OBJECTIVES: To establish whether elderly people with impaired cognition are at greater risk for the de-velopment of type 2 diabetes. DESIGN: Prospective population-based cohort study. SETTING: The El-derly Nutrition and Health Survey in Taiwan (NAHSIT Elderly). PARTICIPANTS: One thousand and four hundred ninety-three diabetes-free people >=65 years were followed for incident diabetes in relation to cognitive status for up to 8 years. MEASUREMENTS: The association between cognitive impairment and diabetes incidence was analyzed with Cox proportional hazards models with exclusion of people who had diabetes within one year of cognitive function assessments. RESULTS: Cognitively-impaired women, but not men, had increased diabetes incidence density (DID). Age, gender, ethnicity and personal behavior adjusted hazard ratios (HR) and 95% confidence intervals (CI) for type 2 diabetes with normal cognition as referent were 2.43 (95% CI: 1.27-4.63) for women and 1.55 (95% CI: 0.48-5.07) for men. These gender differences and the HR significances remained with adjustments for age, ethnicity, financial status, dietary quality as a dietary diversity score, physical function, physical activity, fasting glucose, indices of body composition, body mass index, waist circumference, mid-arm muscle circumference, perceived and mental health status. There were extensive significant interactions with the covariates in women. CONCLUSION: Cognitive impairment in later life is associated with greater risk of type 2 diabetes in women and considerable potential risk enhancement.