Determinants of phenotypic difference between populations
The log ratio values for the phenotypic traits included in the analysis ranged from 0, indicating cases with no difference between populations, to 3.7 which, when back transformed from log space, corresponds to a ratio of approximately 39:1 between populations for the given measure. As expected, we found higher differences in phenotypic traits between island populations, than between mainland populations, mainland populations having a log ratio 0.189 lower than island populations (mode = -0.19, 95%CI = -0.37, -0.01, Table 1a, Fig. 3, Table S6.1a). This difference in the level of variation corresponds to a ratio of trait values of approximately 1.44:1 between island populations and 1.15:1 for mainland populations. We found some support for an interaction between system type (island or mainland) and geographic distance, i.e. phenotypic variability between populations tended to increase with increasing geographic distance on the mainland, but it was constant for all geographic distances on islands. This effect size corresponds to variation between mainland populations approximately 500 km apart matching the variation found between island system populations at any distance (Table 1a, Table S6.1a, Fig. S6.2a). The effect of other variables was even weaker (Table 1a, Table S6.1a). Across the random terms included in the model, most of the variation was associated with the residual terms, less variation was associated with the study or the species, and very little effect was attributed towards the phylogenetic term and response type (Table 1a). The zero adjusted models (Table S6.1b, Table S6.3a), the reduced model that included phenotypic traits associated with size (Fig. S6.4a) and models which excluded the macroclimatic or geographic distance (Table S6.5a, Table S6.6a) produced qualitatively similar results to the main phenotypic model.