Information on drivers of dispersal is critical for wildlife conservation but is rare for long-lived marine mammal species with large geographic ranges. We fit multi-state mark-recapture models to resighting data of 369 known-aged Steller sea lion (Eumetopias jubatus) females marked as pups on their natal rookeries in southeastern Alaska from 1994-2005 and monitored from 2001-15. We estimated probabilities of females being first observed parous at their natal site (natal philopatry), and of not moving breeding sites among years (breeding philopatry) at large (> 400 km, all five rookeries in southeastern Alaska) and small (< 4 km, all islands within the largest rookery, Forrester Island Complex, F) spatial scales. At the rookery scale, natal philopatry was moderately high (0.776-0.859) for most rookeries and breeding philopatry was nearly 1, with < 3% of females switching breeding rookeries between years. At more populous islands at F, natal philopatry was 0.500-0.684 versus 0.295-0.437 at less populous islands, and breeding philopatry was 0.919-0.926 versus 0.604-0.858. At both spatial scales, the probability of pupping at a non-natal site increased with population size of, and declined with distance from, the destination site. Natal philopatry of < 1 would increase gene flow, improve population resilience, and promote population recovery after decline in a heterogeneous environment. Very high breeding philopatry suggests that familiarity with neighboring females and knowledge of the breeding site (the topography of pupping sites and nearby foraging locations) may be a critical component to reproductive strategies of sea lions.