaplan and colleagues’ own numbers, so much larger in absolute terms than the food grandmothers provide; and Peccei’s critique has to explain why the postmenopausal-provisioning pattern recurs, independently, in killer whales, where there is no equivalent human-style male-provisioning economy to compete with it. My reading is that these are not fully substitutable theories competing for one open slot — they are competing for different shares of one energy budget, and the honest position is that each has real data behind part of it.
Finnish and Canadian church books: what a natural experiment in grandmothering actually shows
Ethnographic time-allocation data and formal simulation models both have a shared weakness: they cannot directly show that a grandmother’s survival changed her descendants’ fitness, only that her foraging behavior correlates with outcomes that plausibly matter. Lahdenperä, Lummaa, Helle, Tremblay and Russell’s 2004 study closes that gap with a genuine natural experiment, built from complete multi-generational demographic registers rather than observational fieldwork: Lutheran church records from Finland covering 537 women born between 1702 and 1823, and Catholic parish and civil registers from the Saguenay region of Quebec covering 3,290 women born in the region between 1850 and 1879 [8].
Restricting the analysis to women who lived past menopause — 339 in the Finnish sample, 2,362 in the Canadian one — the paper finds that the length of a woman’s postreproductive lifespan predicts the number of grandchildren she has, and that the relationship’s strength is nearly identical across the two very different populations: surviving ten additional years past age 50 is associated with roughly two additional grandchildren, in both countries [8]. The Finnish table alone shows the scale of the underlying numbers: women averaged 6.8 children born and 11.3 grandchildren; Canadian women, in a higher-fertility population, averaged 9.1 children and 38.2 grandchildren [8]. Mean lifespan for women who reached postreproductive age was 68 years in Finland and 74 in Canada [8].
The paper’s strongest methodological move is checking whether the correlation could simply reflect that women who happened to have easier lives lived longer and also had more successful children, rather than the grandmother’s presence causing anything. Three separate checks argue against that confound. First, offspring who had a living post-reproductive mother began reproducing 2.4 years earlier, on average, than offspring whose mother had died, controlling for the mother’s age or potential age if she had died [8] — a difference in the offspring’s own life-history timing that a general “good genes” story does not obviously predict. Second, offspring living in the same village as their post-reproductive mother produced more grandchildren than offspring of the same women who had moved more than twenty kilometers away [8], which controls for the mother’s own quality while varying only proximity. Third, and most directly, a grandmother’s cause of death mattered: dying of an infectious disease, which the authors treat as effectively random with respect to the mother’s underlying quality, was unrelated to either her offspring’s probability of dying of the same cause or to her offspring’s fertility [8].
The grandchild-survival result carries the sharpest edge relevant to the “stopping reproduction matters” claim at the center of the grandmother hypothesis. Grandchildren’s odds of surviving to adulthood were about 12 percent higher when their grandmother was alive and under 60 at their birth, but only about 3 percent higher when she was alive but over 60 [8] — the benefit is concentrated in the years right after a woman’s own fertility has ended, not spread evenly across old age, which is close to what the grandmother hypothesis predicts and not obviously what a generic “wealthy, healthy family” confound would produce. The paper’s authors read their own results as supporting the idea that selection continues to favor delayed aging in women only until their own offspring have themselves finished reproducing, since mortality rates in their data begin accelerating from that point in a woman’s life onward [8].
The limits are real ones, not throat-clearing. Both study populations are agrarian, Christian, patrilineal-adjacent farming societies within the last four centuries, not foragers, and both show grandmother effects transmitted overwhelmingly through co-residence and proximity — conditions that may not generalize to more mobile forager-band social structures, including the ones the grandmother hypothesis was originally built to explain. The design also cannot separate a grandmother’s direct childcare contribution from her households’ general economic stability, her potential role in resource inheritance, or purely social effects of having an older female relative present, since the historical record does not contain the fine-grained foraging-effort measurements the Hadza studies do. What it establishes cleanly is the fitness correlation and, through the three confound checks, a serious case that some of that correlation is causal. What it cannot establish is which specific mechanism — Hawkes’s foraging subsidy, Kaplan’s general adult-productivity story, or something else entirely tied to inheritance and household economics in an agrarian society — is doing the causal work.
Three hypotheses, one contested trait, and a childhood that still needs explaining
None of the three positions in this article accounts for the whole pattern on its own, and treating any one of them as the settled answer would misstate the state of the field.
The grandmother hypothesis owns the ethnographic mechanism most precisely: the Hadza foraging-effort data are specific, the correlation between grandmother output and weaned grandchildren’s growth is directly measured rather than inferred, and Kim, Coxworth and Hawkes’s formal model shows the mechanism is at least mathematically sufficient to move a chimpanzee-like lifespan toward a human one under a defensible, if contestable, set of assumptions about the cost of male longevity [4]. What it does not own outright is the magnitude question: Kaplan and colleagues’ own energy-production data show adult male provisioning dwarfing grandmothers’ foraging output in absolute caloric terms in the same forager populations [3], and Cant and Johnstone’s reproductive-conflict model reproduces the timing of human menopause from an entirely different mechanism, using real cross-primate comparative data on reproductive overlap that the grandmother hypothesis does not, on its own, need to explain [6].
Embodied-capital theory owns the energy-budget arithmetic and the male-provisioning data most directly, and it is the only one of the three that offers a unified account of why brains, diet, and lifespan should have coevolved together rather than treating the lifespan extension as a side effect of one specific kin-helping behavior [3]. What it does not own is a compelling account of why the Hadza correlation between grandmother foraging and grandchild growth exists at all if male provisioning is doing the real work, and it has less to say about why menopause specifically, rather than a gradual decline in fertility matching the gradual decline in other physical capacities, is the pattern actually observed.
The reproductive-conflict and toothed-whale evidence together arbitrate a narrower but genuinely settled question: post-reproductive lifespan is not a fluke of human medicine or culture, because it has evolved independently at least four times in a taxonomically distant lineage, under a directional signature — extended total lifespan without extended reproductive lifespan, concentrated benefit to grandoffspring, a survival cost when a postreproductive matriarch dies — that matches the shape, though not the specific mechanism, of what human data show [11, 10]. That convergence is real evidence that “grandmothering-style” kin investment after fertility ends is a viable evolutionary strategy in a highly social, long-lived mammal, which is more than any single terrestrial primate dataset could show on its own. It does not tell us whether the human version runs through tuber digging, general household productivity, reproductive-conflict avoidance, or some combination the field has not yet cleanly separated.
What would change this assessment is specifiable rather than rhetorical. If future work using genetic-relatedness and paternity data from additional forager populations found that grandmother foraging effort predicts grandchild survival independent of household wealth and male provisioning levels — replicating the Hadza pattern outside the Hadza — that would meaningfully strengthen the grandmother hypothesis’s claim to be more than a locally specific finding. If instead cross-cultural surveys found that societies with high male provisioning and low grandmother co-residence show life-history packages just as extended as societies with the reverse pattern, that would favor embodied-capital theory’s claim that the lifespan extension is general rather than grandmother-specific. And if the ancestral patrilocality assumption underneath Cant and Johnstone’s model were either confirmed or conclusively falsified by future paleogenetic or isotopic residence-pattern evidence from Pleistocene populations, rather than argued from modern hunter-gatherer analogues on both sides as it currently is, that would resolve the one clean empirical disagreement identified above rather than leaving it as a standoff between two readings of the ethnographic record. None of those tests has been run in a way that settles the matter, which is the honest reason this remains, in Hawkes and Coxworth’s own restrained framing, an active research question rather than a solved one — and why a trait this central to what makes human life histories distinctive is still being argued over sample by sample, register by register, four decades after the Hadza data first made it strange.
