Data collection
We constructed an FST dataset through a systematic search in google scholar (key words: “genetic structure”, “population differentiation”, “population genetics”, “genetic diversity”, “population gene flow”) for articles published up until June 2018. The search yielded 356 peer-reviewed publications on seed plants for which measures of population genetic structure (FST) based on nuclear markers were available. When multiple studies reported FST values for the same species, we recorded the FST from the study with the largest geographic range, as this may better represent the genetic diversity found in the species (Cavers et al., 2005). By this criterion, we compiled a dataset that included 337 unique species. We extracted information for the predictor variables directly from the publications, and infrequently complemented this, where necessary, with information from peer-reviewed literature on the studied species (see Appendix S1 and Table S1 in Supporting Information). Predictor variables were included in multiple regressions to explain variation in FST values (see section FST models). We included three factors that pertained to the sampling scheme of each study and that can potentially affect FST (Nybom, 2004; Nybom & Bartish, 2000): genetic marker used, maximum distance between populations, mean sample size per population. We used them to construct a null model to be compared against models with our factors of interest. Factors of interest consisted of five categorical variables with 2–4 levels: mating system (outcrossing, mixed-mating), growth form (non-woody, shrub, tree), pollination mode (large insects, small insects, vertebrates, wind), seed dispersal mode (animal, gravity, wind), and latitudinal region (tropics, sub-tropics, temperate). Below we explain the FSTestimates and all eight factors used in this study in greater detail.