There has been considerable emphasis on understanding isotopic discrimination for diet estimation in omnivores. However, discrimination may differ for carnivores, particularly species that consume lipid-rich diets. Here, we examined the potential implications of several factors when using stable isotopes to estimate the diets of bears, which can consume lipid-rich diets and, alternatively, fast for weeks to months. We conducted feeding trials with captive brown bears (Ursus arctos) and polar bears (Ursus maritimus). As dietary lipid content increased to ∼90%, we observed increasing differences between blood plasma and diets that had not been lipid extracted (∆¹³C_{tissue-bulk diet}) and slightly decreasing differences between plasma δ¹³C and lipid-extracted diet. Plasma Δ¹⁵N_{tissue-bulk diet} increased with increasing protein content for the four polar bears in this study and data for other mammals from previous studies that were fed purely carnivorous diets. Four adult and four yearling brown bears that fasted 120 d had plasma δ¹⁵N values that changed by <±2‰. Fasting bears exhibited no trend in plasma δ¹³C. Isotopic incorporation in red blood cells and whole blood was ≥6 mo in subadult and adult bears, which is considerably longer than previously measured in younger and smaller black bears (Ursus americanus). Our results suggest that short-term fasting in carnivores has minimal effects on δ¹³C and δ¹⁵N discrimination between predators and their prey but that dietary lipid content is an important factor directly affecting δ¹³C discrimination and indirectly affecting δ¹⁵N discrimination via the inverse relationship with dietary protein content.