Non-drinkers and heavy drinkers tend to have higher total and cardiovascular mortality rates than light or moderate drinkers. The finding is not disputed; it is the interpretation of this U-shaped curve that is controversial, and in particular the belief that light and moderate drinking protects against coronary heart disease. The British Regional Heart Study of middle-aged British men has shown that 70% of non-drinkers are ex-drinkers. Those ex-drinkers have high rates of doctor-diagnosed illnesses including heart disease, hypertension, diabetes and bronchitis as well as high prevalence rates of measured hypertension, obesity, current smoking and regular medical treatment. Over a five-year period men who were diagnosed as having heart disease, had multiple diagnoses or were put on regular medication had an increased likelihood of becoming non-drinkers or occasional drinkers. The study suggests a downward drift from heavy and moderate drinking towards non-drinking under the influence of accumulating ill health. The data strongly suggest that the observed alcohol-mortality relationships in prospective studies are produced by symptoms and disease present at the time of screening, and by the prior movement of men with such disorders into non-drinking or occasional drinking categories. The concept of a protective effect on mortality which ignores the dynamic relationship between ill health and drinking behaviour is likely to be ill-founded. A review of the major prospective studies reveals an inadequate exploration of the nature of non-drinkers, who are clearly unsuitable for use as a baseline in studies of the effects of alcohol on health.