It is unclear whether persons with diabetes are at increased risk for dementia, including Alzheimer's disease. Existing studies are limited by small sample size, selection bias, and case-control designs. This population-based historical cohort study provides estimates of the risk of dementia and Alzheimer's disease associated with adult onset diabetes mellitus (AODM). The sample included all persons with AODM residing in Rochester, Minnesota, on January 1, 1970, plus all persons diagnosed in Rochester or who moved to Rochester with the diagnosis between January 1, 1970, and December 31, 1984. Individuals were followed through review of their complete medical records from AODM diagnosis until dementia onset, emigration, death, or January 1, 1985. Standardized morbidity ratios for dementia and Alzheimer's disease were calculated, using an expected incidence based on age- and sex-specific rates for the Rochester population. Poisson regression was used to estimate risks for persons with AODM relative to those without. Of the 1, 455 cases of AODM followed for 9,981 person-years, 101 developed dementia, including 77 who met criteria for Alzheimer's disease. Persons with AODM exhibited significantly increased risk of all dementia (Poisson regression relative risk (RR) = 1.66, 95% confidence interval (Cl) 1.34–2.05). Risk of Alzheimer's disease was also elevated (for men, RR = 2.27, 95% Cl 1.55–3.31; for women, RR = 1.37, 95% Cl 0.94–2.01). These findings emphasize the importance of AODM prevention and prompt additional investigation of the relation between AODM and dementia. Am J Epidemiol 1997; 145: 301–8.