We assessed the effects of long-term habitat fragmentation on genetic (random amplified polymorphic DNA) diversity in 11 Peromyscus maniculatus populations in the Lake Superior watershed. We analysed genetic structure at two spatial scales and the effect of island size and isolation on genetic diversity. At the regional scale, island populations differed from mainland populations (FST = 0.36), but mainland populations did not differ from each other (FST = 0.01). At the local scale, populations of the main island of Isle Royale differed from adjacent islet populations (P < 0.001; Monte Carlo approximation of Fisher’s exact test), but not from each other (combined P = 0.63). Although geographical distance and genetic distance were positively correlated (P < 0.01; Mantel test), cluster analysis revealed some inconsistencies. Finally, genetic diversity was inversely related to isolation (P = 0.01), but had an unexpectedly negative relationship with island area (P = 0.03). The genetic structure of P. maniculatus populations in portions of the Lake Superior watershed appears to have been affected by long-term habitat fragmentation.