Insulin resistance increases the risk of cognitive impairment and dementia, but higher insulin levels may be cognitively protective after a diagnosis of Alzheimer's disease (AD). The role of peripheral insulin as a predictor of cognitive decline both before and after an AD diagnosis needs further elucidation. We studied 197 AD cases and 198 normal controls enrolled in the Texas Alzheimer's Research and Care Consortium. Standardized protocols were used to collect age, gender, education, body mass index (BMI), serum insulin (not restricted to fasting), hemoglobin A1c (HbA1c), lipids, smoking and cardiovascular disease history, and neuropsychological tests including Mini-Mental State Examination, American National Adult Reading Test (AMNART) errors, Controlled Word Association Test (COWAT), Boston Naming Test, Wechsler Memory Scale-Revised (WMSR) Digit Span, Trails A and B, WMSR Logical Memory (LM) I and II, and Visual Reproduction (VR) I and II. We used linear regression to test the contribution of log-transformed serum insulin to each score, adjusting for age, gender, education, and BMI. In the AD cases, higher serum insulin was associated with worse performance on the COWAT (p < 0.001) and Trails B (p = 0.04). In controls, higher serum insulin was associated with worse performance on the AMNART (p = 0.001), COWAT (p = 0.007), Digit Span (p = 0.004), LM I (p = 0.004), LM II (p = 0.009), and marginally with VR II (p = 0.076). Adjustment for HbA1c, APOE4, and cardiovascular disease, or restricting the sample to mild AD, did not alter these associations. In non-demented older individuals, higher peripheral insulin appears to be associated with worse cognitive performance in multiple domains, but is not a consistent predictor in AD cases. These findings indicate the need for additional research on the role of insulin in the transition between normal and impaired cognitive function.