Alcohol consumption has been linked with considerable mortality, and reduction of drinking, especially of heavy drinking, has been suggested as one of the main measures to reduce alcohol-attributable mortality. Aggregate-level studies including but not limited to natural experiments support this suggestion; however, causality cannot be established in ecological analysis. The results of individual-level cohort studies are ambiguous. On the other hand, randomized clinical trials with problem drinkers show that brief interventions leading to a reduction of average drinking also led to a reduction of all-cause mortality within 1 year. The results of these studies were pooled and a model for reduction of drinking in heavy drinkers and its consequences for all-cause mortality risk was estimated. Ceteris paribus, the higher the level of drinking, the stronger the effects of a given reduction. Implications for interventions and public health are discussed.