Persons with type 2 diabetes are at increased risk of cognitive dysfunction. Less is known about which cognitive abilities are affected and how undiagnosed diabetes and impaired fasting glucose relate to cognitive performance. The authors explored this question using data from 1,917 nondemented men and women (average age = 76 years) in the population-based Age, Gene/Environment Susceptibility-Reykjavik Study (2002-2006). Glycemic status groups included diagnosed diabetes (self-reported diabetes or diabetic medication use; n = 163 (8.5%)), undiagnosed diabetes (fasting blood glucose >or=7.0 mmol/L without diagnosed diabetes; n = 55 (2.9%)), and impaired fasting glucose (fasting blood glucose 5.6-6.9 mmol/L; n = 744 (38.8%)). Composites of memory, processing speed (PS), and executive function were constructed from a neuropsychological battery. Linear regression was used to investigate cross-sectional differences in cognitive performance between glycemic groups, adjusted for demographic and health factors. Persons with diagnosed diabetes had slower PS than normoglycemics (beta = -0.12; P < 0.05); diabetes duration of >or=15 years was associated with significantly poorer PS and executive function. Undiagnosed diabetics had slower PS (beta = -0.22; P < 0.01) and poorer memory performance (beta = -0.22; P < 0.05). Persons with type 2 diabetes have poorer cognitive performance than normoglycemics, particularly in PS. Those with undiagnosed diabetes have the lowest cognitive performance.