A primary goal of epidemiological research is to describe how the exposure of members of a population to certain risk factors influences quent disease incidence. Such a description is relatively straightforward for cohort studies, where exposed and unexposed members of the study group are followed forward in time so as to permit direct measurement of the relevant disease rates. Cohort studies of rare diseases are extremely costly and time consuming, however, since large groups must be assembled and followed up many years in order to collect enough cases for meaningful statistical analysis. The case-control study, wherein one compares cases and disease-free controls vis-a-vis exposure histories obtained by interview or other retrospective means, has been gaining favor in recent years as a valid and cost-effective alternative method (1). Case-control studies have contributed to the solution of important public health problems. This is exemplified ing endometrial cancer were more likely than controls to have received treatment with exogenous estrogens for menopausal symptoms (2, 3). How ever, the fierce debate that followed publication of these findings reminds us that such studies must be designed carefully in order to provide a true picture of the effect of exposure on incidence. Major issues that arise during the planning stages are:(a) the selection of subjects;(b) the choice of variables for measurement and analysis;(c) the extent to which comparabil ity of cases and controls is to be assured through stratification or matching; and (d) the size of the case and control samples. Recent developments in biostatistical theory and methodology have con tributed to the resolution of some of these design issues and have also