Primates and the Evolution of Long-Slow Life Histories

Summary

Primates are characterized by relatively late ages at first reproduction, long lives and low fertility. Together, these traits define a life-history of reduced reproductive effort. Understanding the optimal allocation of reproductive effort, and specifically reduced reproductive effort, has been one of the key problems motivating the development of life history theory. Because of their unusual constellation of life-history traits, primates play an important role in the continued development of life history theory. In this review, I present the evidence for the reduced reproductive effort life histories of primates and discuss the ways that such life-history tactics are understood in contemporary theory. Such tactics are particularly consistent with the predictions of stochastic demographic models, suggesting a key role for environmental variability in the evolution of primate life histories. The tendency for primates to specialize in high-quality, high-variability food items may make them particularly susceptible to environmental variability and explain their low reproductive-effort tactics. I discuss recent applications of life history theory to human evolution and emphasize the continuity between models used to explain peculiarities of human reproduction and senescence with the long, slow life histories of primates more generally.

Introduction

Explaining the great diversity of forms and lifestyles is the central goal of evolutionary biology. Natural selection is the primary force behind adaptive diversification, and selection favors those that persist and increase over time. Evolutionary success is ultimately founded on two fundamental demographic processes: first, surviving to an age at which reproduction is possible (‘recruitment’), and second, reproducing successfully. Life-history theory seeks to explain the diversity of tactics through which different organisms achieve evolutionary persistence and increase, why the tempo and mode of reproduction can vary so much across taxa, and why life cycles vary from species to species. R. A. Fisher[1] defined the modern study of life-history theory saying that: “it would be instructive to know…what circumstances in the life-history and the environment would render profitable the diversion of a greater or lesser share of the available resources toward reproduction”. In an influential essay, Stephen Stearns [2] suggested that studies of life-history phenomena “naturally elicit a research viewpoint that combines the study of reproduction, growth, and genetics in an ecological setting to produce hypotheses concerning evolutionary changes.” As successful reproduction is immediately proximate to fitness, life-history theory lies at the heart of any understanding of adaptation in evolutionary biology.

Primates are mammals and, as such, are not characterized by particularly exotic life cycles: like other mammals, they grow until some age of maturity, when they cease growth, and begin their reproductive lives, dedicating the energy used as immatures for growth to reproduction [3]. They reproduce sexually and retain their original sex throughout life. These features limit the range of possible life-history tactics, but still leave plenty of room for variation. A major axis of such variation was identified by Dobzshansky [4] and later MacArthur and Wilson [5] as the speed and productivity of the life cycle. At one extreme — a tactic pursued by many rodents and lagomorphs — an organism lives for a short time and breeds extensively, producing an abundance of low-quality offspring. The probability of recruitment success of any one of these offspring may be low, but the sheer number of offspring produced makes it likely that at least two can be recruited — the average number of recruits a diploid organism must place in the next generation to persist. At the other end of this spectrum lie primates, along with dolphins, whales and elephants. They are characterized by long lives, modest reproductive rates, and extensive parental care. primates are grow more slowly, have later ages at first reproduction, longer life spans and lower fertility than most other mammals. The nature of primate life histories has been extensively reviewed [6–11]. Rather than covering the well-worn ground of these previous reviews, I will focus here on how primate life histories are understood by contemporary life-history theory.

Fitness is determined by survival and reproduction, so we naïvely might expect organisms to always maximize both. However, organisms are energy-limited, and this prevents the simultaneous maximization of both components. Thus, trade-offs are a central feature of life-history theory. An organism can use a given endowment of energy for current metabolic expenditures or invest in its growth, its survival or its reproduction. Energy dedicated to one of these tasks cannot be used for another, so organisms necessarily face trade-offs. For example, a large investment in survival will compromise an organism’s ability to produce bountiful offspring. As a result, we tend to see groups with common syndromes or suites of life-history traits.

A key concept for understanding different solutions to the problem of persistence and increase is that of ‘reproductive investment.’ An investment represents the diversion of energy from immediate use to some other fitness-related end. ‘Reproductive effort’ (RE) refers to the proportion of an individual’s total possible energy invested in a time period that is devoted to reproduction. An individual who engages in 100% reproductive effort, reproduces suicidally, leaving no investment for its survival. It is reproductive effort to which Fisher [1] alluded and this concept has continued to dominate thinking in modern life-history theory [12].

Life-history theory uses optimality models as a framework for understanding different life-history tactics [12–14]. An optimality model requires the researcher to specify several things [15]: first, an objective function to be maximized; second, a set of constraints that define what is biologically feasible; and third, a strategy set that lays out the possible alternative tactics. Specifying these requirements defines both the theoretical and empirical tasks of life-history theory. Fitness increases with all fitness components – i.e., age-specific fertility rates and survival probabilities – but increases more rapidly with some than with others. Trade-offs arising from the finiteness of energy create the opportunity for different solutions to the problem of ensuring evolutionary persistence and relative increase. To understand the evolution of life-history tactics, optimality models are used to find the maximum difference between the fitness benefits and costs of a tactic.[AU: ref 16 missing either insert elsewhere or delete and renumber]

Perhaps the canonical optimality model was developed by Gadgil and Bossert [13], who first noted that age-specific fertility and survival are both likely to be functions of reproductive effort and therefore trade-off. High effort leads to high fertility, but a cost is paid in reduced survival probability. Using a simple graphical model, they showed that intermediate reproductive effort can only be favored if there are diminishing marginal benefits and that there are either increasing or constant marginal costs with increasing effort (Figure 1). Only under these conditions can the greatest difference between costs and benefits lie anywhere but at the extremes (i.e., RE= 0 or RE=1).

Figure 1.

Models of reproductive effort following the classical analysis of Gadgil & Bosse [13]. Optimal RE is indicated by the vertical grey line. The left panel shows a situation (such as during juvenile or subadult stage) where the costs of any reproductive effort exceed the corresponding benefits and the predicted level of RE is zero. In the central panel, benefits are concave with effort (i.e., show diminishing marginal benefits) while costs are convex. This leads to an intermediate optimum level of RE. In the left panel, both benefits and costs are concave and the benefits always exceed the costs. The difference is greatest at maximum effort, so the optimal effort is suicidal reproduction, what Gadgil & Bossert call ‘big bang’ reproduction. The only possibility for intermediate effort occurs when benefits show diminishing marginal returns to effort and costs are either marginally increasing or linear with effort.

Understanding how different models yield differing predictions requires first knowing the objective function used in the optimization problem. In evolutionary optimization problems, the objective function is ultimately fitness, but fitness can be approximated in different ways. All models used in life-history theory are ultimately representations of demographic processes of birth and death and they differ primarily in the assumptions they make about the distribution of demographic events, the relevant time scale over which they are measured, and whether or not demographic rates are constant or vary probabilistically.

The objective function that arises from these demographic models is the rate of increase (i.e., of the individual, genotype, group). The two major axes of demographic models used in life history analysis are: scalar vs. structured and deterministic vs. stochastic. A scalar life history is one in which fertility rates and survival probabilities do not vary by age or stage, while a structured model allows demographic rates to vary. The long life spans and extensive iteroparity (i.e., repeated breeding) of primates mean that their life histories can really only be understood using structured models. A deterministic model is one where demographic rates are assumed to maintain constant mean values, while a stochastic model allows rates to vary probabilistically between time periods. Given the long lives and the dietary requirements that make primates susceptible to variability in food availability [16], stochastic models probably better explain primate life histories. The Euler-Lotka equation is the basic model used in life-history studies of age-structured populations (Box 1). While deterministic, this model is the foundation for understanding structured stochastic models and I will briefly highlight one extension of the model to variable demographic rates that holds particular relevance for understanding primate life histories.

Box 1.

The most important demographic model in life-history theory is the Euler-Lotka equation, which implicitly defines the instantaneous rate of increase for a population with a fixed age at first reproduction, and age-specific schedules of mortality and fertility. The Euler-Lotka equation is an integral equation defined by:

$$1={\int }_{\alpha }^{\beta }{e}^{-rx}l(x)m(x)dx,$$ (1)

where α is age at first reproduction, β is age at last reproduction, l(x) is the fraction of all live births surviving to exact age x, m(x) is the fertility rate at age x, and r is the intrinsic rate of increase. The age-specific product φ(x) = l (x)m (x), is known in the demographic literature as the ‘net maternity rate,’ emphasizing the fact that survival to reproductive age is a necessary predicate for fitness. The sum of the age-specific net maternities is the ‘net reproduction ratio’ and is the population average lifetime reproductive success (i.e., the number of live-born offspring to an individual), $R_0={\int }_{\alpha }^{\beta }l(x)m(x)dx$. A key observation is that the formula for the net reproduction ratio is simply the Euler-Lotka equation when r=0. This implies that when r=0, R₀=1. When r=0, a population neither increases nor decreases and is referred to as ‘stationary.’ This point is of fundamental importance since it means that any time lifetime reproductive success is used as an objective function to be maximized in a life-history analysis, a (sometimes hidden) assumption is that the population is stationary – that is, deaths exactly balance births. This is a problematic assumption because, among other things, it implies that the timing of reproduction has no bearing on fitness. Models that assume population stationarity often make qualitatively different predictions than those that all population to grow (or decline).

The Euler-Lotka equation applies to deterministic cases. Extension of the model to cases where vital rates vary probabilistically are straightforward if somewhat more complex. The probabilistic nature of the viral rates can arise either because the environment is inherently variable and this variability affects vital rates or because of a small population where the realized birth and death rates are subject to sampling variability (akin to genetic drift). Mean fitness (r) as given by equation 1 is a rate of increase. In principle these models can be solved for a ‘stochatic growth rate,’ which is the expected long-run growth rate for a stochastic model analogous to r in equation 1. Such analytic solutions can be difficult but they frequently yield useful approximations to the long-run rate of increase that provide important insights about life-history evolution. Tuljapurkar’s ‘small noise’ approximation for the stochastic growth rate is particularly applicable to primate life histories. Assuming a random environment in which variability primarily affects juvenile survival, Tuljapurkar [28] showed that the appropriate fitness measure is

$$a=\overline{r}-\frac{c^2}{2T_0^2},$$ (2)

where r̄ is the fitness of the mean life history (from equation 1), c² is the coefficient of variation on juvenile survival, and T₀ is the mean age of reproduction (i.e., the generation length).

Figure B1.

Graphical demonstration of the Euler-Lotka equation (equation 1) using demographic data for Aché hunter-gatherers [50]. The left-hand panel plots age-specific survival, l(x). The central panel plots age-specific fertility, m(x), of female births (assumed to be half the recorded births reported in [50]). The right-hand panel plots e^−rxl(x)m(x), the product of age-specific survival age-specific fertility, and the age-discounting term of the Euler-Lotka equation. The area under this curve sums to unity, as indicated in equation 1.

In this review, I will discuss the unusual life-history traits of primates and show how they are understood by contemporary life-history theory, singling out one popular life-history model. I will pay special attention to humans—the best-studied primate and the one most difficult for theorists to understand. Throughout the paper, I will emphasize how the peculiarities of primates provide a real opportunity to further life-history theory in general.

Primates and the Mystery of Long, Slow Life Histories

Charnov and Berrigan [3] asked the central question for understanding primate life histories: “why do female primates live so long and have so few babies?” For their body size, primates mature later, live longer and have lower fertility than most other mammals [17–22]. While there are certainly other mammals with a late age at first reproduction, long life spans, and low fertility relative to their body mass, primates as an order are consistently at the extreme of the bivariate distributions (Figure 2).

Figure 2.

Scaling relationships between adult female body mass in mammals and three fundamental life-history variables: (a) age at first reproduction (AFR), (b) maximum life span, and (c) annual fertility and. All are plotted on double-logarithmic axes. Primate points are drawn in red, while other non-volant taxa are drawn in black. Bats, another long-lived order despite its small average size, are colored in blue. Data for non-volant mammals (including primates) from [22] and bats from the An Age database [132].

The key question is what favors delayed maturity and low fertility? Early work in life-history theory suggested that age at first reproduction is a crucial life history variable and that, all things being equal, earlier maturity is better for fitness [23,24]. Furthermore, intuition suggests that high fertility should be better for fitness than low fertility. These two observations make the pattern of low fertility and late maturity in primates paradoxical. Both late maturity and low fertility indicate life-history tactics with a low reproductive effort. This specific question about why we observe reduced reproductive effort tactics in primates is a special case of the fundamental question in life-history theory, that of the optimal allocation of reproductive effort [12,13, 24].

The consensus from a variety of life history modeling approaches is that delayed reproduction and low fertility can most readily be seen as adaptations to juvenile recruitment uncertainty. Three classic works in life-history theory support this interpretation: first, the seminal paper of Gadgil and Bossert [13] showed that reduced reproductive effort can only evolve when the shapes of the cost and benefit functions with respect to reproductive effort allow intermediate reproductive effort optima; second, Schaffer [25] showed that variability in recruitment success favors reduced reproductive effort; third, resolving the famous paradox of Cole [24] — repeated breeding is ubiquitous in nature despite the ease with which a single, suicidal bout of reproduction can increase fitness — Charnov and Schaffer [26] showed that when the external mortality rate of juveniles exceeded that of adults, reduced reproductive effort is favored.

Low fertility can be seen as an adaptation to uncertainty in juvenile recruitment because of trade-offs. By holding back on reproductive effort, mothers are able to further invest in their own survival and reap the benefits of a longer reproductive span. That is, they reduce the effort at any given age to ensure more reproductive events overall. Any given year might be bad for reproduction, but by having more (and more dispersed) reproductive events, primates seek to ‘get lucky’ in finding a few good years. Primates do indeed live longer than most other mammals both absolutely and for their body mass (Figure 1) [21]. Low fertility can thus be seen as an adaptation to uncertainty or variability in juvenile recruitment.

But how does delayed reproduction fit in? In deterministic environments, the classical theory of life histories predicts that early reproduction is highly favored [23]. Solving the Euler-Lotka equation (Box 1)for ∂r/∂α = 0 to find the optimal age at first reproduction yields an optimal value at α = 0, i.e at birth. Obviously, biological realities make this solution an absurdity, but the fact remains that we expect selection to push α to be the lowest it can possibly be, all things equal. There are two primary reasons why delayed reproduction could be favored: first, survival and/or fertility are frequently size-dependent and it takes time to grow [27]; and second, delaying maturity increases generation time which permits individuals to average over a longer time frame in stochastic environments [28].

Features of primate ecology make this latter approach a promising line of theoretical reasoning about delayed age at first reproduction in primates [28,29]. Tuljapurkar’s small-noise approximation to the stochastic growth rate is an appropriate fitness measure in a variable environment (Box1) [28]. As is clear from equation 2, when the coefficient of variation in juvenile survival is greater than zero (i.e., when there is variability in survival), realized fitness will be less than the fitness of the mean life history (i.e., the value of r that would arise by using equation 1 with the population averages for mortality and fertility schedules). The difference between these is mediated by generation length. When generations are long, the reduction of mean fitness due to variability is attenuated. This leads to the expectation that when variability affects juvenile survival substantially more than it does adult survival, delayed reproduction is favored, a result particularly favored if age at first reproduction is causally linked to longer reproductive life span as suggested by some authors [14,30]. This mathematical abstraction is backed by a very straightforward intuition: longer generation times mean that individuals sample a longer range of temporal environments with their reproductive events. If the environment is highly variable in its suitability for juvenile recruitment, temporally sampling more environments makes it more likely that an individual will indeed ‘get lucky’ and reproduce during propitious periods for offspring survival.

Is there a common ecological feature (or at least ancestral feature) of primates that might make them susceptible to especially high levels of environmental variance? Primates are primarily frugivores and those with the longest and slowest life histories are specialists in ripe fruit. Fruiting phenology in the tropics is remarkably variable, often stemming from life-history strategies of fruiting plants themselves [31–33]. Peaks of fruit abundance can be unpredictable and periods of fruit scarcity can be common, leading to substantial energetic shortfalls for primates and other frugivores [34–36]. Recent work from a variety of long-term primate demography projects strongly suggests that adult survival in primates exhibits lower variance that that of other mammals, potentially owing to the behavioral, social and dietary flexibility arising from primates’ cognitive abilities[37]. In contrast, Janson and van Schaik [38] argue that frugivorous juvenile primates often have lower foraging success due to scramble competition inherent in foraging for fruit in groups [39] and general inefficiency due to small size, a point further emphasized for human foragers [40]. Supporting this idea, folivorous primates have faster growth rates than frugivores [41]. Similarly, the life-histories of folivorous Thomas langurs are faster than those of sympatric frugivores [42]. Chimpanzees, who are obligate ripe-fruit specialists, have later age at first reproduction and lower fertility than sympatric gorillas [43–45]. Orangutans, who live in the especially variable mast-fruiting forests of Borneo and Sumatra, have even lower fertility than chimpanzees, indeed the lowest fertility of any mammal [46,47]. These considerations suggest that the model of equation 2 (Box1) may be particularly applicable to the evolution of primate life histories.

Extreme environmental variation and frequent population crashes raise the possibility that primate life histories are generally adapted to non-equilibrium ecological conditions [48,49], a suggestion also put forward for human hunter gatherers [50]. This is a rather different interpretation of the match between primate life histories and ecology than the traditional model for understanding the evolution of fast as opposed to slow life histories [5] But it certainly fits the facts of tropical phenology and feeding competition better than the vague notions of the tropics being ‘constant’ from early work on life-history evolution [4]. When a species spends more of its time on average in decline (which can be true even in populations where the average long-run growth rate is positive [28]), long/slow life histories are favored. By Contrast, in a species that spends more time in growth, precocity and high productivity will be favored [48]. An important implication of this is that models of primate life-histories that rely on equilibrium arguments could be wrong either quantitatively or, more distressingly, qualitatively.

Senescence and the Evolution of Long Lives

Senescence is conventionally defined as the decline in physiological function with age. Senescence arises because the force of selection declines with age [51–53]. This explanation is consistent with explanations that are both adaptive — senescence because of trade-offs between genes with beneficial early effects and deleterious late effects — or non-adaptive ones— deleterious mutations accumulating with age as the force of purging selection approaches zero (Figure 2A)[54]. Operationally, senescence is often measured either by life expectancy or maximum recorded age. Primates generally live longer than other similar-sized mammals(Figure 1B) [21], suggesting that selection has slowed the rate of senescence in this lineage because of putative fitness gains associated with longer lifespan.

Williams [52] first suggested the hypothesis that lower ‘external’ mortality (i.e., mortality due to predation, starvation, etc.) will lead to lower rates of senescence. Senescence is indeed an evolutionarily labile trait in primates and other mammals [55], responding to the selective pressures of specific ecologies. For example, opossums living on predator-free islands indeed have lower rates of senescence [56], while life span is longer than expected in flying and gliding mammals and birds live much longer for their body size than mammals [57–59]. These examples suggest that species that can escape high external mortality by evolving the ability to fly, glide or by living in predator-free environments have longer life spans. Primates are, for the most part, arboreal and may reduce mortality by escaping terrestrial predators. Recent phylogenetic analysis of the longevity of mammals supports the general prediction that arboreal species have greater longevity [19]. However, within primates, arboreal taxa don’t live longer than terrestrial ones, a result attributed to the long history of arboreality in the primate lineage [19].

Despite its empirical successes, the Williams hypothesis [52] is not without critics [54]. Importantly for organisms living in non-equilibrium environments, Williams’ hypothesis depends on population stationarity but fails to account for density-dependence. The incorporation of density-dependence can change the predictions substantially [54]. However, if population regulation occurs primarily through juvenile mortality, the Williams hypothesis can hold. Even if juvenile mortality does not ensure population stationarity (as assumed in e.g., in [14]), it is more variable than adult mortality [37], something human demographers have known for a long time [60]. It is an open theoretical question if the Williams hypothesis applies to non-equilibrium, populations in stochastic environments where density-dependent regulation is not dominant. Given the premium placed on generation length by equation 2 (Box 1), we should expect substantial investment in adult mortality reduction in taxa such as primates in which recruitment is highly variable, leading to increased life span.

Humans have an unusually extended life span after the end of their reproductive phase. A variety of authors have argued that such female post-reproductive survival is a general mammalian trait [61,62]. However, its extent in humans is qualitatively different than for most other mammals. For example, in two hunter-gatherer groups, the Aché of Paraguay [50] and the !Kung of Botswana [63], women aged 45 (the conventional last age class with recorded fertility in much demographic work) have around 20 years to live. Thus, in these populations, the length of post-reproductive nearly equals that of reproductive life, and it even exceeds it in industrial populations.

There is confusion in much of the evolutionary literature on reproductive termination in humans. At least two phenomena are involved in human post-reproductive life span: reproductive cessation and post-reproductive survival. Human reproductive cessation is due to follicular atresia, the loss of oocytes throughout a woman’s life [61,64]. As chimpanzee reproductive senescence is broadly similar to that of humans [65,66], it seems that extensive post-reproductive life span is the phenomenon that requires explanation. The leading hypothesis for the evolution of post-reproductive life span — known as ‘the grandmother hypothesis’— argues that the subsidies older women provide their daughters and grandchildren provide a greater marginal benefit to their fitness than their own continued reproduction would. Lee [67] presents a model in which economic transfers, not births, shape patterns of senescence. This model generalizes the grandmother hypothesis and accounts for some characteristic features of mortality patterns that are not addressed by other models. In particular, this model predicts that mortality rates should decline initially and then increase steadily with age, giving rise to the ‘bathtub-shaped’ mortality pattern characteristic of humans [68]. Recently, demographic simulations have suggested that the probability of grandmothers surviving is not high enough to make grand maternal transfers a viable force for selection of post-reproductive life span [69], but the existence of more generalized inter-generational transfer as envisioned by Lee [67,70] may save the more general form of the grandmother hypothesis.

There is a great deal of interest in energetic restriction as a means of increasing life span, especially our own. More important from a theoretical perspective, understanding the evolution of primate life histories highlights the likely shortcomings of this approach to life-span extension. While energetic restriction has been shown to improve certain biomarkers of aging in captive primates [71], it is too early to assess the actual demographic impact of such interventions. The actual life span gain that arises from energetic restriction may depend on the steepness of the reaction norm between energy availability and life span[72]. Critically, this reaction norm for primates, and humans in particular, is shallow, so the benefits to life span are likely to be modest at best. This is due to primate life histories being already evolved towards reduced reproductive effort and extended life span. In effect, primates expect variability in environmental quality (and consequently energy availability) because of their selective diets, and their long, low-reproductive-effort life cycles already are adaptations for smoothing over variability in energy availability.

Pesky Primate Brains

Primates have large brains relative to their body size and thus brain size may place a constraint on, or even serve as a regulator of primate life history [73]. However, a comprehensive review of variation in life histories of strepsirhines (i.e., lemurs and lorises),] found no relationship between life-history variability and either brain size or metabolic rate [74]. Evidence suggests that there is a grade-shift between strepsirhines and haplorhines (i.e., tarsiers and anthropoid primates) in primate diversification in which relatively larger brains exert a real constraint on life-history tactics [75]. Like so many other problems in the analysis of life histories, the relationships between brain size, body size, and life-history traits are endogenous and hopelessly confounded. The general trend in current work seems to favor the hypothesis that large-brains cause slow life histories [76]. However, based on the theory of reduced reproductive effort, the alternative hypothesis cannot be rejected: primates, especially frugivorous ones, are selected for slow growth and their relatively large brains are the result, rather than the cause of the slow life-history [77,78].

The greater canalization of brain ontogeny has been suggested to lead to high levels of ‘encephalization’ in species that are selected for smaller size or slower growth [79]. The larger relative brain size emerges not because of selection on brains getting bigger but on ontogenetic lag in brain size as bodies get small. This argument is supported by selection experiments with rodents [78, 80, 81]. Furthermore, bats, which have extremely long life spans for their size (Figure 1B), have relatively small brains [76]. Similarly, highly non-brainy squamate reptiles have substantially longer life spans for their body size than mammals [59]. Is it possible the masterpiece of human higher-intelligence was, in fact, painted on the spandrel of delayed maturation that arose as a response to unpredictable food supplies? Answering this important question in primate life-history evolution will undoubtedly require evidence from a deeper investigation into the ontogeny of primate bodies and brains [82].

Human Life Histories — Allo-Maternal Investment Overcomes Constraints

Humans are paradoxical primates. For instance, despite the very long periods of juvenile dependency and late age at first reproduction (Figure 3), life-cycle traits typically associated with low fertility, humans have a much higher fertility than other great apes [18,45]. This highlights the fact that suites of life-history traits are not absolute — there is still considerable room for variation.

Figure 3.

Fitness elasticities (survival in left panel, fertility in right panel) of two long-lived nonhuman primates (baboons, chimpanzees) and Aché hunter-gatherers, plotted against relative age (i.e., scaled to maximum reproductive age). An elasticity represents the force of selection on a proportional scale. The sum of elasticities within a given life-cycle is unity, making an individual elasticity a measure of the total force of selection in the life-cycle on that particular transition, conditional on the other life-cycle transitions. Red ticks along the abscissa indicate the average ages of first reproduction for the three populations. Notable is the late average age of first reproduction of the Aché. Nonzero elasticities for ages younger than the average AFR mean that (1) reproduction (naturally) does occur before the average and (2) that selection could push AFR early. The fact that AFR is so late suggests that it is constrained through negative covariances with other, higher-elasticity traits [133]. Baboon demographic data from [134]; chimpanzee demographic data a composite from [65, 135].

The seemingly paradoxical high fertility of long-lived humans relates to another major thread in life-history theory. Saether et al. [83] present a useful classification of life-history tactics based on how a species’ ecology affects the most basic inputs for fitness, survival and reproduction (Table 1). The diagonal of this table corresponds to the standard fast–slow continuum, with high mortality paired with high fertility and low mortality with low fertility. The viable off-diagonal element (high survival–high fertility) is known as ‘bet-hedging’. In the context of reproductive tactics, bet-hedging is defined as a risk-management strategy in which an individual attempts multiple breeding bouts with relatively large clutch size. This is a common strategy in raptors, for example, where large fluctuations in food supply lead to high variance in recruitment success. By having clutches of more than one chick, although both rarely survive, a variety of bird species hedge against recruitment variance by having two opportunities to succeed and potentially having two chicks recruit if conditions are exceptionally good [83].

Table 1.

Modification of [85]. Bet-hedging species combine high fertility with high adult survival when recruitment success is strongly limiting. Humans are a relatively high-fertility species, especially when compared to other hominoids, despite a suite of life-history traits that otherwise seem classically long/slow. This reproductive strategy is hedged even further by staggering the developmental states offspring at any given point in time, a feat carried out with the apparently uniquely human tactic of extensively overlapping the dependent stages of successive offspring.

		Survival
		Low	High
Fertility	High	High-reproduction species	Bet-hedging
	Low	Extinct	Low-reproduction species

In this sense, human reproductive patterns can be seen as a type of bet-hedging. Human natural fertility and age-specific survival are substantially higher than those of the great apes [45,84–86], and while we typically bear singletons, reproductive bouts overlap substantially. Humans manage this feat by having extensively overlapping periods of juvenile dependency [87]. By contrast, a chimpanzee mother will invest heavily in her offspring until it is weaned, and once this is accomplished, she is done with direct energetic investment. Given the state of extended dependency of their children, this is not an option for human mothers [88]. A variety of authors have shown that children are a net economic liability in subsistence societies until they are in their late teens to early twenties [71,89,90]. This pattern of overlapping periods of dependency makes human reproduction doubly hedged. On the one hand, human mothers have relatively high fertility where an individual woman will have a ‘clutch’ of multiple dependent offspring simultaneously [91]. On the other hand, this clutch consists of offspring of different ages, vulnerabilities and lower parental investment. When conditions are favorable, this can lead to explosive population growth — a potential that has been realized in recent history [92].

The peculiarities of the human life cycle, the extremely late age at first reproduction and the relative helplessness of our infants and juveniles, are frequently explained in terms of allo-maternal investment, the investment in offspring by individuals other than the mother. Various authors have shown the importance of older siblings on the survival of children or the fertility of mothers [91,92,93]: male provisioning may allow women to have high fertility despite the dependency of their offspring[86]. For example, Hadza men provision their wives dependent on their reproductive state [93]. Post-reproductive women subsidize their daughters energy budgets, thereby increasing their fertility and, possibly, their grandchildren’s survival [30]. All of this argues that humans could be viewed as cooperative breeders [94,95]. Cooperative breeding provides humans tremendous reproductive potential not seen in other primates [95], which, when combined with the mutli-layered bet-hedging reproductive tactic discussed above, has led humans to dominate nearly all terrestrial environments [96]. The necessity of cooperative breeding could even explain the very low fertility seen in contemporary cities, where traditional women’s support networks are frequently broken down [97], providing theoretical leverage for the evolutionarily vexing problem of demographic transitions in contemporary (and recent historical) human populations [98].

The human life-history complex, with its late age at first reproduction, long reproductive span, and overlapping periods of juvenile dependency is illustrated schematically in figure 3. The proposed mechanism leading to such slowing and elongation of the human life-cycle is the low-constancy/low-contingency [94] variation in energy supplies that arises from preferential foraging on energy-dense resources in tropical environments.

Charnov’s Model and Primate Life Histories

The comprehensive life-history model for female mammals has been developed by Charnov [14] and has been applied in particular to understanding long/slow primate life-histories and to the peculiar events of human evolution. The model suggests that extrinsic adult mortality rates set the life history. Age at first reproduction (α) is optimized by selection to balance the trade-off between reproductive power (arising from body size) and recruitment success. At α, the instantaneous mortality rate and marginal fertility rate with respect to α are equal. Both these quantities are, in turn, given by a scaling relationship with body mass, since at α, all physiological power that was used for growth is channeled into reproduction. Population stationarity is maintained by density-dependent mortality of juveniles.

Charnov’s model fits fairly well demographic data collected from Aché hunter-gatherers — predicted optimal age at growth cessation (i.e., where marginal benefits for production of increased size just balance out the marginal survival costs of later α) fits observed ages [50]. Charnov’s model was also used as a framework for understanding growth in Hadza hunter-gatherers, though not formally tested [100], and for understanding the slower growth rates of human children compared to chimpanzees [18]. Furthermore, the model has been used as a means for arguing the importance of post-reproductive mothers in increasing their own daughters’ fertility [30], a classical explanation for human female reproductive senescence [52].

Perhaps more important than the specific applications of this model is the general adoption of its underlying logic. Specifically, Charnov [14] posited that adult mortality rates exclusively determine an organism’s life history. This logic is presented in comprehensive reviews and narrative scenarios of human life-history evolution [86,101–103]. Charnov’s later life-history model contradicts his earlier model [26] in which age-specific mortality of adults and juveniles jointly determine optimal reproductive effort.

A second feature central to Charnov’s model, and implicit in other, less formal life-history models, is the idea that metabolic power dedicated to growth or reproduction is equivalent. This is a particular instantiation of the fundamental notion that energy devoted to one type of reproductive investment (e.g., growth) cannot also be used for another investment (e.g., fertility). These energetic trade-off models assume a fixed energy pool from which investments are made. Kramer and colleagues [84,104] have suggested pooled energy models to help account for the puzzling fact that human fertility is much greater than expected from our somatic growth rates, which are relatively slow compared to other primates [19, 88].

Charnov’s model has some shortcomings. Mortality is not generally age- or size-independent. Relative weaning mass is not constant, but appears to scale with body mass [105]. Two critical theoretical problems also afflict the Charnov model: first, the assumption of constant relative weaning mass makes the critical trade-off between adult fertility and juvenile recruitment impossible [106]; second, the model assumes population stationarity, which is accomplished by density-dependent juvenile mortality. That is, since the life history is set exclusively by adult mortality rates, juvenile mortality must adjust itself to ensure population stationarity. It is problematic to relegate juvenile mortality to book keeping in this way when it is the part of the life-cycle under strongest selection (Figure 2)[107].

Moving Forward

In his important monograph, Stearns [108] lamented the growing disconnect between the mathematical sophistication of life-history theory and the paucity of data available to test these theories, singling out age-structured stochastic models in particular [109]. In fact, the data demands of age-structured life-history modes in stochastic environments are not unreasonable. Life-history theory is not string theory. However, such models do require measures of variability, both of vital rates and of environments. They also require long periods of observation. Rare events can have substantial impacts on fitness in long-lived organisms [110, 111]:humans, for instance, passed through at least one very severe bottleneck [112]. Similarly, studies of contemporary hunter-gatherers fail to reveal signs of rapid Pleistocene population expansion characteristic of nearly all other populations in the world [113], suggesting population crashes. Thus more long-term demographic studies of free-ranging primates and subsistence populations of humans are needed [37,66,114]. One of the many additional benefits of such long-term studies is the ability to discern individual heterogeneity in life-history traits.

Longitudinal studies should include careful measurement of features of the biophysical environment as well as vital rates as they change with the environment. Ideally this would be coupled with other phenotypic measurements, allowing for greater integration of demography and evolutionary genetics [117,118]. Technological advances make this increasingly possible, even with forest species [119]. There are several exemplary long-term projects that have monitored both demographic data and environmental conditions and these projects have yielded data that have had an enormous impact on our understanding of life history evolution in primates, and overall [37,55,66].

Life-history traits are hopelessly confounded and the direction of causality is often ambiguous. For example, do big brains cause slow life histories or vice versa? Randomized experiments are the gold standard for establishing causality, and experimental manipulations have yielded extremely convincing tests of life-history models in other, short-lived taxa [120–122]. However, in many primates, such experiments are both infeasible and unethical. Statistical techniques have been developed to allow the separation of causation from mere correlation [123], such as the use of instrumental variables. An instrumental variable is one that affects the dependent variable only through its effects on some potentially endogenous explanatory variable. The use of instrumental variables is increasingly common in economic anthropology and demography [124,125], but has yet to make an impact in primatology and evolutionary anthropology. As arboreality, dietary niche, and brain size are hopelessly confounded within primates, who are among the best studied tropical mammals, what is really needed are data on other tropical, arboreal mammals.

Like other complex phenotypic traits, the evolution of life-history traits must ultimately be understood in the context of quantitative genetics. The quantitative genetic theory of life-history evolution requires the calculation of the additive genetic covariance matrix, also known as the ‘G-matrix.’ The trade-offs specified in optimality models for life-histories are ultimately given by the negative genetic covariances between life-cycle transitions. Primatologists have been among the leaders in using quantitative genetics to study morphological development and phenotypic integration [126–128], so expanding this work to life histories should be straightforward. Some progress in measuring important quantitative genetic quantities for life-history traits has already been made [129–131].

Studies on primates have the potential not only to help us understand the life-history of these fascinating animals, but to improve life-history theory in general. The features of primate biology that have traditionally been seen an impediment to life-history studies — long lives, slow reproduction, complex sociality, arboreality— may ultimately prove to be virtues for understanding the most fundamental question of life-history theory, that of allocation of reproductive effort.

Figure 4.

Schematic illustration of the fundamental issues surrounding the human life history. Reliance on high-quality, energy-dense resources (e.g., ripe fruit, seeds, game). Such resources are subject to highly volatile, low-contingency availability, which selects for slow growth (to minimize starvation risk) and late age at maturity (to increase generation length), as indicated by the volatile time-series of energy availability. Long-lived humans adopt an unusual bet-hedging reproductive strategy by caring for multiple offspring with overlapping periods of dependency. Overlapping periods of dependency of a hypothetical woman’s eight births are shown at the bottom, with the result that a woman nursing an infant frequently must also care for one or more needy children.