The movement of organisms between subdivided populations is considered a key influence on the persistence of species in modified landscapes. In particular, the ability to recolonize ‘empty’ fragments of habitat is directly relevant to conservation management, and to understanding the link between pattern and process in metapopulations. We studied the movement and recolonization ability of the bush rat, Rattus fuscipes, in a highly fragmented agricultural landscape in south-western Victoria, Australia. Populations were monitored in seven small (<2.5 ha) and two large (>49 ha) forest fragments before removal of all residents from four of the small fragments. Subsequent monitoring (for up to 16 months) allowed the detection of colonizing individuals, and comparisons between ‘experimental’ and ‘control’ fragments. Rattus fuscipes readily moved between fragments and successfully recolonized (i.e. both males and females arrived) two of four fragments in which extinctions were simulated. A single male moved into a third experimental fragment. In one fragment, new animals were detected 1 month after the removal of residents, indicating that recolonization can occur rapidly. Dispersers were not a random sample of the population: although both males and females, and adults and sub-adults were recorded dispersing, adult males in reproductive condition predominated. Functional connectivity appears to be high for R. fuscipes in the study landscape. Results from this manipulative experiment provide direct empirical evidence that a capacity for movement allows recolonization of fragments of suitable habitat and is a key process responsible for species persistence in fragmented landscapes, as predicted by theory.