We evaluated the sensitivity and specificity of positron emission tomography for diagnosis of probable Alzheimer's disease under conditions similar to those encountered in the routine clinical practice of nuclear medicine. We obtained tomographic images of regional cerebral blood flow from three groups of subjects: (1) 13 subjects, ages 69 to 84, who had probable Alzheimer's disease diagnosed by validated clinical criteria; (2) 15 subjects, ages 57 to 77, who had Parkinson's disease without dementia; and (3) 11 subjects, ages 65 to 83, who were normal. Three blinded reviewers, who had not previously seen the images, categorized them as normal, bilateral temporoparietal flow defects typical of Alzheimer's disease, or other abnormality. Consensus interpretation demonstrated sensitivity of 0.38 (5/13) and specificity of 0.88 (23/26) for identifying patients with probable Alzheimer's disease. Thus, the criterion of bilateral temporoparietal reduction in cerebral blood flow used in this study did not have sufficient sensitivity to be of clinical value. While other criteria may be developed to improve diagnostic accuracy, clinical utility can be established only by testing for validity in patients with a full spectrum of complicating neurologic and psychiatric conditions for whom diagnosis is uncertain and who are then followed longitudinally to determine clinical outcome or pathologic findings.