Objective: To assess the accuracy of clock drawing for detecting dementia in a multicultural, non-English-speaking-background population. Design: A prospective cohort study. Setting: A general geriatric medical outpatient clinic in southwest Sydney, Australia. Participants: Ninety-three consecutive new patients to the clinic who had a non-English-speaking-background country of birth (mean age 78.0 years). Measurements: The clock drawing test was conducted at the beginning of each clinic visit by a blinded investigator. Each patient was then assessed by a geriatrician who collected demographic data, administered the Modified Barthel Index, the Geriatric Depression Scale, and the Folstein Mini-Mental State Examination, and categorized each patient as normal or demented, according to DSM-IV criteria. Interpreters were used for participants who spoke a language other than English or who requested them. Each clock drawing was scored according to the 4-point CERAD scale and the previously published methods of Mendez, Shulman, Sunderland, Watson, and Wolf-Klein. Scoring was evaluated for reliability and predictive accuracy, using receiver operating characteristic (ROC) curve analysis. Logistic regression analysis was used to assess the potential interaction between level of education and each of the clock scoring methods. Results: Using ROC curve analysis, there was no significant difference between the clock scoring methods (area under the curve varied from 0.60 to 0.72). The most sensitive was the Mendez scoring method (98%), with a specificity of 16%. Specificity above 50% was found only for the Wolf-Klein method, with an intermediate sensitivity of 78%. Conclusions: There were no significant differences in the clock scoring methods used to detect dementia. Performance of the clock drawing test was modest at best with low levels of specificity across all methods. Scored according to these methods, clock drawing was not a useful predictor of dementia in our multicultural population.