Indirect genetic effects among neighbors promote cooperation and accelerate adaptation in a small-scale human society

Abstract

Explaining the rapid evolution of human cooperation and its role in our species' biodemographic success remains a major evolutionary puzzle. To address this challenge, we tested a social drive hypothesis, which predicts that social plasticity and social selection in human groups cause indirect genetic effects that accelerate the adaptation of fitness, promoting population growth via feedback between the environmental causes and evolutionary consequences of cooperation. Using Bayesian multilevel models to analyze fertility data from a small-scale society, we demonstrate that density- and frequency-dependent indirect genetic effects on fitness promote the evolution of cooperation among neighboring women, increasing the rate of contemporary adaptation by ~5×. Our results show how interactions between the genetic and socioecological processes shaping cooperation in reproduction can drive rapid growth and social evolution in human populations.

Fig. 1.. IGEs and the adaptation of fitness in social environments.

(A) The total DGE on an individual’s fitness $W$ is determined by and proportional to $\propto$ the total magnitude of nonsocial selection ${\mathbf{\beta }}_{\mathbf{N}}$ acting on heritable variation $\mathbf{A}$ in all their fitness-relevant phenotypes $\mathbf{\eta }$ ; the total IGE on their fitness is in turn proportional to the total magnitude of social selection ${\mathbf{\beta }}_{\mathbf{S}}$ acting on heritable variation ${\mathbf{A}}^{\prime }$ in all fitness-relevant phenotypes ${\mathbf{\eta }}^{\mathbf{\prime }}$ expressed by cooperators or competitors in their environment. These heritable effects are shaped both by social plasticity $\mathbf{\Psi }$ and average assortment/relatedness $\mathbf{R}$ across phenotypes. See Table 1 for an overview of key notation and terminology. Evolvability (potential for adaptation) of fitness in social environments $e_{W\mid \overline{r}}$ is in turn determined by the total variances of IGEs and DGEs among individuals in the population (which are always positive), the sign and magnitude of covariance between individuals’ DGEs and IGEs scaled by the expected number of social partners $\overline{n}$ , and the average relatedness $\overline{r}$ between individuals and their social partners [expanding Fisher’s fundamental theorem; (31, 40)]. Note that $100\ast e_{W\mid \overline{r}}$ can be interpreted as the expected % change in fitness attributable to natural selection, which reflects adaptive change in the instantaneous growth rate of the population. See Eqs. 3 to 18 for mathematical details. (B) Given a fixed level of DGEs, here $\text{var}\left({\mathbf{W}}_{\mathbf{D}}\right)=0.3$ with $\overline{W}=1$ and $\overline{n}=1$ , the covariance of DGEs and IGEs $\text{cov}({\mathbf{W}}_{\mathbf{D}}\mathit{,}{\mathbf{W}}_{\mathbf{I}})$ can accelerate or inhibit the expected rate of change in the heritable component of average population fitness [i.e., adaptation; (41)]. (C) Conceptual overview of the social drive hypothesis of rapid human adaptation that is tested empirically in this paper.

Fig. 2.. The Tsimane people of Bolivia.

The Tsimane are an Indigenous people that occupy a broad territory within and surrounding the lowland Amazonian forests of Bolivia, who largely rely on foraging, hunting, and horticulture for their subsistence (59). Multiple Tsimane families tend to live in close spatial proximity together, forming clusters or “neighborhoods,” and cooperate in daily resource production, childcare, manual labor, and other fitness-relevant activities. These neighborhoods are nested within broader communities. (A) Neighboring kin socializing together, with a grandmother sitting in contact with her daughters and grooming her granddaughter. (B) Children from two neighboring households play together while their mothers prepare a shared meal. (C) Many families gather for a community-wide educational event. (D) Local ecological conditions can vary appreciably across the Tsimane territory, from forests to open grasslands and riverine habitats. (E) For the 93 communities included in our fertility dataset (see Materials and Methods), those lacking information on reproductively active neighbors are filled orange, while those with neighbor information are filled yellow. The radiuses of yellow circles are scaled by the total number of reproductively active neighbors meeting our selection criteria that contributed to the dataset, ranging from 2 to 83 women per community (median = 8, mean = 14). Photo credit: Jordan Scott Martin. All photographs were taken and used for scientific publication with community and family consent.

Fig. 3.. Social selection, IGEs, and the evolvability of Tsimane fertility.

(A) Conceptual overview of our Bayesian multilevel IGE model [Eqs. 1A to 1C; based on (58)]. IGEs were estimated using the slope ${\mathrm{\beta }}_{\text{SW}}$ of a focal woman’s fertility on the mean DGE of her neighbors ${\overline{W^{\prime }}}_D$ scaled by the total number of neighbors $n$ , such that $\text{var}\left({\mathbf{W}}_{\mathbf{I}\mid \overline{\mathit{n}}}\right)=\text{var}\left({\mathbf{W}}_{\mathbf{D}}\right){\mathrm{\beta }}_{\text{SW}}^2{\overline{n}}^2$ and $\text{cov}({\mathbf{W}}_{\mathbf{D}},{\mathbf{W}}_{\mathbf{I}})=\text{var}\left({\mathbf{W}}_{\mathbf{D}}\right){\mathrm{\beta }}_{\text{SW}}$ (Eq. 2), where the slope on $W_D$ is fixed to 1 by construction (see Materials and Methods for mathematical details). (B) Posterior distributions for average quantitative genetic effects on fertility, shown both for the variance in marginal IGEs due to the average effect on a single neighbor (n = 1) and the total IGEs due to the average number of neighbors in the sample ( $\overline{n}=2.4$ ). (C) Posterior distribution of the average ${\mathrm{\beta }}_{\text{SW}}$ among Tsimane women, where ${\mathrm{\beta }}_{\text{SW}}>0\equiv \text{cov}({\mathbf{W}}_{\mathbf{D}},{\mathbf{W}}_{\mathbf{I}})>0$ indicates net selection on heritable fertility variation for cooperation among neighbors (positive-sum payoffs), while ${\mathrm{\beta }}_{\text{SW}}<0\equiv \text{cov}({\mathbf{W}}_{\mathbf{D}},{\mathbf{W}}_{\mathbf{I}})<0$ indicates net selection for conflict (zero-sum payoffs). (D) Posterior distributions for the expected evolvability of fertility due to selection on heritable fertility variation among Tsimane women. The nonsocial evolvability $e_{W0}$ ignores the evolutionary consequences of IGEs, the inclusive fitness evolvability $e_{W_{\mathit{\text{IF}}}\mid \overline{r}}$ only accounts for IGEs among kin ( $\overline{r}=0.16$ ), while the social evolvability $e_{W\mid \overline{r}}$ accounts for IGEs among all neighbors.

Fig. 4.. Fluctuating social selection across Tsimane neighborhoods and communities.

Posterior estimates for patterns of fluctuating social selection on heritable fertility variation ${\mathrm{\beta }}_{\text{SW}}$ and resulting community- and neighborhood-level variation in the quantitative genetic covariance $\text{cov}({\mathbf{W}}_{\mathbf{D}},{\mathbf{W}}_{\mathbf{I}})$ (Eqs. 1D and 1E). (A) Posterior medians (points) and 10 to 90% CIs (shaded lines) for community-specific (y axis) genetic covariances (x axis), with values at 0 indicating no net selection on heritable fertility variation and values below/above 0 indicating net selection for conflict/cooperation among neighbors within the community. Note that because of partial pooling of random effects, communities with fewer observed neighbors tend to cluster more around the average across communities (0.03, solid line). Therefore, the plotted predictions are likely to underestimate the true magnitude of variation among communities. (B) Posterior distributions (top) for density-dependent selection ${\mathrm{\beta }}_{\mathrm{D}}$ and its resulting effect (bottom) on the magnitude of $\text{cov}({\mathbf{W}}_{\mathbf{D}},{\mathbf{W}}_{\mathbf{I}})$ across neighborhoods of varying size, with median predictions shown by the dark line and 10 to 90% Bayesian CIs indicated by the shaded bands. (C) The same is shown for frequency-dependent selection ${\mathrm{\beta }}_{\mathrm{I}}$ (blue) and its moderation by neighbor relatedness ${\mathrm{\beta }}_{\text{Ir}}$ (purple), with predictions plotted as function of the interaction between women’s fertility DGEs and the average fertility DGE of their neighbors. The increased slope in higher relatedness neighborhoods (shown for coresiding sisters, $\overline{r}$ = 0.5) indicates synergy (positive frequency dependence) among kin.