A major criticism of the hypothesis which states that non–drinkers are mainly a selected group of ex–drinkers who have given up for reasons of ill health, particularly cardiovascular disease, is that such a high selective migration seems unlikely. It is emphasized that ex–drinking status is related to a wide range of disorders although it is particularly marked for cardiovascular disorders. In addition, mortality in middle–aged men is affected most strongly by cardiovascular disease. Virtually all previous prospective studies have not paid sufficient attention to the reasons for people changing their drinking behaviour, or to the burden of disease in non–drinkers. The increase in HDL–cholesterol associated with light drinking (the ‘narrow window’ of protective effect) is small and unlikely to account for significantly reduced coronary heart disease mortality. Studies claiming a protective effect for alcohol are entreated to examine closely the burden of disease in their alcohol intake groups, particularly the non– and occasional drinkers.