OBJECTIVE To validate a self-administrable version of the SMAST questionnaire for diagnosing alcoholism at the primary care level. METHODS A descriptive study of a crossover type. METHODS An urban health centre. METHODS 497 patients over 15, selected at random from the log-book of daily demand, answered the questionnaire. 46 patients with 2 or more positive replies concerning alcoholism (SMAST positive) were invited to a subsequent appointment, as were 40 with a lower score (SMAST negative). 30 from each group attended. RESULTS The presence of alcoholism was examined by looking at the compliance with the diagnostic criteria of the DSM-III-R and/or the existence of alcohol consumption equal to or above 30 grams a day in women and 60 in men, in those patients who attended the subsequent appointment. 20 SMAST++ positives and two negatives were diagnosed as alcoholics. There was good correlation between daily alcohol consumption in grams and SMAST scoring (r = 0.60; CI's 95%: 0.41-0.74). SMAST showed 50.48% sensitivity, 96.48% specificity and positive and negative predictive values of 66.67% and 93.33%, respectively. By means of a logistical regression analysis, it was clearly shown that the SMAST questions which best forecast the existence of alcoholism were the first two. CONCLUSIONS The SMAST questionnaire is a very specific test for diagnosing alcoholism within primary care, although its only moderate sensitivity does not recommend it for screening purposes.