Social network centrality predicts dietary decisions in a wild bird population

Summary How individuals balance costs and benefits of group living remains central to understanding sociality. In relation to diet, social foraging provides many advantages but also increases competition. Nevertheless, social individuals may offset increased competition by broadening their diet and consuming novel foods. Despite the expected relationships between social behavior and dietary decisions, how sociality shapes individuals’ novel food consumption remains largely untested in natural populations. Here, we use wild great tits to experimentally test how sociality predicts dietary decisions. We show that individuals with more social connections have higher propensity to use novel foods compared to socially peripheral individuals, and this is unrelated to neophobia, observations, and demographic factors. These findings indicate sociable individuals may offset potential costs of competition by foraging more broadly. We discuss how social environments may drive behavioral change in natural populations, and the implications for the causes and consequences of social strategies and dietary decisions.

) eigenvector centrality, and subsequent novel food usage (proportion of novel food usage -y axis) for the (a,c) first trial, and (b,d) second trial.The point positions show the individual data points, point colour shows the colour of the novel food (red or green dyed peanut), point shape shows which experimental site the individual was at (site 1 or site 2), and point size indicates weight of the data point i.e. the total number of detections (at both the novel, and familiar food feeder).The lines show the GLM fit, and the surrounding polygons show the associated standard error around this estimate, with the red lines showing the fit for the red novel food site, the green line showing the fit for the green novel food site, and the black line denoting the overall fit.See Table S3 & S4 for full model details.

SUPPLEMENTARY TABLES
Table S1: Summary of experimental procedure, related to Figure 1.The study protocol at each of the sites, showing the phase of the study and food-types used over the data-collection days and the fine-scaling positioning of the feeders within the feeding sites.

Figure S1 :
Figure S1: Social centrality metrics and novel food usage, related to Figure 2. Prior social centrality (x axis), as measured as (a-b) Average edge weight and (c-d) eigenvector centrality, and subsequent novel food usage (proportion of novel food usage -y axis) for the (a,c) first trial, and (b,d) second trial.The point positions show the individual data points, point colour shows the colour of the novel food (red or green dyed peanut), point shape shows which experimental site the individual was at (site 1 or site 2), and point size indicates weight of the data point i.e. the total number of detections (at both the novel, and familiar food feeder).The lines show the GLM fit, and the surrounding polygons show the associated standard error around this estimate, with the red lines showing the fit for the red novel food site, the green line showing the fit for the green novel food site, and the black line denoting the overall fit.See TableS3& S4 for full model details.

Figure S2 :
Figure S2: Basic social measures and novel food usage, related to Figure 2. Average group size is the average size of the flocking event that the individual was observed in, and flockmates is the total number of unique individuals the individual was observed occurring with in at least one flocking event.Prior basic measures (x axis), as measured as (a-b) Average group size and (c-d) number of flockmates, and subsequent novel food usage (proportion of novel food usage -y axis) for the (a,c) first trial, and (b,d) second trial.The point positions show the individual data points, point colourshows the colour of the novel food (red or green dyed peanut), point shape shows which experimental site the individual was at (site 1 or site 2), and point size indicates weight of the data point i.e. the total number of detections (at both the novel, and familiar food feeder).The lines show the GLM fit, and the surrounding polygons show the associated standard error around this estimate, with the red lines showing the fit for the red novel food site, the green line showing the fit for the green novel food site, and the black line denoting the overall fit.See TableS5& S6 for full model details.

Table S3 : Average edge weight and novel food usage model outputs, related to Figure 2 and Figure S1.
Output of GLMs assessing the relationship between individuals' propensity to use novel food (response variable) and individuals' average edge weight (FigureS1a;S1b), along with the other fitted explanatory variables.Each column holds the test statistics for (A) experimental trial 1 and (B) experimental trial 2. Each row gives the result for each explanatory variable, with 'Sex' in relation to female birds, Age in relation to adult birds, Immigrant status in relation to residents, and the 'Edge' as average non-zero edge weight directly prior to each experimental trial (see Methods) and 'Observations' as the number of records.

Table S4 : Eigenvector centrality and novel food usage model outputs, related to Figure 2 and Figure S1.
Output of GLMs assessing the relationship between individuals' propensity to use novel food (response variable) and individuals' eigenvector centrality (FigureS1c;S1d), along with the other fitted explanatory variables.Each column holds the test statistics for (A) experimental trial 1 and (B) experimental trial 2. Each row gives the result for each explanatory variable, with 'Sex' in relation to female birds, Age in relation to adult birds, Immigrant status in relation to residents, and the 'Eigenvector' as weighted eigenvector centrality directly prior to each experimental trial (see Methods) and 'Observations' as the number of records.

Table S5 : Mean gathering event size and novel food usage model outputs, related to Figure 2 and Figure S2.
Output of GLMs assessing the relationship between individuals' propensity to use novel food (response variable) and individuals' average flock size (FigureS2a;S2b), along with the other fitted explanatory variables.Each column holds the test statistics for (A) experimental trial 1 and (B) experimental trial 2. Each row gives the result for each explanatory variable, with 'Sex' in relation to female birds, Age in relation to adult birds, Immigrant status in relation to residents, and the 'Flock size' as mean number of individuals within each flocking event the individual was observed in directly prior to each experimental trial (see Methods) and 'Observations' as the number of records.

Table S6 : Unique flockmates and novel food usage model outputs, related to Figure 2 and Figure S2.
Output of GLMs assessing the relationship between individuals' propensity to use novel food (response variable) and their number of unique flockmates (FigureS2c;S2d), along with the other fitted explanatory variables.Each column holds the test statistics for (A) experimental trial 1 and (B) experimental trial 2. Each row gives the result for each explanatory variable, with 'Sex' in relation to female birds, Age in relation to adult birds, Immigrant status in relation to residents, and the 'Flockmates' as sum of the number of unique individuals seen in the same flocking events as themselves directly prior to each experimental trial (see Methods) and 'Observations' as the number of records.

Table S7 : Social network strength and first feeder used model outputs, related to Figure 3.
Output of GLMs assessing the relationship between whether individuals are first detected on the novel food feeder when they first arrive at the experimental trial and their prior network strength (Figure3-Main Text), along with the other fitted explanatory variables.Each column holds the test statistics for (A) experimental trial 1 and (B) experimental trial 2. Each row gives the result for each explanatory variable, with 'Sex' in relation to female birds, Age in relation to adult birds, Immigrant status in relation to residents, and the 'Strength' as weighted network degree directly prior to each experimental trial (see Methods) and 'Observations' as the number of records.

Table S8 : Social network strength and time delay to use novel food, related to Figure 3.
Output ofLMs assessing the relationship between the amount of time taken for each individual to first land on the feeding perch of the novel food (quantified as time of day they were first recorded on the novel food), and their prior network strength, along with the other fitted explanatory variables.Each column holds the test statistics for (A) experimental trial 1 and (B) experimental trial 2. Each row gives the result for each explanatory variable, with 'Sex' in relation to female birds, Age in relation to adult birds, Immigrant status in relation to residents, and the 'Strength' as weighted network degree directly prior to each experimental trial (see Methods) and 'Observations' as the number of records.

Table S9 : Social network strength and overall time delay to use novel food, related to Figure 3.
Output of LMs assessing the relationship between the amount of time taken for each individual to first land on the feeding perch of the novel food (quantified as total elapsed foraging time since they were first detected at the site during the trial -log transformed), and their prior network strength, along with the other fitted explanatory variables.Each column holds the test statistics for (A) experimental trial 1 and (B) experimental trial 2. Each row gives the result for each explanatory variable, with 'Sex' in relation to female birds, Age in relation to adult birds, Immigrant status in relation to residents, and the 'Strength' as weighted network degree directly prior to each experimental trial (see Methods) and 'Observations' as the number of records.

Table S10 : Social network strength and novel food exploitation after first use, related to Figure 4.
Output of GLMs assessing the relationship between individuals' propensity to use novel food after they had already first tried the novel food feeder (response variable) and individuals' prior network strength (Figure4-Main Text), along with the other fitted explanatory variables.Each column holds the test statistics for (A) experimental trial 1 and (B) experimental trial 2. Each row gives the result for each explanatory variable, with 'Sex' in relation to female birds, Age in relation to adult birds, Immigrant status in relation to residents, and the 'Strength' as weighted network degree directly prior to each experimental trial (see Methods) and 'Observations' as the number of records.