OBJECTIVE: Patients with type 2 diabetes have an increased risk of a reduction in cognitive function. We investigated the hypothesis that plasma glucose is associated with a reduction in episodic and/or semantic memory already in nondiabetic subjects. RESEARCH DESIGN AND METHODS: We linked two large population-based datasets in Sweden: the Betula study, in which a random sample from the population aged 35-85 years was investigated for cognitive function, including episodic and semantic memory; and the Västerbotten Intervention Program, a health survey with subjects aged 40, 50, and 60 years, that includes measuring of fasting and 2-h plasma glucose, along with other risk factors for diabetes and cardiovascular disease. We identified 411 (179 men and 232 women, mean age 50.6 +/- 8.0 years) nondiabetic subjects, free from dementia, who had participated in the two surveys within 6 months. RESULTS: Women had better episodic (score 7.37 +/- 1.42) and semantic memory (score 16.05 +/- 2.76) than men (score 6.59 +/- 1.29 and 15.15 +/- 2.92, respectively, P < 0.001 for both). In an adjusted multivariate model, fasting plasma glucose (fPG) and 2-h plasma glucose (2hPG) were significantly negatively associated with episodic memory (fPG: B -0.198, SE 0.068, beta -0.209, P = 0.004; and 2hPG: B -0.061, SE 0.031, beta -0.148, P = 0.048, respectively) in women but not in men. The association was not found in relation to semantic memory. CONCLUSIONS: We conclude that an increase in plasma glucose is associated with impairment in episodic memory in women. This could be explained by a negative effect on the hippocampus caused by raised plasma glucose levels.