Simple foraging rules in competitive environments can generate socially structured populations

Abstract

Social vertebrates commonly form foraging groups whose members repeatedly interact with one another and are often genetically related. Many species also exhibit within-population specializations, which can range from preferences to forage in particular areas through to specializing on the type of prey they catch. However, within-population structure in foraging groups, behavioral homogeneity in foraging behavior, and relatedness could be outcomes of behavioral interactions rather than underlying drivers. We present a simple process by which grouping among foragers emerges and is maintained across generations. We introduce agent-based models to investigate (1) whether a simple rule (keep foraging with the same individuals when you were successful) leads to stable social community structure, and (2) whether this structure is robust to demographic changes and becomes kin-structured over time. We find the rapid emergence of kin-structured populations and the presence of foraging groups that control, or specialize on, a particular food resource. This pattern is strongest in small populations, mirroring empirical observations. Our results suggest that group stability can emerge as a product of network self-organization and, in doing so, may provide the necessary conditions for the evolution of more sophisticated processes, such as social learning. This taxonomically general social process has implications for our understanding of the links between population, genetic, and social structures.

Keywords: cooperation; foraging specialization; group dynamics; self‐organization; social network.

Figure 1

Emergent properties of simple social foraging rules. The (a) number of foraging groups, (b) average group size, and (c) average individual payoff, rapidly stabilized over time, resulting in a single group of size R where each individual receives an optimal payoff. Probability distributions are calculated as the observed value divided by the total value of the metric at each time step t. Inset plots represent three snapshots of the simulation time (vertical dashed lines in (c) indicate the optimal payoff r _i = 1). The simulations were run for 500 replicates of the baseline model with population size N = 40, resource patch size R = 5, initial connectivity among nodes T _prob = 0.2, and 200 simulated years (time steps = N*5, but for clearer visualization, the plots were truncated at time step t = 100). These patterns are consistent with other areas of the parameter space (Figure S1)

Figure 2

The emergence of foraging groups in the baseline model. In the surface plot (a), exclusivity (proportion of network density among the top connected individuals) is plotted as a function of initial population sizes (i.e., number of nodes in the network, N) and resource patch sizes (R). Each point is the result of one replicate of the simulation. In the social networks representative of four areas of the parameter space (b–e), individuals (nodes) are connected by the proportion of times (links) they are part of a group (links <0.3 are filtered out for better visualization, but included in all analyses). Red nodes indicate members of the emergent specialized foraging group at the end of the simulation. Data are based on 500 replicates of the model for each combination of N and R parameters. The initial network connectivity is T _prob = 0.2 (i.e., ~20% of possible links are randomly assigned when the network is initialized), but the results are independent of initial connectivity values (Figure S2)

Figure 3

Evolution of pedigree, relatedness, and social relationships in the reproductive model. Simulations were run for the parameter space area representing small population and resource patch sizes (see Figures 2b and 4b). The pedigree networks (a), showing all individuals that are alive at a given time step (x‐axis), show that related individuals (nodes connected by links) are frequently found in foraging groups. Red nodes (and shading) indicate individuals currently part of the single foraging group. Similarly, the relatedness networks (b) show that individuals within social groups are often highly related (the thicknesses of links are proportional to their relatedness). However, the network depicting individuals are connected by the proportion of times they have been part of a specialized foraging group (c) suggests that foraging groups often contain individuals from different genetic lineages (here node color represents unique genetic lineages; for better visualization links whose weights <0.3 are filtered out). The simulations are based on a population size N = 40, resource patch size R = 15, initial connectivity T _prob = 0.2, and run for 1,000 time steps, and these patterns are consistent across the parameter space (Figure S3)

Figure 4

Relatedness among members and nonmembers of emergent specialized foraging groups. Environments with a smaller resource patch lead to greater relatedness among group members, suggesting that the more specialized a population, the more related individuals are simply by chance. The surface plot (a) displays the log of the ratio between the relatedness among members of the emergent foraging group (“members”) and the average relatedness among all other individuals (“nonmembers”) at the end of the simulation (time step t = 1,000) as a function of initial population size N and resource patch size R. Positive log‐ratio indicates higher relatedness among members than nonmembers. The scatterplots (b–e) display relatedness outputs of 100 model replicates throughout the entire simulation (1,000 time steps each), representative of the four areas of the parameter space. In the colored scatterplots, black circles represent the average relatedness among foraging group members; blue circles represent average relatedness among nonmembers; and red circles represent the average relatedness among 100 randomly chosen sets of individuals of the same size of the foraging groups. In the other scatterplots, black circles represent the log of the relatedness ratio between foraging group members and sets of randomly chosen individuals throughout the simulation time for 100 model replicates. Here, positive log‐ratio indicates higher relatedness among members than expected by chance. Green lines represent the mean log‐ratio across model replicates, and green polygons indicate 95% confidence intervals. Horizontal dashed lines indicate log‐ratio = 0. Cases of undefined relatedness ratio (l _m,t = l _r,t = 0 or l _r,t = 0) and zeroed ratio (l _m,t = 0 → ln(0) = −∞), making the plots for small population sparser (b,c). In all cases, the initial network connectivity is T _prob = 0.5. Each combination of N and R parameters is run for 100 model replicates, and each model is run for 1,000 time steps