Many alcoholics deny abuse. To screen greater samples for alcohol dependence, short questionnaires, e.g. the CAGE or MAST are often applied. Frequently laboratory parameters [i.e. 'alcohol markers', such as carbohydrate-deficient transferrin (CDT), gamma-glutamyl transferase or mean corpuscular volume of erythrocytes] are used to support the diagnosis of long-standing heavy alcohol consumption. In this study, the self-ratings (CAGE and MAST) were compared with the above laboratory parameters in an unselected sample of 204 patients admitted to a general hospital. The sensitivities, specificities, and positive (PPV) as well as negative predictive values of the CAGE, the MAST, and the alcohol markers were calculated along with the reported alcohol consumption or the ICD-10 diagnosis as standard. According to recent harmful alcohol consumption levels (women >225 g/week: men >350 g/week), the sensitivities and the PPVs were rather low in all tests (sensitivity <60%; PPV <50%). With the ICD-10 diagnosis as standard, the CAGE and MAST showed a rather high specificity (>95%) and PPV (about 90%). CDT revealed the best PPV of all alcohol markers (60%). However, the sensitivity of the CAGE, MAST, and the alcohol markers for the ICD-10 diagnosis was rather poor (<60%). This low sensitivity impedes the usefulness of these questionnaires and alcohol markers as screening tests for alcoholism in general hospitals.