Estimating the rate and scale of dispersal is essential for predicting the dynamics of fragmented populations, yet empirical estimates are typically imprecise and often negatively biased. We maximized detection of dispersal events between small, subdivided populations of water voles (Arvicola terrestris) using a novel method that combined direct capture–mark–recapture with microsatellite genotyping to identify parents and offspring in different populations and hence infer dispersal. We validated the method using individuals known from trapping data to have dispersed between populations. Local populations were linked by high rates of juvenile dispersal but much lower levels of adult dispersal. In the spring breeding population, 19% of females and 33% of males had left their natal population of the previous year. The average interpopulation dispersal distance was 1.8 km (range 0.3–5.2 km). Overall, patterns of dispersal fitted a negative exponential function. Information from genotyping increased the estimated rate and scale of dispersal by three- and twofold, respectively, and hence represents a powerful tool to provide more realistic estimates of dispersal parameters.