Influence of growth form on FST
We found that trees have populations with significantly lower FST than both shrubs and non-woody plants (Fig. 1b). Even though most trees are outcrossing in our dataset, our results show that growth form contributes to the variation in FSTindependently from mating system, contrary to the findings of Duminil et al. (2007, 2009). The inherent difference in scale between growth forms may contribute to this pattern: a given geographic distance between subpopulations may restrict gene flow much more for an herb than for a tree. In fact, neighborhood size, i.e. the spatial extent of closely related individuals, is larger in trees than shrubs and herbs (Vekemans & Hardy, 2004). Furthermore, trees usually have greater longevity than shrubs and non-woody plants (Duminil et al., 2009), which may increase the chances of gene flow between tree subpopulations, more than for other growth forms. Finally, the fact that growth form and habitat are tightly linked may also contribute; many non-woody plants and shrubs in our dataset occur in the forest understory, while many trees reach the canopy. Givnish (2010) and Theim, Shirk, and Givnish (2014) hypothesized that the understory imposes more limits to gene flow than the canopy because of the sedentary lifestyle of animal mutualists in the understory.