The concept of tolerance to ethanol has evolved gradually over the past two centuries, and all of the basic clinical features, as they are now understood, have been clearly recognized for nearly 100 years. The basic mechanisms involved in central nervous system tolerance, however, have been elucidated only in the past 20 to 30 years. Little progress was made as long as tolerance was viewed as a purely cellular or physiological adaptation to alcohol, and researchers used overly simple paradigms based on mere exposure to the drug. With the recognition that learning, both operant and associative, can play a major role in the development of tolerance to alcohol and cross-tolerance to other drugs, a radical change in research approaches became possible. Most of the neural mechanisms related to learning and memory are now known to be involved in the development and retention of tolerance, and the simplistic models used in earlier research must now be abandoned. Nevertheless, a review of the history of past research points to a number of important lessons for future work, including the following: (1) many of the present concepts were enunciated by astute observers many decades ago, and research was hindered because this older literature was forgotten; (2) for many decades progress was slow because of a narrow focus on specific techniques, questions, and hypotheses that overlooked important research in related disciplines; (3) the course of research is often irregular, and past questions may have to be revisited with new approaches--but these are more likely to be fruitful if based on knowledge of past history; and (4) excellent researchers often obtain apparently contradictory findings, but the disagreements may hold the key to deeper understanding of the phenomena, and should not be brushed over by ignoring the minority findings and interpretations. As in all scientific research, the most important requirement for major progress is the formulation of good questions or hypotheses: the results yielded by the best available techniques can be only as good as the questions they are meant to answer.