Using laboratory experiments, simulation models, and analytical techniques, we examined the impact of dispersal on the mean densities of patchily distributed populations. Even when dispersal leads to no net additions or removals of individuals from a population, it may nonetheless increase mean population densities if the net immigration rate is positive when populations are growing and negative when they are declining. As a model system for exploring this phenomenon, we used the yeastlike fungus Aureobasidium pullulans. In a laboratory experiment, we showed that dispersal can both ensure persistence and increase mean population densities even when dispersal among populations causes no direct addition or loss of fungal cells. From the laboratory data, we constructed a plausible model of A. pullulans dynamics among apple leaves within an orchard. This simulation model demonstrated that the effect of dispersal on mean densities is enhanced by three factors: weak density dependence of the dynamics within populations, high environmental variability affecting population growth rates, and lack of synchrony among the fluctuations of populations. Using an analytical model, we showed that the underlying mechanisms for this phenomenon are general, suggesting that a large effect of dispersal on mean population densities is possible in many natural systems.