To examine the hypothesis that caloric intake in mid-life is associated with later dementia or cognitive impairment not dementia (CIND). A prospective cohort study was conducted in Caerphilly, South Wales, United Kingdom. Men aged 45-59 years were identified from the electoral roll and general practice. 2,512 men were examined between July 1979 until September 1983. Four follow-up examinations were conducted every 4-5 years until 2004. Participants were categorized on the basis of their average daily caloric intake over each of the first three phases. Outcomes were CIND and dementia ascertained at phase five (2004). 192 men (15% of 1,248 participants at phase five) had CIND and 100 (8%) dementia. Age adjusted odds ratios demonstrated strongest associations between average energy consumption and vascular CIND or dementia (OR 1.62 95% CI 1.25-2.10). Adjustment for nutritional factors, vascular disease, diabetes, smoking, BP and BMI if anything increased the association (OR 1.64, 95% CI 1.03-2.60). After adjusting for social class, associations were attenuated and consistent with chance (OR 1.48, 95% CI 0.92-2.38). When adjusted for social class, the previously observed association between caloric intake and cognitive outcomes is modest, consistent with chance, and may be due to residual confounding.