ABSTRACT Aims  To investigate the relationships between alcohol consumption and mortality and morbidity risk by specific causes. Design  Prospective cohort study. Setting  Twenty‐seven work‐places in West and Central Scotland. Participants  A total of 6000 men aged 21–64 years at screening in 1970–1973, median follow‐up 29 years. Measurements  Relative rates, using Cox's proportional hazard models, by weekly reported units of alcohol consumption for all cause, coronary heart disease (CHD), stroke, respiratory, digestive, liver disease and alcohol‐related causes of mortality and for specific causes of acute hospital admissions. Findings  Mortality risk was increased for men drinking 15–21 or more units per week for all causes, stroke, liver disease and alcohol‐related causes. For respiratory mortality, drinkers of 35 or more units had double the risk compared to non‐drinkers. CHD mortality showed increasing trends with consumption when adjusted for age and after full adjustment showed no clear patterns, although the 8–14 units group had a lower risk than non‐drinkers relative rate 0.81 (0.68–0.97). Hospital admissions had similar patterns to mortality for stroke and liver disease. Increased risk began at 8–14 units for alcohol‐related admissions, and at 15–21 units for respiratory admissions. Non‐drinkers had higher risks of having a CHD admission than drinkers and there were decreasing trends with increasing consumption (P = 0.019). Conclusions  Consumption of 15–21 units per week and over was associated with increased mortality from most causes and increased risk of hospital admissions from stroke, liver disease and respiratory diseases. Alcohol‐related admissions were raised from 8 to 14 units. Alcohol use may have been under‐reported in our study, but it was similar to other studies of the time. The apparent protective effect of alcohol with CHD admissions could be due partly to detrimental effects of heavy drinking causing sudden deaths. The associations, including that with respiratory disease, may arise from inadequate adjustment for confounding by other factors such as smoking.