Oceanographic features influence the transport and delivery of marine larvae, and physical retention mechanisms, such as eddies, can enhance self-recruitment (i.e. the return of larvae to their natal population). Knowledge of exact locations of hatching (origin) and settlement (arrival) of larvae of reef animals provides a means to compare observed patterns of self-recruitment ‘connectivity’ with those expected from water circulation patterns. Using parentage inference based on multiple sampling years in Moorea, French Polynesia, we describe spatial and temporal variation in self-recruitment of the anemonefish Amphiprion chrysopterus, evaluate the consistency of net dispersal distances of self-recruits against the null expectation of passive particle dispersal and test the hypothesis that larvae originating in certain reef habitats (lagoons and passes) would be retained and thus more likely to self-recruit than those originating on the outer (fore) reef. Estimates of known self-recruitment were consistent across the sampling years (~25–27% of sampled recruits). For most (88%) of these self-recruits, the net distance between hatching and settlement locations was within the maximum dispersal distance expected for a neutrally buoyant passive particle based on the longest duration of the larval dispersive phase and the average direction and speed of current flow around Moorea. Furthermore, a parent of a given body size on the outer (fore) reef of Moorea was less likely to produce self-recruits than those in passes. Our findings show that even a simple dispersal model based on net average flow and direction of alongshore currents can provide insight into landscape-scale retention patterns of reef fishes.