This paper proposes a standard definition of endometriotic disease for epidemiologic studies and suggests subject-selection strategies to increase the validity of clinic- or population-based studies of the disease. Although endometriosis can be defined simply as the presence of ectopic endometrial tissue, emerging evidence indicates that to be pathologic, such tissue must persist and progress. The proposed disease definition incorporates the concepts of persistence and progression, and its use may increase the likelihood of observing true associations in etiologic studies. Potential threats to validity of substantial magnitude exist in both clinic- and population-based epidemiologic studies of endometriosis. In clinic-based studies, control subjects (infertility clinic patients, women delivering infants, or women undergoing tubal ligation) often are not representative of the population from which the cases arose, and bias can be considerable for behavioral and hormone-related exposures. In population-based studies, substantial case underascertainment may exist, and diagnosed cases may be a biased sample of all potential cases in the population. Although neither the ideal design nor the ideal case and control groups are likely to be achievable in epidemiologic studies of endometriosis, the subject-selection strategies suggested may improve the validity of studies that are obliged to depart from the ideal.