INTRODUCTION: We investigated the association between glucose tolerance status and trajectories of change in blood glucose, and cognitive function in adults aged 25 to 85. METHODS: The sample (n = 4547) was drawn from a national, population-based cohort study in Australia (AusDiab). Fasting plasma glucose (FPG), glycated haemoglobin (HbA1c) and general health were assessed at 0, 5 and 12 years. Covariates included age, education, body mass index, blood pressure and physical activity. At 12 years, participants completed assessments of memory, processing speed and verbal ability. RESULTS: Known diabetes at baseline was associated with slower processing speed at 12 years in both younger (25-59 years) and older (>60 years) age-groups. After 12 years of follow-up, adults aged < 60 with diabetes at baseline had a mean speed score of 49.17 (SE = 1.09) compared with 52.39 (SE = 0.20) in normals. Among younger males without diagnosed diabetes, reduced memory at 12 years was associated with higher HbA1c at 5 years (β = -0.91, SE = 0.26, p < 0.001). No effects were apparent for females or older males. Adjusting for insulin sensitivity (HOMA-%S) and hs-C reactive protein attenuated these associations, but depression and CVD risk did not. Latent class analysis was used to analyse the associations between trajectories of HbA1C and glucose over 12 years, and cognition. Identified classes were described as 1) normal and stable blood glucose over time (reference), 2) high intercept but stable blood glucose over time, and 3) increasing blood glucose over time. In both young males and females, high stable glucose measures were associated with poorer cognitive function after 12 years. CONCLUSION: Those with type 2 diabetes, younger males with high non-diabetic HbA1c, and adults with high stable blood glucose are at increased risk of poorer cognition. The findings reinforce the need for management of diabetes risk factors in midlife.