AIMS: To examine associations between amount and frequency of alcohol consumption, and Type 2 diabetes. METHODS: A prospective study of 36 527 adults aged 40-69 at baseline. Incident cases of Type 2 diabetes were identified by questionnaire 4 years later. Sex-specific logistic regression models, adjusting for country of birth, dietary glycaemic index, energy intake and age, and in a second model body mass index (BMI) and waist-hip ratio (WHR), were used. RESULTS: Diabetes status was ascertained for 31 422 (86%) participants, and 362 cases identified. Former drinkers had higher risks than lifetime abstainers. Female drinkers had lower risk than lifetime abstainers (ORs < 10 g/day 0.54, 95% CI 0.36-0.82; 10-19.9 g/day 0.57, 0.34-0.94; > or = 20 g/day 0.46, 0.24-0.88, P trend = 0.005). There was no relationship after adjustment for body size. For men, a weak inverse association was observed after adjustment for body size (ORs relative to lifetime abstainers: < 10 g/day 1.56, 0.95-2.55; 10-19.9 g/day 1.21, 0.69-2.10; 20-29.9 g/day 0.80, 0.40-1.60; = 30 g/day 0.86, 0.50-1.58, P trend = 0.036). Wine was the only beverage for which an inverse association was observed. Compared with men who did not drink in the week before baseline, men who drank > or = 210 g over 1-3 days had an increased risk of diabetes (OR 5.21, 1.79-15.19), while the same amount over more days did not increase risk. CONCLUSIONS: Total alcohol intake was associated with reduced risk only in women. Alcohol from wine was associated with reduced risk of Type 2 diabetes. A high daily intake of alcohol, even on only 1-3 days a week, may increase the risk of diabetes in men.