Species richness of habitat fragments is affected by spatial isolation. However, the scale of this phenomenon, and its interactions with the species' seed dispersal potential has remained underexplored. By integrating seed trap and species distribution data, Koh et al., in this issue of the Journal of Vegetation Science, make a compelling case for scale-dependent species' responses to forest fragmentation.It has been almost half a century now since Robert MacArthur and Edward Wilson suggested that the same principles that govern the species richness of oceanic islands also apply to formerly continuous terrestrial habitats that are broken up by encroaching civilization. Their ideas resulted in an avalanche of research papers studying fragmented ecosystems across all biomes, and have greatly contributed to the maturing of the discipline we know as conservation biology. Especially in the highly fragmented agricultural landscapes of Europe, where species-rich semi-natural grasslands and forests typically occur as islands in a hostile landscape matrix, a wealth of studies of increasing sophistication has been produced since the 1990s, basically building on the island theory. As a result, we have acquired profound insights not only into the landscape and fragment characteristics that affect plant species richness and community and functional trait composition of habitat fragments (e.g. Marini et al. 2012;Wulf & Kolk 2014), but also into the genetic and demographic processes that drive small and isolated plant populations towards extinction (e.g. Vandepitte et al. 2007). Although it is known that well-connected habitat fragments, ceteris paribus, contain more plant species than more isolated ones, the scale at which the surrounding landscape mediates this phenomenon, and how scale interacts with the ability of seeds to disperse across large distances has remained largely unexplored. Koh et al. (2015) elegantly shed light on this scale issue (Fig. 1). The authors studied plant species richness of small traditional Korean village forests (maeulsoops), which occur as isolated patches in the agricultural landscape matrix in the valleys between mountain forests. Instead of simply adding the connectivity of the study fragments at different spatial scales as explanatory variables for species richness to their models, they evaluated model performance at 13 different landscape buffer sectors, ranging in size from a 50-m to a 3500-m radius. The authors show that species that have seeds that are specifically adapted for long distance dispersal (LDD) positively responded to deciduous forest cover in a radius of 3000 m around the target fragment, whereas species with no such adaptations (short distance dispersal (SDD) species), positively responded to total forest cover in a 150-m radius. For both species groups, the main share of the variation in species richness across forest fragments was explained by the degree of anthropogenic disturbance of the maeulsoops, suggesting an important role for recruitment limitation as wel...