The goals of this study were to determine whether improvements in metabolic control can ameliorate the cognitive dysfunction associated with type 2 diabetes and evaluate the possibility that such improvements are mediated by changes in circulating insulin or insulin resistance. This randomized double-blind trial enrolled 145 subjects at 18 centers in the U.S. Older adults with type 2 diabetes receiving metformin monotherapy received add-on therapy with either rosiglitazone, a thiazolidinedione insulin sensitizer, or glyburide. Cognitive function was assessed at baseline and week 24 using the Digit Symbol Substitution Test, the Rey Auditory Verbal Learning Test, and the Cambridge Neuropsychological Test Automated Battery. Pretreatment fasting plasma glucose (FPG) in both groups was similar, and after 24 weeks both treatment groups showed similar significant reductions in FPG (2.1-2.3 mmol/l). Working memory improved with both rosiglitazone (P < 0.001) and glyburide (P = 0.017). Improvement (25-31% reduction in errors) was most evident on the Paired Associates Learning Test and was significantly correlated (r = 0.30) with improved glycemic control as measured by FPG. Similar and statistically significant cognitive improvement was observed with both rosiglitazone and glyburide therapy, and the magnitude of this effect was correlated with the degree to which FPG improved. These results suggest that a cognitive benefit is achievable with pharmacological interventions targeting glycemic control.