Interspecific social networks promote information transmission in wild songbirds

Abstract

Understanding the functional links between social structure and population processes is a central aim of evolutionary ecology. Multiple types of interactions can be represented by networks drawn for the same population, such as kinship, dominance or affiliative networks, but the relative importance of alternative networks in modulating population processes may not be clear. We illustrate this problem, and a solution, by developing a framework for testing the importance of different types of association in facilitating the transmission of information. We apply this framework to experimental data from wild songbirds that form mixed-species flocks, recording the arrival (patch discovery) of individuals to novel foraging sites. We tested whether intraspecific and interspecific social networks predicted the spread of information about novel food sites, and found that both contributed to transmission. The likelihood of acquiring information per unit of connection to knowledgeable individuals increased 22-fold for conspecifics, and 12-fold for heterospecifics. We also found that species varied in how much information they produced, suggesting that some species play a keystone role in winter foraging flocks. More generally, these analyses demonstrate that this method provides a powerful approach, using social networks to quantify the relative transmission rates across different social relationships.

Keywords: mixed-species flocking; network-based diffusion analysis; public information; social information; social networks; transmission networks.

Figure 1.

For each of the two areas in the study ((a,b) Cammoor/Simpsons Copse; (c,d) Higgins Copse) we formed two candidate social networks. One network contained all of the associations between conspecifics (a,c), with all the edges that were observed between nodes of the same species. The other network contained all of the associations between heterospecifics (b,d), with all the edges that were observed between nodes of different species. Node colour and label represents species (blue, B: blue tits; yellow, G: great tits; grey, M: marsh tits). Similarly, edge colour is the combination of the connecting nodes (e.g. green are edges between great tits and blue tits). Node size represents eigenvector centrality, which was calculated in the original study [7]. (Online version in colour.)

Figure 2.

Arrival time and order for each experimental diffusion. Each diffusion was tested against the networks from that area (Cammoor/Stimpsons Copse with figure 1a,b and Higgins Copse with figure 1c,d) to estimate social and asocial rates of information acquisition. Each newly arrived individual is shown by a coloured point (blue: blue tit, yellow: great tit, grey: marsh tit). Arrival times were binned by hour, but the order of arrivals was maintained (from bottom to top). (Online version in colour.)

Figure 3.

(a) Breakdown of discovery events corresponding to the estimated network-based diffusion analysis parameters. The area of each box represents the estimated proportion of individual patch discoveries events (independent arrivals to the patch by each individual) that were a result of transmission within species (38%), transmission between species (23%), or asocial learning (39%). The latter is further broken down by species, with numbers in parentheses giving the observed number of individuals (see table 1). For example, 7% of all arrivals were by marsh tits who discovered the patch without having access to social information. We also calculated the estimated rate of asocial discovery per capita (b). Each individual marsh tit accounted for 2.6% of all asocial discoveries (totalling 18% of all asocial discoveries by just seven individuals), and thus produced on average 3.7 times more new information than individual great tits and 6.5 times more information than individual blue tits. The size of each boxes represents the estimated percentage of total discoveries (a) and asocial discoveries (b).