AIMS: To investigate the relationship between usual daily alcohol intake, beverage type and drinking frequency on cardiovascular (CVD) and coronary heart disease (CHD) mortality, accounting for systematic misclassification of intake. DESIGN: Prospective cohort study with mean follow-up of 11.4 years. Setting The Melbourne Collaborative Cohort Study, Australia. PARTICIPANTS: A total of 38 200 volunteers (23 044 women) aged 40-69 years at baseline (1990-1994). MEASUREMENTS: Self-reported alcohol intake using beverage-specific quantity-frequency questions (usual intake) and drinking diary for previous week. FINDINGS: Compared with life-time abstention, usual daily alcohol intake was associated with lower CVD and CHD mortality risk for women but not men. For women, the hazard ratio [HR (95% CI)] for CVD for those drinking > 20 g/day alcohol was 0.43 (0.19-0.95; P trend = 0.18), and for CHD, 0.19 (0.05-0.82; P trend = 0.24). Male former drinkers had over twice the mortality risk for CVD [HR = 2.58 (1.51-4.41)] and CHD [HR = 2.91 (1.59-5.33)]. Wine was the only beverage associated inversely with mortality for women. Compared with drinkers who consumed no alcohol in the week before baseline, drinking frequency was associated inversely with CVD and CHD mortality risk for men but not women. HR for men drinking 6-7 days/week was 0.49 (0.29-0.81; P trend = 0.02) for CVD, and 0.49 (0.26-0.92: P trend = 0.23) for CHD. CONCLUSIONS: Usual daily alcohol intake was associated with reduced CVD and CHD mortality for women but not men. This benefit appeared to be mainly from wine, although comparison of beverages was not possible. Drinking frequency was associated inversely with CVD and CHD death for men but not women.