We undertook a demographic analysis of the Yellowstone grizzly bears (Ursus arctos) to identify critical environmental factors controlling grizzly bear vital rates, and thereby to help evaluate the effectiveness of past management and to identify future conservation issues. We concluded that, within the limits of uncertainty implied by the available data and our methods of data analysis, the size of the Yellowstone grizzly bear population changed little from 1975 to 1995. We found that grizzly bear mortality rates are about double in years when the whitebark pine crop fails than in mast years, and that the population probably declines when the crop fails and increases in mast years. Our model suggests that natural variation in whitebark pine crop size over the last two decades explains more of the perceived fluctuations in Yellowstone grizzly population size than do other variables. Our analysis used demographic data from 202 radio-telemetered bears followed between 1975 and 1992 and accounted for whitebark pine (Pinus albicaulis) crop failures during 1993-1995. We used a maximum likelihood method to estimate demographic parameters and used the Akaike Information Criteria to judge the significance of various independent variables. We identified no independent variables correlated with grizzly bear fecundity. In order of importance, we found that grizzly bear mortality rates are correlated with season, whitebark pine crop size (mast vs. nonmast year), sex, management-trapping status (never management-trapped vs. management-trapped once or more), and age. The mortality rate of bears that were management-trapped at least once was almost double that of bears that were never management-trapped, implying a source/sink (i.e., never management-trapped/management-trapped) structure. The rate at which bears move between the source and sink, estimated as the management-trapping rate (h), is critical to estimating the finite rate of increase, I>I?. We quantified h by estimating the rate at which bears that have never been management-trapped are management-trapped for the first time. It differed across seasons, was higher in nonmast than mast years, and varied with age. We calculate that I>I?=1.00 from 1975 to 1983 (four mast and five nonmast years) and 1.02 from 1984 to 1995 (seven mast and five nonmast years). Overall, we find that I>I?=1.01A? 0.04 (mean A? 1 SE) from 1975 to 1995. Our models suggest that future management should concentrate on the threats to whitebark pine, such as those posed by white pine blister rust, global warming, and fire suppression. As is currently widely recognized by Yellowstone land managers, our model also suggests that future management must compensate for the increased grizzly bear mortality that is likely to be caused by an increasing number of humans in Yellowstone.