The occurrence of idiosyncratic drug hepatotoxicity is a major problem in all phases of clinical drug development and the most frequent cause of post-marketing warnings and withdrawals. This review examines the clinical signatures of this problem, signals predictive of its occurrence (particularly of more frequent, reversible, low-grade injury) and the role of monitoring in prevention by examining several recent examples (for example, troglitazone). In addition, the failure of preclinical toxicology to predict idiosyncratic reactions, and what can be done to improve this problem, is discussed. Finally, our current understanding of the pathophysiology of experimental drug hepatotoxicity is examined, focusing on acetaminophen, particularly with respect to the role of the innate immune system and control of cell-death pathways, which might provide targets for exploration and identification of risk factors and mechanisms in humans.