Clinical and epidemiologic studies suggest that a number of factors are associated with the risk of migraine and precipitation of an attack. However, the degree to which causal associations can be inferred from reported studies is very limited and is a result of the methodological problems discussed throughout this review. The study of migraine in many ways parallels the pattern seen in early investigations of other conditions, such as rheumatoid arthritis and systemic lupus erythematosus, because a number of methodological problems had to be resolved in the study of these conditions before significant progress could be made. To achieve significant advances in the improvement of our understanding of the causes of migraine, a number of related issues must be addressed and resolved in future studies. Most noteworthy among these are Recognition of the probable heterogeneity of migraine, not merely in the manifestation of symptoms but, more importantly, in the existence of distinct etiologic subtypes. A number of findings suggest that some migraine subtypes are sensitive to certain precipitants, some appear to be a part of a more generalized constitutional disorder, and some are accompanied by a higher prevalence of migraine among family members. Efforts should be made in understanding the relationship between specific biochemical markers and traits (such as monoamine oxidase deficiency and tyramine sensitivity); precipitants related to the migraine attack; and epidemiologic characteristics such as age at onset and sex. Creation of a more precise, reliable, and practically useful definition of migraine. Without such a definition, it is difficult, if not impossible, to compare results between studies, to understand the relationship between risk factors and migraine subtypes, to understand properly associations identified in selected clinic populations, and, in general, to understand the epidemiology of migraine. More accurate characterization of the case group under study. More documentation of the age at onset, symptoms, frequency of attacks, and other characteristics related to migraine would be very useful to compare properly results between studies. Additional descriptive epidemiologic studies of migraine which would include reliable estimates of age-specific incidence, prevalence, remission rates, and natural history of the various migraine subtypes.(ABSTRACT TRUNCATED AT 400 WORDS)