Stable producer-scrounger dynamics in wild birds: sociability and learning speed covary with scrounging behaviour

Abstract

There has been extensive game-theoretic modelling of conditions leading to equilibria of producer-scrounger dichotomies in groups. However there is a surprising paucity of experimental evidence in wild populations. Here, we examine producer-scrounger games in five subpopulations of birds feeding at a socially learnt foraging task. Over four weeks, a bimodal distribution of producers and scroungers emerged in all areas, with pronounced and consistent individual tactic specialization persisting over 3 years. Tactics were unrelated to exploratory personality, but correlated with latency to contact and learn the foraging task, with the late arrivers and slower learners more likely to adopt the scrounging role. Additionally, the social environment was also important: at the broad scale, larger subpopulations with a higher social density contained proportionally more scroungers, while within subpopulations scroungers tended to be central in the social network and be observed in larger foraging flocks. This study thus provides a rare example of a stable, dimorphic distribution of producer-scrounger tactics in a wild population. It further gives support across multiple scales for a major prediction of social foraging theory; that the frequency of scroungers increases with group size.

Keywords: Parus major; animal personality; producer–scrounger games; social foraging; social learning; social network analysis.

Figure 1.

(a) Distribution of scrounging behaviour in individuals across subpopulations T1–T5 in an initial four-week observation period. Bars are semi-transparent and overlaid to show all replicates, and only individuals with ≥50 visits are shown. (b) Example of individual longitudinal trajectories from T1, showing a different coloured line for each individual of the running mean of the proportion of their last 20 rewarded visits that were scrounges, plotted proportional to their total rewarded visit count; again only individuals with ≥50 visits are shown. A longitudinal clustering algorithm identifies two distinct clusters, shown in black (‘scroungers’) and grey (‘producers’). (c) Example of individual longitudinal trajectories from T3. Note that in both (b) and (c) switches are infrequent.

Figure 2.

Individual (non-social) predictors of tactic choice, showing all birds with ≥50 visits. Asterisks highlight significant differences. (a) Adults and females were more likely than males and juveniles to be classed as scroungers in a longitudinal clustering model. Vertical bars show standard error. (b) Birds that arrived later in the experimental period were more likely to adopt the scrounging role. (c) Scroungers also tended to have a longer latency to learn than producers. (d) There was no relationship between scrounging and personality score. Horizontal lines show median in (b), (c) and (d). (Online version in colour.)

Figure 3.

Social predictors of tactic choice, showing all birds with ≥50 visits. (a) The proportion of birds that were classed as scroungers in a longitudinal clustering model was higher in replicates with higher population sizes. Lines show linear model fit and 95% CI (p = 0.01, R² = 0.39). (b) Unweighted degree was higher for birds classed as scroungers in a longitudinal clustering model. However sociality interacted with the time of contact in the experimental period; birds that visited on day 1 (magenta) and in the first week (day 2–5, orange) showed a positive relationship between degree centrality and likelihood of scrounging, but birds that arrived later (day 6–20, yellow) showed no relationship. Points represent median and size is relative to the number of individuals in each category (day 1, n = 46; day 2–5, n = 89, day 6–20, n = 100), vertical bars show 95% CI. (Online version in colour.)