A large number of prospective population studies from many countries have described a J- or U-shaped relation between alcohol intake and mortality. Both heavy drinkers and abstainers are at a higher risk of dying from all causes than individuals with light to moderate alcohol intake. This makes information to the public about sensible drinking limits more complex than, eg, that concerning smoking. The present paper aims at identifying upper thresholds for harmless alcohol intake. The review is mainly based on epidemiological evidence concerning somatic morbidity and mortality. It is concluded that the present Danish recommendations--14 drinks per week for women and 21 drinks per week for men--should be maintained. It is emphasized that these limits apply to adults who are at no risk of dependency and that they do not apply to pregnant women. Information about a potentially beneficial effect of a moderate alcohol intake should be reserved for individuals already at risk of coronary events.