Events that occur during the pelagic larval stage are thought to be important determinants of reef fish population dynamics. Recent research contradicts the early paradigm of larvae being advected as passive propagules and indicates that many late stage larvae have well-developed sensory and locomotory capabilities. Whether and how larvae use these capabilities to influence their dispersal is unknown. We compare alternative hypotheses regarding larval behavior. Contrary to the trend in dispersal modeling, we focus on larval biology rather than physical oceanographic considerations. Specifically, we present two streams of models: one that describes a return-based strategy and one in which dispersal is a central component. The models depend on different sets of behavioral assumptions for a pomacentrid species and for acanthurids, two groups with contrasting early life histories. Whether dispersal or return-based strategies are favored depends on the efficiency and sustainability of larval swimming methods and the environmental conditions experienced during dispersal. We argue that dispersal models should consider a variety of behavioral hypotheses and that the sensitivity of results to the behavioral assumptions made should be quantified.