We investigated factors associated with the distribution of grizzly bears (Ursus arctos horribilis) in 1850 and their extirpation during 1850–1920 and 1920–1970 in the contiguous United States. We used autologistic regression to describe relations between grizzly bear range in 1850, 1920, and 1970 and potential explanatory factors specified for a comprehensive grid of cells, each 900 km² in size. We also related persistence, 1920–1970, to range size and shape. Grizzly bear range in 1850 was positively related to occurrence in mountainous ecoregions and the ranges of oaks (Quercus spp.), piñon pines (Pinus edulis and P. monophylla), whitebark pine (P. albicaulis), and bison (Bos bison) and negatively related to occurrence in prairie and hot desert ecoregions. Relations with salmon (Oncorynchus spp.) range and human factors were complex. Persistence of grizzly bear range, 1850–1970, was positively related to occurrence in the Rocky Mountains, whitebark pine range, and local size of grizzly bear range at the beginning of each period, and negatively related to number of humans and the ranges of bison, salmon, and piñon pines. We speculate that foods affected persistence primarily by influencing the frequency of contact between humans and bears. With respect to current conservation, grizzly bears survived from 1920 to 1970 most often where ranges at the beginning of this period were either larger than 20,000 km² or larger than 7,000 km² but with a ratio of perimeter to area of <2. Without reductions in human lethality after 1970, there would have been no chance that core grizzly bear range would be as extensive as it is now. Although grizzly bear range in the Yellowstone region is currently the most robust of any to potential future increases in human lethality, bears in this region are threatened by the loss of whitebark pine.