It has long been recognized that, as populations increase in density, ecological processes affecting growth and survival reduce per capita recruitment in the next generation. In contrast to the evidence for such "compensatory" density dependence, the alternative "depensatory" process (reduced per capita recruitment at low density) has proven more difficult to demonstrate in the field. To test for such depensation, we measured the spawner-recruit relationship over five decades for a sockeye salmon (Oncorhynchus nerka) population in Alaska breeding in high-quality, unaltered habitat. Twenty-five years of detailed estimates of predation by brown bears, Ursus arctos, revealed strong density dependence in predation rate; the bears killed ca. 80% of the salmon in years of low salmon spawning abundance. Nevertheless, the reconstructed spawner-recruit relationship, adjusted to include salmon intercepted in the commercial fishery, provided no evidence of demographic depensation. That is, in years when few salmon returned and the great majority were killed by bears, the few that spawned were successful enough that the population remained highly productive, even when those killed by bears were included as potential spawners. We conclude that the high quality of breeding habitat at this site and the productive nature of semelparous Pacific salmon allowed this population to avoid the hypothesized depressed recruitment from depensatory processes expected at low density. The observed lack of demographic depensation is encouraging from a conservation standpoint because it implies that depleted populations may have the potential to rebound successfully given suitable spawning and rearing habitat, even in the presence of strong predation pressure.