The question of how much time a foraging herbivore should spend in a patch of food poses a central challenge in classical foraging theory. However, there remains uncertainty about the relevance of the patch paradigm to foraging decisions by large herbivores. This paper examines evidence for successfully predicting and quantifying patch departure decisions for large mammalian herbivores foraging across several spatial and temporal scales. Departure decisions at fine scales are influenced by tradeoffs between maximizing intake rate and food quality. Classical models for departure decisions at larger spatial scales, particularly the marginal value theorem, appear inadequate. We advocate exploring alternative models for predictions of residence time at the patch scale.