The purpose of this article was to examine the kinds of alcohol use disorder the AUDIT most accurately screens for since the literature is inconsistent in the use of the AUDIT. Sometimes it is viewed as a measure of hazardous or harmful drinking and sometimes as a measure of dependence. The performance of its subsets (consumption items, AUDIT-C; and problem items, AUDIT-P) and of the full AUDIT (AUDIT-10) was tested against four criteria: high-volume drinking, alcohol-related social problems, alcohol-related health problems, and alcohol dependence. A general population sample of 600 Swedish subjects was interviewed during the winter 2000-01. The results document that, at the recommended cutoff score of 8+, the AUDIT-10 performed well against all four criteria, even if less well against the alcohol-related health problems. The AUDIT-C also performed well against all the problem criteria, showing high areas under the ROC curve, even though significantly lower than the full scale. When measuring high-volume drinking, the AUDIT-C outperformed the full instrument. Scoring at least 1 on the AUDIT-P improved sensitivity of the instrument when screening for social problems and dependence and made it a satisfactory measure of health problems. It is suggested that, when using the full AUDIT to screen for problems more severe than high-volume drinking, the criterion of scoring at least 1 on the AUDIT-P should be applied in combination with a cutoff score on the AUDIT-C.