To understand the impact of various factors on the maintenance of genetic variation in natural populations, we need to focus on situations where at least some of these factors are removed or controlled. In this study, we used highly variable, presumably neutral, microsatellite and mtDNA markers to assess the nature of genetic variation in 14 island and two mainland populations of the Australian bush rat, where there is no migration between islands. Thus we are controlling for selection and gene flow. Both marker sets revealed low levels of diversity within the small island populations and extreme differentiation between populations. For six microsatellite loci, all of the small island populations had less genetic variation than the mainland populations; reduction in allelic diversity was more pronounced than loss of heterozygosity. Kangaroo Island, the large island population, had similar levels of diversity to the mainland populations. A 442 base pair (bp) section of the mtDNA control region was screened for variation by outgroup heteroduplex analysis/temperature gradient gel electrophoresis (OHA/TGGE). Only three of the 13 small island populations showed haplotypic diversity: Gambier (2), Waldegrave (2), and Eyere (3). The level of haplotypic diversity in the small island populations was similar to that on the mainland, most likely reflecting a recent population bottleneck on the mainland. In contrast, Kangaroo Island had 9 mtDNA haplotypes. The dominant factor influencing genetic diversity on the islands was island size. No correlation was detected between genetic diversity and the time since isolation or distance form the mainland. The combination of genetic drift within and complete isolation among the small island populations has resulted in rapid and extreme population divergence. Population pair-wise comparisons of allele frequency distributions showed significant differences for all populations for all loci (Fst = 0.11–0.84, Rst = 0.07–0.99). For the mtDNA control region, 92.6% of variation was apportioned between populations; only the Pearson islands shared a haplotype. Mantel tests of pair-wise genetic distance with pair-wise geographic distance showed no significant geographical clustering of haplotypes. However, population substructuring was detected within populations where sampling was conducted over a broader geographical range, as indicated by departures from Hardy-Weinberg equilibrium. Thus substructuring in the ancestral population cannot be ruled out. The dominant evolutionary forces on the islands, after the initial founder event, are stochastic population processes such as genetic drift and mutation.