This review discusses research methodology in the relation between diet and disease. Medical research can be divided into two types: complex research (the detailed study of disease mechanisms using such methods as biochemistry and molecular genetics) and simple research (the investigation of the factors that cause or prevent disease using methods such as epidemiology, intervention trials, and analagous studies on animals). Although complex research has received the bulk of resources, the large majority of our information of practical value has come from simple research. This general principle is illustrated in the area of diet and disease by examples from various areas: selenium, carotenoids, and cancer; vitamin E, omega-3 fatty acids, and coronary heart disease; dietary fat and obesity; dietary sodium and hypertension; and alcohol and stroke. Discussion then turns to aspects of the design of cohort (prospective) studies. Because of problems of sample size and relative lack of diversity, previous studies often failed to give clear-cut results. Suggestions are made concerning the design of cohort studies, notably the use of much larger numbers of subjects and with greater diversity in their diets. The problem of confounding also is discussed. Lifestyle factors often cluster together but cohort studies may not have fully unraveled this.