OBJECTIVE: To examine the association of anxiety with incident dementia and cognitive impairment not dementia (CIND). METHODS: We conducted a prospective study of men aged 48 to 67 years at baseline anxiety assessment; we measured cognition 17 years later. We studied 1481 men who were either eligible for examination or were known to have dementia. Trait Anxiety was assessed using the Spielberger State Trait Anxiety Inventory. Psychological distress was assessed using the 30-item general health questionnaire. Cognitive screening was followed by a clinical examination. Medical notes and death certificates of those not seen were also examined. Outcomes were CIND and Diagnostic and Statistical Manual of Mental Disorders, 4th Edition (DSM-IV) dementia. RESULTS: Of 1160 men who were cognitively screened, 174 cases of CIND and 69 cases of dementia were identified. A further 21 cases of dementia were identified from medical records. After adjustment for age, vascular risk factors and premorbid cognitive function associations with higher anxiety (31st-95th centile) were for CIND odds ratio (OR) 2.31 (95% Confidence Interval (CI) = 1.20-4.44) and for dementia OR 2.37 (95% CI = 0.98-5.71). These associations were slightly stronger for nonvascular (OR = 2.45; 95% CI = 1.28-4.68) than for vascular impairment (OR = 1.94; 95% CI = 0.77-4.89). Analyses of change in cognitive performance, assessed by the Cambridge Cognitive Examination of the Elderly subscales found some evidence for decline in learning memory with higher anxiety score (b(age adj) = -0.291 (-0.551, -0.032), but not for any other subscale. CONCLUSIONS: Anxiety is a risk factor for CIND and dementia. The extent to which the association is independent of depression and whether or not it is causal requires further study.