Habitat fragmentation threatens biodiversity by disrupting dispersal. The mechanisms and consequences of this disruption are controversial, primarily because most organisms are difficult to track. We examined the effect of habitat corridors on long-distance dispersal of seeds by birds, and tested whether small-scale (G20 meters) movements of birds could be scaled up to predict dispersal of seeds across hundreds of meters in eight experimentally fragmented landscapes. A simulation model accurately predicted the observed pattern of seed rain and revealed that corridors functioned through edgefollowing behavior of birds. Our study shows how models based on easily observed behaviors can be scaled up to predict landscape-level processes.Habitat fragmentation poses a widespread threat to biodiversity by disrupting the dispersal of organisms (1, 2). Corridors-narrow strips of habitat that join patches of similar habitatare thought to provide a general solution by restoring dispersal among patches, thereby increasing gene flow and reducing the probability of local extinctions (3, 4). Yet corridors are controversial (5, 6). Their efficacy can vary greatly among systems, depending on the complex interaction between disperser behavior and landscape structure (3, 7). Controversy about corridors has been difficult to resolve because corridors operate at a landscape scale, where both experimental and observational (tracking) studies are difficult (8).Here we test and validate an alternative approach to examine corridor function: individualbased behavioral models (9). Our goal was to predict corridor effects on long-distance (9250 m) dispersal of seeds by birds. We collected data on small-scale (G20 m) movements of seeddispersing birds in experimental landscapes and used these movements to parameterize a model that predicted the effects of corridors on seed dispersal at the landscape scale. We then tested our model using data on actual seed rain from the same large-scale experimental landscapes. Importantly, our model links observations of local bird behavior to population-level impacts on the recruitment of plants, showing how corridor effects on one taxon can affect the other. More generally, the control and replication provided by our experimental landscapes allow a rigorous validation of individual-based models, which are widely applicable to other systems.Our experimental landscapes were designed to test two alternative hypotheses about how corridors function. The traditional corridor hypothesis posits that corridors act as dispersal conduits, channeling organisms between connected patches (3, 4). The drift-fence hypothesis posits that corridors intercept organisms dispersing through matrix habitat and direct them into associated patches, thereby increasing colonization of patches with corridors, regardless of whether the corridors connect patches (10, 11). These hypotheses are not mutually exclusive.Each of our experimental landscapes contained five patches of regenerating vegetation in a matrix of mature pine forest ( F...