Previous studies have consistently found a selective delay of the P2 flash visual evoked potential (VEP) component among groups of patients with Alzheimer's dementia (AD) compared with control groups. Several authors have termed the selective P2 delay a "marker" for AD and have called for its use in clinical diagnosis. This study examined the diagnostic utility of the selective P2 delay in a retrospective sample of 45 AD patients and 60 age-equivalent healthy control subjects. Although significant between-group differences were found, classification accuracies for individual patients and controls were too low for the P2 delay to contribute meaningfully to clinical diagnosis.