The effect of Pleistocene climate change on the organisms of tropical and subtropical regions is rather poorly understood. We therefore studied the land snail genus Mandarina (Bradybaenidae) of oceanic Ogasawara (Bonin Islands, Japan), with the aim of using population genetic data to understand their recent history. Our analysis of a mitochondrial 16S ribosomal RNA region from more than 600 snails in five ground-living species suggests that populations on the small islands of Mukoujima, Anejima, Imotojima and Meijima, as well as on the low-lying southern and central parts of Hahajima, have probably undergone recent bottlenecks followed by subsequent expansions. Except between the main island of Hahajima and Mukouijima, there is almost no evidence for gene flow among islands even though the islands were connected repeatedly by land bridges through the Pleistocene. Within islands the population structure is severe, suggestive of a long-term, low level of gene flow (F(ST) is frequently greater than 0.5 among geographically close populations). Finally, there is a marked genetic patchiness, meaning that genetically close populations are sometimes separated by genetically distant populations. These patterns could be a consequence of expansion from bottlenecks, low active dispersal and founder effects caused by rare long-distance migrants. Unfortunately, the exact nature of the refugia and bottlenecks remains unknown because the palaeoclimate of this region is poorly understood. Dating the population size changes is also challenging because the molecular clock is uncertain. We suggest, however, that arid conditions or deforestation induced by decreased atmospheric CO(2) may have been the main factor in determining population size.