Records of grizzly bear (Ursus arctos) deaths are currently used by managers to indicate trends in actual grizzly bear mortality and to judge the effectiveness of management. Two assumptions underlie these current uses: first, that recorded mortality is an unbiased indicator of actual mortality, and second, that changes in mortality after implementation of management strategies are sufficient grounds to infer the effects of management. I examined the defensibility of these 2 assumptions relative to alternate explanations, circumstantial evidence, and the potential costs of error. The potentially complex relation between actual and recorded mortality, as currently tallied and used, was reason to expect that the association between these 2 values would be weak. This expectation was supported by the prevalence (60-76%) of radio-marked bears among recorded deaths, the variation in apparent likelihood of documentation among causes of death, and variation in the prevalence of different causes over time. For these reasons, recorded mortality is likely to be an unreliable indicator of actual mortality. Use of whitebark pine (Pinus albicaulis) seeds by grizzly bears had a major effect on annual variation in recorded mortality. Low numbers of recorded deaths, 1984-92, were attributable to relatively frequent large whitebark pine seed crops. There was little or no residual trend potentially ascribed to management intervention during 1976-92. Management intervention was probably responsible for observed changes in recorded causes of death and stabilized recorded mortality over the period covered by this analysis.