The authors conducted a study of clock drawing test scoring by dementia specialists to determine interrater reliability and diagnostic accuracy. The authors randomly assigned 25 clocks from each of six predetermined groups based on consensus diagnosis (cognitive comparison subjects, subjects with a memory complaint but with normal neuropsychological testing, subjects with probable and possible mild cognitive impairment, and subjects with possible and probable Alzheimer's disease) to dementia specialists for blinded scoring using a binary yes/no impairment system and a 0-10 scale as subjectively determined by each individual clinician rater. The authors collapsed the six groups into three (comparison subjects, mild cognitive impairment patients, and Alzheimer's disease patients) and analyzed interrater reliability, sensitivity, and specificity for consensus diagnosis of mild cognitive impairment and Alzheimer's disease. The authors found excellent interrater reliability, sensitivity, and specificity for predicting consensus diagnosis. The 0-10 clock drawing test rating scale was more predictive of consensus diagnosis than the binary impairment scale. Based on rating systems, clock drawing test scoring by dementia clinicians had excellent interrater reliability and sensitivity for differentiating the mild Alzheimer's disease subjects from comparison subjects.