Severe population declines led to the listing of southern California Rana muscosa (Ranidae) as endangered in 2002. Nine small populations inhabit watersheds in three isolated mountain ranges, the San Gabriel, San Bernardino and San Jacinto. One population from the Dark Canyon tributary in the San Jacinto Mountains has been used to establish a captive breeding population at the San Diego Zoo Institute for Conservation Research. Because these populations may still be declining, it is critical to gather information on how genetic variation is structured in these populations and what historical inter-population connectivity existed between populations. Additionally, it is not clear whether these populations are rapidly losing genetic diversity due to population bottlenecks. Using mitochondrial and microsatellite data, we examine patterns of genetic variation in southern California and one of the last remaining populations of R. muscosa in the southern Sierra Nevada. We find low levels of genetic variation within each population and evidence of genetic bottlenecks. Additionally, substantial population structure is evident, suggesting a high degree of historical isolation within and between mountain ranges. Based on estimates from a multi-population isolation with migration analysis, these populations diversified during glacial episodes of the Pleistocene, with little gene flow during population divergence. Our data demonstrate that unique evolutionary lineages of R. muscosa occupy each mountain range in southern California and should be managed separately. The captive breeding program at Dark Canyon is promising, although mitigating the loss of neutral genetic diversity relative to the natural population might require additional breeding frogs.