OBJECTIVE: Self-management of type 2 diabetes including avoidance of hypoglycemia is complex, but the impact of cognition on safe self-management is not well understood. This study aimed to assess the effect of baseline cognitive function and cognitive decline on subsequent risk of severe hypoglycemia and to assess the effect of different glycemic strategies on these relationships. RESEARCH DESIGN AND METHODS: Prospective cohort analysis of data from the ACCORD trial included 2,956 adults aged >/=55 years with type 2 diabetes and additional cardiovascular risk factors. Cognitive tests (Digit Symbol Substitution Test [DSST], Rey Auditory Verbal Learning Test, Stroop Test, and Mini Mental Status Examination) were conducted at baseline and 20 months. Study outcomes were incident confirmed severe hypoglycemia requiring medical assistance (HMA) and hypoglycemia requiring any assistance (HAA). RESULTS: After a median 3.25-year follow-up, a 5-point-poorer baseline score on the DSST was predictive of a first episode of HMA (hazard ratio 1.13 [95% CI 1.08-1.18]). Analyses of the other cognitive tests and of HAA were consistent with the DSST results. Cognitive decline over 20 months increased the risk of subsequent hypoglycemia to a greater extent in those with lower baseline cognitive function (P(interaction) = 0.037). Randomization to an intensive versus standard glycemic strategy had no impact on the relationship between cognitive function and the risk of severe hypoglycemia. CONCLUSIONS: Poor cognitive function increases the risk of severe hypoglycemia in patients with type 2 diabetes. Clinicians should consider cognitive function in assessing and guiding their patients regarding safe diabetes self-management regardless of their glycemic targets.