A randomized controlled trial examined whether the diagnostic process for Alzheimer's disease and other dementias may be influenced by knowledge of the patient's education and/or self-reported race. Four conditions were implemented: diagnostic team knows (a) race and education, (b) education only, (c) race only, or (d) neither. Diagnosis and clinical staging was established at baseline, at Wave 2, and for a random sample of Wave 3 respondents by a consensus panel. At study end, a longitudinal, "gold standard" diagnosis was made for patients with follow-up data (71%). Group differences in "Diagnostic and Statistical Manual of Mental Disorders" (4th ed.; American Psychiatric Association, 1994) diagnosis were estimated using logistic regression and generalized estimating equations. Sensitivity and specificity were examined for baseline diagnosis in relation to the gold standard, longitudinal diagnosis. Despite equivalent status on all measured variables across waves, members of the "knows race only" group were less likely than those of other groups to receive a diagnosis of dementia. At final diagnosis, 19% of the "knows race only" group was diagnosed with dementia versus 38% to 40% in the other 3 conditions (p = 0.038). Examination of sensitivities and specificities of baseline diagnosis in relation to the gold standard diagnosis showed a nonsignificant trend for lower sensitivities in the knowing race conditions (0.3846), as contrasted with knowing education only (0.480) or neither (0.600). The finding that knowledge of race may influence the diagnostic process in some unknown way is timely, given the recent State-of-the-Science conference on Alzheimer's disease prevention, the authors of which called for information about and standardization of the diagnostic process. (Contains 2 figures and 5 tables.)