A culturally adapted Chinese version of the Blessed-Roth Information-Memory-Concentration test (CIMC) was used in a dementia screening survey of a probability sample of 5,055 elderly Shanghai residents. The individual items on the CIMC that best predicted the overall score were similar to the best predictor of an American version of the IMC. Performance on the CIMC was markedly affected by the level of education or lack thereof. In a subsample for whom clinical diagnoses were obtained, it was possible to establish cutoff values on the CIMC by stratifying the sample according to education.