BACKGROUND Old age is associated with reduced cognitive performance. Nutritional factors may contribute to this association. OBJECTIVE We tested associations between cognitive performance and plasma vitamin B-12, folate, and homocysteine concentrations in the elderly. METHODS We studied survivors of the Scottish Mental Surveys of 1932 (Aberdeen 1921 Birth Cohort, or ABC21) and 1947 (Aberdeen 1936 Birth Cohort, or ABC36), which surveyed childhood intelligence quotient. We measured folate, vitamin B-12, and homocysteine concentrations in fasting blood samples and cognitive performance by the Mini Mental State Examination (MMSE), National Adult Reading Test (NART), Raven's Progressive Matrices (RPM), Auditory Verbal Learning Test (AVLT), digit symbol (DS) subtest, and block design (BD) subtest. RESULTS Homocysteine was higher in the ABC21 than in the ABC36 (P < 0.0001). There were positive correlations between folate and vitamin B-12 and negative correlations between homocysteine and both folate and vitamin B-12. MMSE, RPM, AVLT, DS, and BD scores were higher in the ABC36. In the ABC21, folate, vitamin B-12, and MMSE score were positively correlated and homocysteine was negatively correlated with RPM, DS, and BD scores. Folic acid was positively correlated with AVLT and DS scores. In the ABC36, folate was positively correlated with BD score. After adjustment for childhood intelligence quotient, partial correlations were strengthened between vitamin B-12 and NART score and between homocysteine and RPM score but weakened between red blood cell folate and DS score. CONCLUSIONS B vitamins and homocysteine are associated with cognitive variation in old age. In the ABC21 but not the ABC36, homocysteine accounted for approximately 7-8% of the variance in cognitive performance. This may prove relevant to the design of neuroprotective studies in late life.