Human uniqueness? Life history diversity among small-scale societies and chimpanzees

Background Humans life histories have been described as “slow”, patterned by slow growth, delayed maturity, and long life span. While it is known that human life history diverged from that of a recent common chimpanzee-human ancestor some ~4–8 mya, it is unclear how selection pressures led to these distinct traits. To provide insight, we compare wild chimpanzees and human subsistence societies in order to identify the age-specific vital rates that best explain fitness variation, selection pressures and species divergence. Methods We employ Life Table Response Experiments to quantify vital rate contributions to population growth rate differences. Although widespread in ecology, these methods have not been applied to human populations or to inform differences between humans and chimpanzees. We also estimate correlations between vital rate elasticities and life history traits to investigate differences in selection pressures and test several predictions based on life history theory. Results Chimpanzees’ earlier maturity and higher adult mortality drive species differences in population growth, whereas infant mortality and fertility variation explain differences between human populations. Human fitness is decoupled from longevity by postreproductive survival, while chimpanzees forfeit higher potential lifetime fertility due to adult mortality attrition. Infant survival is often lower among humans, but lost fitness is recouped via short birth spacing and high peak fertility, thereby reducing selection on infant survival. Lastly, longevity and delayed maturity reduce selection on child survival, but among humans, recruitment selection is unexpectedly highest in longer-lived populations, which are also faster-growing due to high fertility. Conclusion Humans differ from chimpanzees more because of delayed maturity and lower adult mortality than from differences in juvenile mortality or fertility. In both species, high child mortality reflects bet-hedging costs of quality/quantity tradeoffs borne by offspring, with high and variable child mortality likely regulating human population growth over evolutionary history. Positive correlations between survival and fertility among human subsistence populations leads to selection pressures in human subsistence societies that differ from those in modern populations undergoing demographic transition.

The authors are later more explicit when referring to LTRE where they are defined as the "vital rate contributions to observed differences in population growth rates" between two projection matrices (please note that vital rates are not individual measures as mentioned in l71 since there are population aggregates). In this sense, they are not "contribution to [a population] fitness" but how differences in entries of two matrices translate into difference in change of population reference growth rate. I strongly suggest the authors to define it more clearly. A way to do it is that sensitivity sij is the impact on λ of one unit of change in matrix entry aij. If we multiply sij by Δaij, it tells us how such a change would have modify the reference population growth λ. Second, the authors then states l74-75 that "Differences between realized fitness contributions and the potential suggested by elasticities may indicate constraints on life history evolution" (also [405][406]. This can be a fantastic idea and I can intuit what the authors have in mind. Yet it is not trivial to me, and it makes me wonder if this has been already theorized elsewhere. If it has, the author should clearly state it and explain why (I think not shying away equations). If it has not, I would strongly encourage the authors to developand if possible demonstrate -this idea. For instance, Cij, is a given amount of change between two matrix waited by sensitivities. Does this idea relates to the long lasting debate on the difference between using sensitivities and elasticities?

Main comment 2
The authors used the ratio between contribution and elasticity to measure (if I understand well) these possible constraints. But, I would strongly suggest the authors to check the resulting equation. First, l146, I think there is a mistake: eij is not equal to sij*(λ/aij) but to sij*(aij/λ) (I guess that this is a typo because elasticities look ok in fig 1) .
Then Z is the 'percentage' of difference between the reference and the analysed matrices divided by the growth rate. I am far on being clear on what does this mean and how this allows identifying constraint on a vital rate. I therefore strongly suggest the authors to explicit this metric and how/why it is used to solve their research question.
I am also not clear on whether Z should be sum(Cij)/sum(Eij) or rather sum(Cij/Eij), which can be substantially different.
Finally I don' t understand the values for the Zs in table 1. For instance for Ache, Zc=Cc/Ec=7/42=0.16, not 95. Or, am I missing something?
I would suggest to incorporate Table S5 into the main text.

Main comment 3
I am not sure that I understand prediction 1 and it may be there a conceptual mistake. Canalization is the fact that vital rates impacting the more fitness (here λ) should exhibit lower temporal variance than those under weaker selection. The authors rightfully quote [22] and [23] testing this by somehow correlating the estimation of the variation in time of matrix entries to the variation of λ (but variance is in time, not between populations, isn't it?).
Note also that, if I am not mistaken, [23] performed elasticity analysis not LTRE (as suggested in sentence l84) such that the effect of variance on LTRE is also not that clear to me. Anyway, I cannot see how LTRE between populations (without temporal variance accounted for) can allow identifying life-history constraint and how the concept of canalization is involved into this. If I am mistaken, I strongly suggest the authors to make their point more clear.

Main comment 4
I find that P1 (l81-82) is not well formulated. If I am not mistaken, it is a property of elasticity to be strictly declining with age in an age-structured model, infant and children survival elasticity always being constant and the largest. Metric have to be twisted and parameters very different that those of mammals to find alternative pattern (Baudish, 2005, PNAS). It is between species that relative magnitude of elasticities can be compared and I would strongly suggest to cite Heppell et al., 2000, Ecology for a comparison in mammals across the slow-fast continuum. Also why not refering to and using a classical Silvertown triangle to represented this (Silvertown, J., et al. 1993. Jouranl of Ecology 81:465-476)?
I am not sure what the authors want to test with prediction 2 which is the obvious fact that both increasing survival and fertility should increase population growth rate. Evidencing trade-off between fertility and survival?

Main comment 5
I would suggest the authors to discuss limitations of elasticities analyses in general and apply to humans in particular. (1) First elasticities are only one hand of the evolutionary GxE equation (Lande 1982;Charlesworth 1990;Steppan et al. 2002). Evolution also need genetic variance and this could be acknowledged.
(2) The authors are comparing Leslie matrix, but any sub structuring (as individual heterogeneity) or hidden trade-offs may change the results. (3) It the most important, it has been shown that intergenerational transfers between age-class or parental investment can strongly impacts elasticities on survival and fertility in humans (Lee 2003, PNAS, Pavard et al. 2007, Evolution, Pavard & Branger 2012. For instance, magnitude of elasticities on adult survival may be strongly underestimated when maternal or grand-maternal care is not implemented. Elasticities on fertilities by age can also exhibit very different patterns. Because such intergenerational transfers have been proposed as a very important drivers of the evolution of human life-history, the authors should at least discuss it. (4) As the authors wonderfully argued in a recent article, only periodic catastrophes in humans can explain the human forager paradox. It also means that all elasticity analysis in constant and infinite environment is somehow incomplete and elasticity should be considered into a stochastic model.

Minor comments & Détails
l418-420 -Isn't there a contradiction is stating that juvenile survival is under canalization effect and stating later on that it varies more in time than adult survival? l 32 -I am not sure that reference [2] did anything in calculating the divergence time between humans and chimpanzees. Please check carefully this reference. I think it should instead be referred to l 33-34. l 35, "human fertility is similar to chimpanzees" and further. Please be more specific. Do you mean the shape of the age-specific fertilities? If yes, both the distribution and the TFR? Is the whole shape the distribution identical? The authors refer to [6] who focus mainly on reproductive senescence and show that if the timing of reproductive senescence is similar rate of reproductive senescence is not the same as well as how it correlates with decline in survival. I suggest to be more precise.
L37-38 -"However, there is great variation among human and chimpanzee life histories". Here again I suggest to be more specific. The authors quote [8]. Although a valid reference, it can be completed by more recent article (as the [2]). Furthermore, I am not a native speaker but is "difference" would be better than "variation"? L39 and many time after -Please change "within species" by between population. In ecology, within species study refers more to the study of variance between individuals than between population as it is investigate here. L153 -this should be sij instead of sj isn't it?
L179 -I am not sure to understand why the fact that Cij and Eij sum to unity allow to calculate the ratio. Figure 3A -I find the figure very complicated to figure out. Are they the mean summed C values between populations? Then why and how is separated positive and negative C values? Or "composite" refers to the mean trajectories for HG, F, WC. But then, again, how does it lead to both (+) and (-) for a same trait (i.e., Infant survival). I am very sorry if I miss this information.