i) Main analyses using an alternate phylogeny
We performed all our main analyses on an additional, separate Heliconiini phylogeny (Cicconardi et al., 2022). Overall, we recorded 19 gregarious species, out of 57 included in Cicconardi et al. (2022). Our model estimated eight independent transitions to gregariousness, and the phylogenetic signal of social behaviour is estimated to be very strong (λ = 1.0). Transitions between solitary and gregarious states are most likely to occur at equal rates (X2(1) = 2.146,p = 0.143, transition rate = 0.033), with four reversals from gregarious larvae back to solitary larvae across the phylogeny. Using categorical social behaviour data, the ‘all rates different’ model is better fitting (model comparison: X2(1) = 29.014,p < 0.001), with the highest transition rate estimated to be from G2 to G3 (0.683 vs next highest G3 to G2 = 0.539).
As reported in our main analysis, solitary larvae feed on a greater variety of host species than gregarious larvae (P-mean = -0.855, 95% CI -1.673 – -0.083, PMCMC = 0.039). After omittingPassiflora hosts used by fewer than four larval species from the data, we found no evidence of an interaction between larval social behaviour and any host species. Additionally, we found that more frequently used hosts by all larvae tend to grow in woods edge habitats (P-mean = 2.989, 95% CI 1.322 – 4.681, PMCMC = 0.001, Table A6). Similar to our main results, we did not identify any interactions between specific host plant traits, including leaf size, and larval social behaviour (Table A6). We also found no interaction between adult wingspan and host leaf size across gregarious larvae (P-mean = -0.031, 95% CI -0.586 – 0.552, PMCMC= 0.914).