The objective of this cross-sectional study was to clarify the association between alcohol and obesity using data from 106 182 adults in England and Scotland (46·7 % male; mean 46·9 (sd 16·9) years). Trained interviewers asked participants about alcohol intake. Obesity was defined as BMI≥30 kg/m2. Potential confounders included age, sex, smoking, physical activity, longstanding illness, psychological distress and socioeconomic status. Compared with those who drank at least five times a week, obesity risk was 1·21 (95 % CI 1·15, 1·27) in those who drank one to four times a week, 1·53 (95 % CI 1·43, 1·62) in those who drank one to two times a month, 1·61 (95 % CI 1·52, 1·71) in those who drank less than once every couple of months, 1·34 (95 % CI 1·23, 1·47) in those who were former drinkers, and 1·03 (95 % CI 0·95, 1·11) in those who were never drinkers. Compared with those who drank a harmful volume, obesity risk was 0·78 (95 % CI 0·68, 0·90) in those who drank within guidelines, 0·69 (95 % CI 0·54, 0·88) in former drinkers and 0·50 (95 % CI 0·40, 0·63) in never drinkers; and, these associations were biased away from the null after adjustment for drinking volume. Abstinence was associated with increased risk of obesity in women. These data suggest that the association between drinking frequency and obesity is bell-shaped, with obesity risk not significantly different in those who drink most often and never drinkers. Drinking volume has a positive confounding effect on the association between drinking frequency and obesity, which may help explain the conflicting findings of other studies.