Islands have traditionally been considered as migratory and evolutionary dead ends for two main reasons: island colonizers are typically assumed to lose their dispersal power, and continental back colonization has been regarded as unlikely because of niche preemption. The hypothesis that islands might actually represent dynamic refugia and migratory stepping stones for species that are effective dispersers, and in particular, for spore-producing plants, is formally tested here, using the archipelagos of the Azores, Canary Islands, and Madeira, as a model. Population genetic analyses based on nuclear microsatellite variation indicate that dispersal ability of the moss Platyhypnidium riparioides does not decrease in the island setting. The analyses further show that, unlike island populations, mainland (southwestern Europe and North Africa) populations underwent a severe bottleneck during the last glacial maximum (LGM). Our results thus refute the traditional view of islands as the end of the colonization road and point to a different perception of North Atlantic archipelagos as major sources of biodiversity for the postglacial recolonization of Europe by spore-producing plants.biogeography | speciation | genetic diversity | island syndromes | isolation-with-migration I sland biogeography has historically been and remains a research area of prime importance for the advance of biology (1). It was in an island setting that Darwin (2) and Wallace (3) presented their theories of evolution by natural selection. It was also in this context that McArthur and Wilson (4) described the "equilibrium model," according to which biodiversity patterns can be explained by isolation from source populations (immigration) and area size (extinction). Oceanic islands hence appear as natural laboratories of evolutionary and biogeographic processes and Wallace (5) suggested that understanding evolutionary processes within an island context is a key to understanding the same mechanisms in the more complex continental settings.Theoretical work on population structure has used island models as paradigms for understanding how interactions between gene flow, effective population size, and genetic drift influence patterns of genetic variation. Although surprisingly few studies have explicitly tested models of population structure and diversity in island and mainland plant populations (6), theory predicts that island populations should be less variable and more genetically differentiated relative to source populations (7). Differentiation of island populations is further enhanced by one of the most striking syndromes displayed by island taxa, that is, their typical loss of dispersal power as a result of either absence of predation and/or competition in the island setting or counter selection against high dispersal ability associated with the risk of massive individual loss to surrounding sea water (8). This process can be fairly rapid. In the annual daisy Lactuca muralis, for example, a substantial and significant loss of dispersal power owing...