The roles that diet and prey abundance play in habitat selection of Canada lynx (Lynx canadensis) in the contiguous United States is poorly understood. From 1998–2002, we back-tracked radiocollared lynx (6 F, 9 M) for a distance of 582 km and we located 86 kills in northwestern Montana, USA. Lynx preyed on 7 species that included blue grouse (Dendragapus obscurus), spruce grouse (Canachites canadensis), northern flying squirrel (Glaucomys sabrinus), red squirrel (Tamiasciurus hudsonicus), snowshoe hare (Lepus americanus), least weasel (Mustela nivalis), and white-tailed deer (Odocoileus virginianus). Snowshoe hares (69 kills) accounted for 96% (4-yr average, range = 94–99%) of prey biomass during the sample period. Red squirrels were the second-most-common prey (11 kills), but they only provided 2% biomass of the winter diet. Red squirrels contributed little to the lynx diet despite low hare densities. A logistic regression model of snowshoe hare, red squirrel, and grouse abundance, as indexed by the number of track crossings of use and available lynx back-tracks, was a significant (Wald statistic = 19.03, df = 3, P < 0.001) predictor of habitat use. As we expected, lynx (P < 0.001) selected use-areas with higher snowshoe hare abundance compared to random expectation. However, the red squirrel index had a weak (P = 0.087) negative relationship to lynx use, and grouse was nonsignificant (P = 0.432). Our results indicate that lynx in western Montana prey almost exclusively on snowshoe hares during the winter with little use of alternative prey. Thus, reductions in horizontal cover for hares would degrade lynx habitat.