Objectives. To examine the reliability and validity of the Clock Drawing Test when used as a cognitive screening instrument for mild to moderate dementia, and to compare different scoring mechanisms. Design. Retrospective analysis of clock drawing performance using three published scoring methods (Shulman, Sunderland and Wolf‐Klein). Setting. Hospital‐based memory disorders clinic. Participants. A sample of 28 consecutive patients attending the memory clinic for assessment who were given a diagnosis of Alzheimer's disease (mild or moderate) and 28 age‐ and sex‐matched control subjects comprising 17 memory clinic attenders found to be normal and 11 community volunteers. Measurements. Sensitivity and specificity of the three clock rating scales against memory clinic diagnoses of dementia using DSM‐III‐R; their respective interrater reliabilities; and comparisons of each with measures of cognitive impairment (the Mini‐Mental State Examination and the Blessed Orientation–Information–Memory–Concentration Test), daily performance of basic and instrumental activities (the Blessed Dementia Scale) and depression (the Hamilton Rating Scale for Depression). Results. All methods of scoring the Clock Drawing Test correlated well with measures of cognitive impairment (r=0.57–0.73) and daily performance (r=0.38–0.48), were independent of mild depression and demonstrated high sensitivity, specificity and interrater reliability. While all clock scales identified mild to moderate dementia reasonably well, the Shulman method performed best. In screening for dementia, clock drawing proved superior to the MMSE: 24/28 vs 20/28 cases identified. When compared with the MMSE, clock drawing provided additional diagnostic discrimination, identifying 7/8 AD patients with MMSE scores ≥24. Conclusions. In a clinic population, clock drawing, especially if scored according to the Shulman scale and combined with the MMSE, is an extremely efficient test screening measure for mild to moderate dementia of the Alzheimer's type with low false negative and false positive rates. This may have implications for screening elderly populations.