The internal validity and the reliability of the CAGE test (cutting down, annoyance by criticism, guilty feeling, and eye-openers) was examined for residents of Bucaramanga (northeastern Colombia).One hundred nine subjects between 18 and 60 years were given the CAGE test on two occasions. Each subject was surveyed for a suite of biological, socioeconomic and dietary variables. The internal consistency of the test was evaluated with Cronbach's alpha, and the reliability with Spearman's correlation coefficient, r, and Cohen's kappa. A binomial regression produced association measures among the risk of alcoholism and the biological, socioeconomic and dietary variables.The risk of alcoholism in the subjects was of 15.6% (CI 9.0-22.5%). The men presented a risk of alcoholism three times that of women (p=0.02). Cronbach's alpha was low for all test questions (min=0.08; max=0.28). In the first application of CAGE, the average for Cronbach's alpha was 0.20, in the second application 0.27. The overall Spearman's r was of 0.65 (CI 0.53-0.75), but with significant sexual difference: 0.58 (CI 0.36-0.73) for men and 0.80 (CI 0.68-0.88) for women. Cohen's kappas ranged from moderate to good. Any biological, socioeconomic and dietary variables was associated positively with the risk of alcoholism.The low internal consistency of CAGE is due to the multiple aspects of the test; however, because of the high reliability and stability of the association measures, the test was recommended for continued use in epidemiological investigations.