Database compilation
We searched the ISI Web of Science in March 2017 for comparative studies that included data on phenotypic traits and/or neutral genetic diversity of populations on marine islands and on mainland sites in any taxonomic group. Search terms were ‘island’ and (‘mainland’ or ‘continental’) and ‘population*’ and (‘demograph*’ or ‘fitness’ or ‘survival’ or ‘growth’ or ‘reproduc*’ or ‘density’ or ‘abundance’ or ‘size’ or ‘genetic diversity’ or ‘genetic structure’ or ‘population genetics’) and (‘plant*’ or ‘tree*’ or ‘shrub*or ‘animal*’ or ‘bird*’ or ‘amphibian*’ or ‘mammal*’ or ‘reptile*’ or ‘lizard*’ or ‘snake*’ or ‘fish’), subsequently refined to the Web of Science categories ‘Ecology’ or ‘Evolutionary Biology’ or ‘Zoology’ or ‘Genetics Heredity’ or ‘Biodiversity Conservation’ or ‘Marine Freshwater Biology’ or ‘Plant Sciences’ or ‘Geography Physical’ or ‘Ornithology’ or ‘Biochemistry Molecular Biology’ or ‘Multidisciplinary Sciences’ or ‘Environmental Sciences’ or ‘Fisheries’ or ‘Oceanography’ or ‘Biology’ or ‘Forestry’ or ‘Reproductive Biology’ or ‘Behavioral Sciences’. The search included the whole text including abstract and title, but only abstracts and titles were searchable for older papers depending on the journal. The search returned 1237 papers which were distributed among co-authors for further analyses.
We chose papers for inclusion in the dataset if the same species was studied on a minimum of two islands and two sites on the mainland (Fig. 1). While we accepted the authors’ judgement about island versus mainland status, we made our own judgement based on the relative size of the island or position relative to the mainland i.e., some islands were reinterpreted as mainland if they were at least four times larger than smaller islands, with the median size difference for islands reclassified as mainland being 126 times larger than other islands in the study (19 papers), or if the distance of the island from the continent very low compared to the rest of the islands (4 km vs. 1700 km in one paper, and 300 m vs. 11 km in another paper; Table S1.1). We eliminated studies comparing populations on several islands where there were no clear island versus mainland comparisons even after reinterpreting the island status, studies referring to migratory species, recent invasions (<50 years), marine species (including coastline organisms) and ex situ populations. The complete selection criteria are presented in Supporting material S1. The initial filter resulted in 235 papers which were then redistributed among co-authors for a second round of filtering. In the second filter, we excluded papers that did not provide both population geographic coordinates and population-level quantitative data, unless data were provided upon contacting the authors or could be obtained from figures using DataThief (Tummers 2006). We visually inspected maps plotted for each study separately and we made minor adjustments to the GPS coordinates when the coordinates placed the focal population off the island or mainland. For this study, we included only responses measured at individual level, therefore we removed papers referring to demographic performance, and we also excluded traits such as immunity, behaviour and diet that are heavily reliant on ecosystem context. We extracted data on population level mean for two broad categories of response: i ) broad phenotypic measures, which included traits such as size and weight of entire body or body parts, morphology (e.g., mandible shape, number of stamens, wood density), metabolism products (e.g., colour of skin), physiology (e.g., digestive efficiency), vital rates (growth, survival, reproduction) and mean age of sampled mature individuals; and ii ) genetic diversity, which included heterozygosity, allelic richness, number of alleles per locus (Fig. S2.2, Table S2.1). The final dataset included 115 studies of 112 species (74 animals and 38 plants) in 1019 island populations (corresponding to 569 unique islands) and 813 mainland populations, with population-level taxonomic and biogeographic information, totalling 7736 records (Fig. 2, Fig. S2.1-S2.2, Table S2.1, Fig. S4.1). The dataset and the corresponding bibliography are available in the Dryad data repository (https://doi.org/10.5061/dryad.h18931zqg) .