Species integrity is maintained only if recurrent allelic exchange between subpopulations occurs by means of migrating specimens. Predictions of this gene flow on the basis of observed or assumed mobility of marine species have proven to be error-prone. Using one mitochondrial gene and seven microsatellite markers, we studied the genetic structure and gene flow in Septemserolis septemcarinata, a strictly benthic species lacking pelagic larvae and the ability to swim. Suitable shallow-water habitats around three remote islands (South Georgia, Bouvet, and Marion Island) are geographically disjunct, isolated by more than 2,000 km of uninhabitable deep sea (east-west) and also separated by the Polar Front (north-south), which serves as a strong demarcation line in many marine taxa. Although we did find genetic differentiation among the three island populations, our results also revealed that a scenario with recent gene flow explains our data best. A model assuming no gene flow after initial colonization of the islands performs significantly worse. The tests also favor an asymmetric gene flow pattern (west to east » east to west) thus mirroring the directionality of major oceanographic currents in the area. We conclude that rare longdistance dispersal rather than vicariance or human-mediated transport must be responsible for the observed patterns. As a mechanism, we propose passive rafting on floating substrata in the Antarctic Circumpolar Current. The results demonstrate that the effectiveness of a physical barrier is not solely a function of its physical parameters but strongly depends on how organisms interact with their environment. © Springer-Verlag 2010.