The retrospective or case-control method of epidemiologic investigation is described and compared with the prospective (cohort) method. Both methods have the weakness of being observational or nonexperimental in nature. 2 methods employed in retrospective studies for reducing (but never eliminating) the likelihood of spurious inferences through indirect association are matching controls to cases with respect to possible confounding factors and statistical adjustment. Some etiologic factors or susceptibility-associated factors that have been identified by studies of this type are smoking in relation to lung cancer and other conditions; oral contraception in relation to intravascular thrombosis; features of the reproductive history in relation to breast and cervical cancer; and maternal rubella in relation to congenital defects. The results of properly designed and conducted retrospective studies rarely disagree to any extent with those of equally well-conducted prospective studies. The retrospective method is both valid and useful. (authors)