The Japanese brown bear (Ursus arctos yesoensis) lives only on Hokkaido Island, where human–bear conflicts are common, but the status of bear populations are unknown. The Shiretoko Peninsula in northeastern Hokkaido supports one small population. The distal half of the 1,000-km² peninsula has been protected as a wildlife reserve since the 1980s. Reduction in harvest and frequent sightings of bears suggests that the population has been increasing since that time. On the other hand, human–bear conflicts and management kills are not uncommon in the proximal half, and occur occasionally even in the protected area. Also, unreported harvest might be substantial. Lacking unbiased survival data from marked bears, we modeled the Shiretoko population using female-only matrix models in a series of “what if?” scenarios that we believe bounded reality. We assumed, alternatively, that true mortality equaled reported kills (best case) and that it double reported kills (worst case). During 1985–2004, known mortality was 12.9 animals/year (1.9 adult females/year), and we doubled this for our scenarios that assumed the detection of mortality was 50%. We estimated reproductive rates by monitoring marked and unmarked — but individually recognizable — females, and estimating transition rates among reproductive states. From 1990 to 2004, 13 identifiable females were observed in west-central Shiretoko for a total of 67 female-years, with 51 unduplicated cubs. Mean litter size was 1.594 (n = 32), and reproductive rate was estimated as 0.604 cubs/female/year (0.302 for female cubs). Our model suggested that if the female population size (N_f) had been 200, it would have been increasing regardless of whether known mortality was 50% or 100% of true mortality. Under the assumption of N_f = 150 and a mortality-detection rate of 50% (worst case), projections suggested the possibility of either an increase or decline (95% CI of λ = 0.986–1.150). The current harvest rate remains unknown because the population size remains unknown, but we could not reject the possibility that current harvest rates are beyond the sustainable level. We call for more efforts to closely monitor the population and harvest, and to promote preventive measures to reduce human–bear conflicts.