There is concern about the worldwide decline of bees, in which genetic factors may play a role. As populations of these haplodiploid insects suffer habitat fragmentation and subsequent isolation, theory predicts an increase in inbreeding and a concomitant increase in inviable or sterile diploid males, a product of reduced allelic diversity at the sex determining locus, which could lead to a diploid male extinction vortex. To test this idea, we genotyped 1,245 males of one orchid bee, Euglossa cordata, a species with low diploid male production on the mainland. We genotyped bees from the Brazilian mainland and three islands using five highly variable microsatellite loci. Allelic richness was lowest on the most remote island 38 km from the mainland and, though the degree of dif-ferentiation across localities was modest (global F ST = 0.034; global G 0 ST ¼ 0:153, both P \ 0.001) and isolation by distance was weak (Mantel test, r = 0.614, P = 0.056), sea was revealed to be a significant barrier to inferred historic gene flow (partial Mantel test of distance over sea, r = 0.831, P = 0.003). Only seven males were diploid (mean diploid male production, DMP, 0.6 %). Though the proportion of diploid males was highest on the most remote island (1.3 %), differences in DMP across study areas were statistically non-significant. Thus island isolation leads to reduced genetic variation at putatively neutral microsatel-lite loci, but E. cordata nevertheless seems to maintain allelic diversity at the sex locus, possibly because of suf-ficient gene flow, or because it is a locus under balancing selection, or because of the joint action of these two evo-lutionary forces: migration and selection. These and other bee species may be able to maintain sufficient variability at the sex locus to avoid entering the DM extinction vortex, even on relatively isolated islands or habitat fragments.