The Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) provides ratings of an individual's changes in everyday cognitive functions during the previous 10 years. Original studies conducted in Australia showed that its score was not influenced by the subjects' educational backgrounds and that it performed at least as well as the Mini-Mental State Examination (MMSE) as a screening instrument for dementia. The subjects of the present study were Chinese and included 399 community residents and 61 dementia patients. Their ages ranged from 50 to 92 years; their education levels ranged from 0 to 19 years, and 63% of them had never attended school. We administered the IQCODE to informants and the Cognitive Abilities Screening Instrument (CASI), from which a CASI-estimated score of the MMSE (MMSE-CE) can be obtained, to the subjects. The diagnosis of dementia was made independently by physicians according to the DSM-III-R criteria based on semistructured interview and testing, neurologic examination, and standardized assessments of cerebral vascular disease, Parkinson's disease, and depression. The Chinese IQCODE showed no association with the subjects' education level or gender, low association with their age, and moderately high association with their MMSE-CE score. The area under the receiver operating characteristic curve of the IQCODE was significantly larger than that of the MMSE-CE for the whole group and for the subgroup with 1 to 19 years of education but not for the subgroup with 0 years of education. Nine of the 26 items of the IQCODE could be deleted without appreciable reduction in sensitivity and specificity. The IQCODE (1) can be shortened to 17 items, (2) had good cross-cultural applicability, and (3) was better than the MMSE-CE as a screening tool for dementia in a population with large variation in educational backgrounds.