The incidence of dementia based on DSM-III criteria and the impact of age and education on the development of dementia using the data of a longitudinal community population survey is discussed. Study subjects were randomly sampled from the community and were assessed by a two-stage procedure twice during a 5-year period. The screening instrument was the Mini-Mental State Examination and the second stage measures include a medical history and physical examination, a neurological examination and an intensive neuropsychiatric interview and testings (i.e. Blessed Dementia Scale, Activity of Daily Living, Fuld Object Memory Test and Block Design Test. All identified cases from the second stage interview were further verified by computed tomographic scan and laboratory tests. A total of 1970 subjects of non-dementia aged 65 or older in 1987 were re-surveyed after 5 years, of whom 114 new cases of dementia were identified. The incidence of dementia in total was 1.15% annually, 0.98% for males and 1.27% for females. The incidence for Alzheimer's disease, vascular dementia and dementia by other causes was 0.74, 0.33 and 0.08%, respectively. The highest rate (2.02%) was found in illiterate individuals. The incidence of dementia in the Shanghai cohort is similar to that from other countries. Age and education are closely related to the occurrence of dementia. While age increases by each 5 years for those aged 65 and over the likelihood of developing dementia increases nearly one-fold (74%). Education shows great protective effect on the development of dementia. The non-educated has a higher risk of suffering from dementia than does the educated.