Many metapopulation models assume rules of population connectivity that are implicitly based on what we know about within-population dispersal, but especially for vertebrates, few data exist to assess whether interpopulation dispersal is just within-population dispersal "scaled up." We extended existing multi-stratum mark-release-recapture models to incorporate the robust design, allowing us to compare patterns of within- and between-population movement in the banner-tailed kangaroo rat (Dipodomys spectabilis). Movement was rare among eight populations separated by only a few hundred meters: seven years of twice-annual sampling captured >1200 individuals but only 26 interpopulation dispersers. We developed a program that implemented models with parameters for capture, survival, and interpopulation movement probability and that evaluated competing hypotheses in a model selection framework. We evaluated variants of the island, stepping-stone, and isolation-by-distance models of interpopulation movement, incorporating effects of age, season, and habitat (short or tall grass). For both sexes, QAICc values clearly favored isolation-by-distance models, or models combining the effects of isolation by distance and habitat. Models with probability of dispersal expressed as linear-logistic functions of distance and as negative exponentials of distance fit the data equally well. Interpopulation movement probabilities were similar among sexes (perhaps slightly biased toward females), greater for juveniles than adults (especially for females), and greater before than during the breeding season (especially for females). These patterns resemble those previously described for within-population dispersal in this species, which we interpret as indicating that the same processes initiate both within- and between-population dispersal.