Figure captions
Figure 1. (a) Figure showing the main study selection criterion. The population (marked with circle) was the observation level for analysis, and only study systems with a minimum of two islands (I1, I2) and two mainland populations (M1, M2) were considered. Circle size indicates possible differences between the mean of a specific phenotypic trait or a neutral genetic diversity measure. We calculated the log ratio between the largest and the smallest values for paired island populations and paired mainland populations respectively for each phenotypic trait and neutral genetic diversity measure (Y axis in (b)). (b) Expectations for a positive effect of geographic and macroclimatic distances on the calculated response variables within island and mainland systems. We expected stronger effects of geographic distance on mean phenotypes and neutral genetic diversity in island systems compared to the mainland systems, and we did not expect the effects of macroclimatic distance to differ between the two system types. We expected stronger responses of phenotypic traits compared to the neutral genetic diversity.
Figure 2. (a) The global representation of 1019 island (turquoise dots) and 813 mainland populations (orange dots) with population-level phenotypic traits and genetic diversity data studied comparatively in island and mainland systems for 112 species. (b) Number of species in different taxonomic groups included in this study (vascular plants were sub-grouped into Tree or Shrub and Herbaceous life forms). (c) Two sample maps showing the geographic distribution of populations studied comparatively in island and mainland systems for a bird (Melospiza melodia ) and a plant (Elymus glaucus ) species.
Figure 3. Log-ratio of phenotypic trait values between island populations (turquoise) and between mainland populations (orange) for 100 sampled datasets. Horizontal black lines represent intercept estimates for each of the 100 sampled datasets, and the point with a vertical dashed line represent the mean and the overall 95% credibility interval. Points are jittered horizontally for better visualisation.