Objective. – To characterize on-the-road, behind-the-wheel driving abilities and related laboratory performances of subjects with mild Alzheimer's disease (AD) and vascular dementia. Design. – Prospective, experimental study involving two mild dementia and three age and health control groups. Road test reliability and validity were assessed. Setting. – Greater western Los Angeles. Subjects were enrolled from the community by referral and from the Veterans Affairs dementia and diabetes clinics. Participants. – Eighty-seven driving subjects were enrolled; 83 completed the study. A sample of eligible dementia clinic subjects consisting of 15 mild AD patients met National Institute of Neurological and Communicative Disorders and Stroke – Alzheimer's disease and Related Disorders Association probable AD criteria, while 12 met Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition and Hachinski diagnostic criteria for multi-infarct dementia (vascular dementia). Clinic control subjects consisted of 15 age-matched patients with diabetes and without a history of stroke or dementia. Community controls consisted of 26 healthy, age-matched, older subjects (>60 years) and 16 young subjects (20 to 35 years). Main Outcome Measures. – Drive score from the Sepulveda (Calif) road test and laboratory measures of attention, perception, and memory. Results. – The drive scores in the mild AD group (mean, 22.1; SD, 3.8) and in the vascular dementia group (mean, 24.0; SD, 7.8) differed significantly (P<.001 studentized range test) from the drive scores in the diabetic control group (mean, 31.5; SD, 3.9), the older control group (mean, 32.6; SD, 2.8), and the young control group (mean, 33.6; SD, 3.2). Drive score among the three control groups did not vary significantly. Short-term memory (Sternberg), visual tracking, and Folstein Mini-Mental State Examination scores correlated best with drive score, with a cumulative R(2) of 0.68. Drive score and number of collisions and moving violations per 1000 miles driven were negatively correlated (r=-0.38; P<.02). Conclusions. – Based on this study, type and degree of cognitive impairment are better predictors of driving skills than age or medical diagnosis per se. Specific testing protocols for drivers with potential cognitive impairment may detect unsafe drivers more effectively than using age or medical diagnosis alone as criteria for license restriction or revocation.; Various psychological tests may be able to detect elderly people with mild dementia who are poor drivers. Researchers gave a driving test and several tests of mental status to 13 people with mild Alzheimer's disease, 12 with mild vascular dementia, 26 healthy people over 60 and 16 young healthy volunteers. Elderly people with mild dementia did worse on the driving test than young and old people without dementia. Low scores on visual tracking, mental status and short-term memory tests were most closely associated with low scores on the driving test in the elderly people with mild dementia.