Epidemiological studies of multiple sclerosis (MS) have shown the importance of genetic susceptibility factors that are modified by as yet unknown environmental influences. The often-cited interrelation between prevalence and latitude is no longer viable, with the important but unexplained exception of Australia and New Zealand. The standardization and increasing use of uniform diagnostic criteria lend new credibility to the results of MS surveys. More precise criteria are offered for symptoms of disease onset, Devic's syndrome, and progressive disease. The role of magnetic resonance imaging in the diagnosis of MS, and the pitfalls of its abuse, are reviewed. Reports of epidemics of MS are examined and discarded because they are based on the biologically meaningless date of diagnosis or that of onset, rather than when MS was probably acquired, that is, before puberty. The concept of onset-adjusted prevalence suggests that patients who are symptomatic should be included retrospectively in epidemiological surveys, even before they have been formally diagnosed as having MS, but individuals who were symptomatic before moving to a study area should not. The importance of ethnic homogeneity of patients and control subjects cannot be overemphasized in prevalence and risk factor studies. In the latter, close attention must be paid to biological plausibility and to prudent statistical interpretation. The hypothesis of the MS trait describes a systemic, asymptomatic condition that does not affect the nervous system, and may explain the low concordance of the disease in monozygotic twins.