Using mtDNA variation in the kangaroo rat Dipodomys stephensi, we found no support for the hypothesis that a species with an historically restricted range will exhibit low levels of genetic polymorphism and little genetic structure. Dipodomys stephensi has long been restricted to a few interior coastal valleys in southern California encompassing an area of approximately 70 x 40 km; however, we found high levels of genetic variation over much of its range and significant genetic structure both within and between regions. We also found evidence for a recent range expansion. Dipodomys stephensi is a federally endangered species that is separated from D. panamintinus, its presumed sister taxon, by a mountain range to the north. We assessed genetic variation by sequencing 645 bases of the mitochondrial d-loop from 61 individuals sampled from 16 locations across the species range and rooted their relationship using two D. panamintinus individuals. Despite its limited geographic range, the level of mtDNA variation in D. stephensi is comparable to that of other rodents, including that of the more widely distributed D. panamintinus. This variation revealed significant regional differentiation. The northern, central, and southern regions of the range differ in both the level and the distribution of genetic variation. Phylogenetic analysis revealed that the center of the range contains the most diversity of lineages, including the most basal. In this region and in the north, most haplotypes were found at only a single location (25/29), or at a pair of nearby locations (3/29). In addition, related haplotypes clustered geographically. These results are consistent with long-term demographic stability characterized by limited dispersal and high local effective population size. Further support for this conclusion is the finding of unique diversity in two northern peripheral populations, Norco and Potrero Creek (PC). However, in sharp contrast, one haplotype (CC) was found at five of 11 central and northern locations and comprised 18% of individuals sampled. The atypical distribution of the CC haplotype reflected a pattern seen more strongly in the southern region. Here the CC haplotype comprised 69% of the sample and was found at all five sampling locations. Consequently, the southern region had very low genetic variability. We propose that this dominance of CC was probably due to a local population bottleneck that occurred during a recent range expansion into the southern region.