Contradictory patterns of density-dependent animal dispersal can potentially be reconciled by integrating the conspecific attraction hypothesis with the traditional competition hypothesis. We propose a hypothesis that predicts a U-shaped relationship between density and both natal and breeding dispersal distance. Using 10 years of observations on a breeding colony of the Blue-footed Booby (Sula nebouxii), the hypothesis was confirmed by documenting simultaneous positive and negative density-dependent dispersal distances in natal and breeding dispersal of males and breeding dispersal of females within the colony. Point-pattern analyses demonstrated that the breeding sites of Blue-footed Boobies were highly aggregated in all years within a large study area, and aggregation presumably resulted in heterogeneity in patch density throughout the colony. As predicted, at moderate to high densities, dispersal distances showed positive density dependence, with individuals moving to lower density patches. In contrast, at low to moderate densities, dispersal distances showed negative density dependence, with individuals moving to higher density patches. In both sexes of the 1994 cohort, the higher the mean density in patches used by an individual over the long term (up to age 11 years), the fewer fledglings it produced. A positive effect of density on long-term reproductive success was not detected, possibly because birds that failed during pair formation or incubation were not sampled. Density of conspecifics may be an important influence on habitat selection of breeders, and dispersal may tend to carry individuals to patches where pair formation opportunities are better and negative effects of competition on reproductive success are reduced.