An easy to use screening instrument for drinking behavior or alcohol-related illness has not been established in Taiwan. The authors examined the cross-cultural validity of the CAGE questionnaire in Chinshan, Taipei County in Taiwan. Study subjects (N=198) were outpatients randomly drawn from the primary care setting in the township. Three psycholinguistic items, equivalent to ”Cut-down” (C1, C2) and ”Annoyance” (A1) of CAGE, thought to be culturally relevant for people in Taiwan, were added to form an experimental Chinese CAGE (C-CAGE). This allowed for tests of five types of scoring combination, CAGE, C1A1GE, C1AGE, C2A1GE and C2AGE. Each subject received a semi-structured clinical interview for diagnosis of alcoholism after he or she completed a C-CAGE questionnaire. The performance of CAGE in our sample was at the lowest limit among those reported in Western countries. Compared to CAGE, C2AGE was superior in robustness for specificity and overall accuracy. C2A1GE tended to be better in sensitivity and performance for the value of the area measured by receiver operating characteristics curve analysis. This indicates that CAGE has room for further improvement under the assumption of psycholinguistic equivalence.