The reliability of 2 methods to estimate weekly alcohol consumption, based on data collected from 2 household surveys in Boston, [Massachusetts, USA] was assessed. In the 1st method, 361 respondents (166 men) met with an interviewer to report their drinking during the preceding week, while in the 2nd method, 270 respondents (111 men) kept their own daily drinking record. Number of drinking events, amount of alcohol consumed and average amount of alcohol consumed at 1 event as reported by interviewees and record keepers for each of the 3 wk of the study were compared. There were no significant differences between the mean levels of alcohol consumption per event estimated by the 2 methods. Respondents using the daily record reported fewer drinking events and less alcohol consumed during week 3 than during weeks 1 and 2. The interview method is the more reliable.