iv) Tests of correlated evolution
We performed tests for correlated evolution between larval social behaviour and the other larval and host plant traits, whilst controlling for phylogenetic effects, using MCMCglmm (Hadfield 2010). We ran MCMC models for 5.1 million iterations, with a 0.1 million burn-in and sample storage frequency of every 500 iterations, with significance of the model calculated as the probability of the parameter value being different from zero (PMCMC ). We also report each model cofactor’s posterior mean (P-mean) and its 95% confidence intervals (CI). All analyses were performed using uninformative, parameter expanded priors for the random effect (G: V = 1, nu = 1, alpha.mu = 0, alpha.V = 1,000; R: V = 1, nu = 0.002) and default priors for the fixed effects. Finally, the Heliconiini dataset on host plant use could potentially contain some uncertainty as it combines records from geographically disperse Heliconiini species. Host plant use can vary across populations of single Heliconiini species (e.g. Merrill et al., 2013), potentially leading to overestimations in host use numbers for some species when data are combined over wide geographical ranges. We therefore also performed a more taxonomically limited analysis of a well studied community of 14 Heliconiini and nine Passifloraspecies in Panama (Merrill et al., 2013), as described above.