OBJECTIVE: To examine the risk and determinants of dementia following a clinically overt stroke in a prospectively followed cohort of elderly subjects. METHODS: We examined the effect of a clinically detectable stroke on the risk of dementia using prospective data from 335 subjects in the Baltimore Longitudinal Study of Aging, all of whom were cognitively and neurologically normal at entry into the study (mean age at entry 75.1 +/- 4.2 years). RESULTS: Clinically overt strokes are common in our cohort (cumulative risk by age 90, 15.4%; 95% CI: 10 to 22%) and confer an increased risk of dementia compared to subjects without stroke (odds ratio [OR] 5.55; 95% CI: 2.76 to 11.4). The majority of patients who became demented after a stroke had evidence of mild cognitive impairment preceding the stroke (14 of 19). Moreover, a clinically symptomatic stroke was a major risk factor for the conversion of mild cognitive impairment to dementia (OR 12.4; 95% CI: 1.5 to 99). When cognitive impairment did not precede the stroke, there was no increase in the risk of subsequent dementia. Pathologic data indicate that both vascular and Alzheimer pathology leads to the prestroke impairment. CONCLUSION: Dementia after stroke may be determined by cognitive impairments that exist prior to the stroke.