We used the medical records linkage system for the population of Rochester, Minnesota, to identify persons in the community who had their first cerebral infarct without previous dementia. In this cohort (n = 971), the incidence of dementia in the first year was nine times greater than expected, but if we did not observe dementia in the first year, the risk of dementia in the cohort each year thereafter was about twice the risk in the population. After the first year, a 50% increase was observed in Alzheimer's disease in the cohort compared with that in the community. Although the incidence of dementia increased with increasing age, the standardized morbidity ratios decreased with increasing age. Age, sex (male), and second stroke were significant independent predictors of dementia in a multivariate Cox proportional hazards model. There was no effect of location or clinical severity of infarct on the rate of occurrence of dementia.