Three hundred and six adult medical and family practice inpatients at a university-affiliated teaching hospital were assessed for problem drinking using: (1) a short questionnaire containing the CAGE (a 4-question, self-report screening measure), and (2) a standardized review of medical records (an examination of specific blood chemistries and the physicians' and nurses' notes). Each patient completed one of two randomly assigned questionnaires containing the CAGE within the first 48 h of admission: a "Health Habits Questionnaire" or an "Alcohol Questionnaire." No significant differences were found between the two questionnaires with regard to the number of CAGE items endorsed. Overall, 16.9% of the subjects endorsed two or more CAGE items. Using the chart screening method, 11.4% of the total subjects were considered "probable" or "definite" problem drinkers. The screener's reading of the physicians' notes was the variable most important in determining overall chart screening assessment. The correlation between the two screening measures was relatively low (r = .38, p less than .001), suggesting that the two measures identify different populations of problem drinkers.