AIM: We investigated how Pleistocene refugia and recent (~12,000 years ago) sea level incursions shaped genetic differentiation in mainland and island populations of the Scinax perpusillus treefrog group. LOCATION: Brazilian Atlantic Forest, São Paulo state, southeastern Brazil. METHODS: Using mitochondrial and microsatellite loci, we surveyed population structure and genetic diversity in Scinax from five mainland populations and three land-bridge islands to understand the roles of Pleistocene forest fragmentation and sea level incursions on genetic differentiation. We calculated metrics of relatedness and genetic diversity to assess whether island populations exhibit signatures of genetic drift and isolation. Two of the three island populations in this study have previously been described as new species based on a combination of distinct morphological and behavioural characters, thus we used the molecular datasets to determine whether phenotypic change is consistent with genetic differentiation. RESULTS: Our analyses recovered three distinct lineages or demes comprised of northern mainland São Paulo populations, southern mainland São Paulo populations, and one divergent island population. The two remaining island populations clustered with samples from adjacent mainland populations. Estimates of allelic richness were significantly lower, and estimates of relatedness were significantly higher, in island populations relative to their mainland counterparts. MAIN CONCLUSIONS: Fine-scale genetic structure across mainland populations indicates the possible existence of local refugia within São Paulo state, underscoring the small geographic scale at which populations diverge in this species rich region of the Atlantic Coastal Forest. Variation in genetic signatures across the three islands indicates that the populations experienced different demographic processes after marine incursions fragmented the distribution of Scinax perpusillus. Genetic signatures of inbreeding and drift in some island populations indicate that small population sizes coupled with strong ecological selection may be important evolutionary forces driving speciation on land-bridge islands.