Over a 9-yr period, we studied dispersal of young banded Savannah Sparrows (Passerculus sandwichensis) from their natal nest to the site where they first bred 1 yr later in a population on an isolated archipelago in the Bay of Fundy, Canada. On a broad geographic scale, young birds were highly philopatric, returning from wintering grounds several thousand kilometers away to nest on the same island and often in the same field where they had hatched the year before. In some cases, birds chose nest sites within a few meters of their natal nest. The median dispersal distance between a bird's natal nest and its first nest as an adult was 228 m, a distance equivalent to about six times the diameter of an average territory. Nearly three-quarters of the breeding birds in the study population had been banded as nestlings or fledglings within a 10-ha area on one island, which suggests that most birds in the population originated within the 127-ha archipelago. Within the archipelago, males and females dispersed similar distances from their natal site. There was no correlation between natal dispersal distances of parents and their offspring, nor was there a correlation between natal dispersal distances of siblings, which indicated that natal dispersal has low heritability. Sex, hatching date, fledging mass, and population densities during the previous and current years were all poor predictors of natal dispersal distance. Males that were strongly philopatric recruited significantly more offspring during their lifetime than males raised outside the study area, although there were no differences between philopatric and dispersing females. Birds tended to shift to distinct parts of the island to breed if their parents of the opposite sex still occupied the territory where they had hatched. Dispersal was not affected by return of same-sex parents. Distances between nests of siblings raised the same year were farther apart than expected by chance, based on Monte Carlo simulations. Although many birds in the population had the opportunity to pair with kin, some mechanism, yet to be determined, enabled birds to avoid inbreeding: in 1073 nesting attempts involving birds of known parentage, no individuals were known to have paired with close relatives (coefficient of kinship >0.125). Because complete inbreeding avoidance occurred in <1% of Monte Carlo simulations, the absence of inbreeding among Kent Island Savannah Sparrows is unlikely to be due to chance. Understanding natal dispersal in birds requires a combination of models: ecogenetic models at broad geographic scales (e.g., adaptation to local environments), ecological constraints and neutral models at smaller spatial scales (e.g., unavailability of territories within particular habitats), and genetic models (e.g., inbreeding avoidance).