BACKGROUND: Smoking is a risk factor for several life-threatening diseases, but its long-term association with dementia is controversial and somewhat understudied. Our objective was to investigate the long-term association of amount of smoking in middle age on the risk of dementia, Alzheimer disease (AD), and vascular dementia (VaD) several decades later in a large, diverse population. METHODS: We analyzed prospective data from a multiethnic population-based cohort of 21,123 members of a health care system who participated in a survey between 1978 and 1985. Diagnoses of dementia, AD, and VaD made in internal medicine, neurology, and neuropsychology were collected from January 1, 1994, to July 31, 2008. Multivariate Cox proportional hazards models were used to investigate the association between midlife smoking and risk of dementia, AD, and VaD. RESULTS: A total of 5367 people (25.4%) were diagnosed as having dementia (including 1136 cases of AD and 416 cases of VaD) during a mean follow-up period of 23 years. Results were adjusted for age, sex, education, race, marital status, hypertension, hyperlipidemia, body mass index, diabetes, heart disease, stroke, and alcohol use. Compared with nonsmokers, those smoking more than 2 packs a day had an elevated risk of dementia (adjusted hazard ratio [HR], 2.14; 95% CI, 1.65-2.78), AD (adjusted HR, 2.57; 95% CI, 1.63-4.03), and VaD (adjusted HR, 2.72; 95% CI, 1.20-6.18). CONCLUSIONS: In this large cohort, heavy smoking in midlife was associated with a greater than 100% increase in risk of dementia, AD, and VaD more than 2 decades later. These results suggest that the brain is not immune to long-term consequences of heavy smoking.