The effects of forest fragmentation on bird populations have been studied primarily as static phenomena. Yet when forests are allowed to regenerate, local edge contrast and landscape matrix composition change with time, and we would expect fragmentation effects to change accordingly. Describing this process is critical for the conservation of avian species sensitive to forest fragmentation, including the Marbled Murrelet (Brachyramphus marmoratus), a seabird threatened by ongoing harvest of old-growth forest nesting habitat. We experimentally assessed potential murrelet nest predation probability in four regions of southwestern British Columbia, Canada. We compared the fates of 448 simulated murrelet nests at paired edge and interior treatments, at sites with "hard" edges (recent clearcuts), "soft" edges (regenerating forest), and natural edges (i.e., riparian areas). Motion-sensitive digital nest cameras enabled us to focus on known predators of real nests, and patterns of nest fates did not differ between real and simulated nests. Using information-theoretic model selection (AIC) with the combined data set (116 sites), we assessed effects at patch (approximately 13 ha), landscape (approximately 1700 ha), and regional (approximately 96000 ha) scales. Nest disturbance probability at hard edges was 2.5 times that of interior sites, but soft edges had less than half the disturbance probability of interiors. There was no edge effect at natural edges. At the landscape scale, overall avian disturbance risk declined by as much as 50% with increasing amounts of regenerating forest in the surrounding matrix. These results indicate that initially negative fragmentation effects decrease as forests regenerate, at both patch and landscape scales. There was no evidence that these patterns differed between regions. Predator surveys suggested that Steller's Jays drive patterns of nest predation risk at the regional scale. Assuming that corvids are the most important predators, larger reserves of habitat will lessen negative hard-edge effects. Smaller reserves should be embedded in a protective matrix of regenerating forest to reduce predation risk at both patch and landscape scales. Our results suggest that dynamic fragmentation effects are generalizable across widespread regions and can be broadly applied to both murrelet management and the conservation of old-growth forest-breeding birds in general.