There has been no validated screening instrument for use in non-psychotic illness of the elderly in Taiwan. This study aims to test the validity of the 12-item Chinese Health Questionnaire (CHQ-12) among the elderly in a community study. The CHQ-12 was administrated via reading-out to 222 subjects aged 65 and over from three communities. Psychiatrists using the Geriatric Mental Status Schedule (GMS) assessed psychiatric condition while the diagnosis was made according to the computerized program, AGECAT. Validity indices of the CHQ-12 were calculated, using the Relative Operating Characteristic (ROC) analysis for its optimal cut-off point. Variables hypothesized to affect its performance were assessed. Validity of the CHQ-12 at optimal cut-off point 2/3 were estimated with a sensitivity of 79.7%, a specificity of 83.6%, a positive predictive value of 68.9%, a negative predictive value of 90.0% and an overall misclassification rate of 17.6%, and an estimate of the area under the ROC curve of 0.81. The performance of the CHQ-12 was better in males, in those who were literate, and in those without any physical illness. This study demonstrated that the use of CHQ-12 in the elderly community is as valid as in the general population survey. However, it should be read out by the investigator rather than self-administered due to the high proportion of illiteracy among the Taiwanese elderly.