During 1990–2000, I investigated patterns of fall migration and harvests of the Jackson elk herd. Specifically, I compared the timing of fall migration of the 2,500–4,500 elk (Cervus elaphus) that summered in Grand Teton National Park (GTNP) and wintered on the National Elk Refuge (NER) with elk that used other summer ranges by monitoring 512 migrations of radiocollared elk and by direct counts of elk arriving at the NER. Annual timing of migrations of GTNP elk was related positively to herbaceous standing crop (r² = 0.89) and residual standing crop (forage remaining at the end of the summer-fall grazing season, r² = 0.82) in GTNP. The number of elk on the NER the previous winter accounted for 39% of the residual variation in migration timing. Migrations of elk from higher elevation summer ranges were inversely related to cumulative snowfall through 10 November each fall, whereas elk from GTNP initiated migrations when snow depths averaged <2 cm. Vegetation offtake averaged 62% in GTNP when elk migrated to access superior foraging areas in the NER. Counter-intuitively, early migration of GTNP elk from their national park sanctuary, through a migratory corridor where they were heavily hunted, may be adaptive. Early migrants from GTNP to the NER experienced 4.8% harvest compared to 11.3% harvest (P = 0.037) of elk migrating after the median migration date each fall. Combined harvest and winter mortality rates were lower among early migrants (P < 0.01). I describe how early migration behavior may be selected over time. Such behavioral adaptation on elk ranges encompassing sanctuaries from hunter harvest challenge wildlife managers' skillfulness to manage elk numbers across heterogeneous topographic and jurisdictional landscapes.