Objective: To assess the reproducibility (reliability and inter-rater agreement> of the client-centred Canadian Occupational Performance Measure (COPM). Design: The COPM was administered twice, with a mean interval of seven days (SD 1.6, range 4-14), by two different occupational therapists. Data analysis was based on intraclass correlation coefficients, the Bland and Altman method and Cohen's weighted kappas. Setting: Occupational therapy departments of two university medical centres. Subjects: Consecutive clients, with various diagnoses, newly referred to the outpatient clinic of two occupational therapy departments, were included. They were all over 18 years of age and perceived limitations in more than one activity of daily life. Complete data on 95 clients were obtained: 31 men and 64 women. Results: Sixty-six per cent of the activities prioritized at the first assessment were also prioritized at the second assessment. The intraclass correlation coefficients were 0.67 (95% confidence interval (CI) 0.54-0.78) for the mean performance score and 0.69 (95% CI 0.56-0.79) for the mean satisfaction score. The limits of agreement were - 2.5 to 2.4 for the mean performance score and - 2.3 to 2.7 for the mean satisfaction score. For the separate prioritized problems, the weighted kappas ranged from 0.37 to 0.49. Conclusions: Inter-rater agreement of the prioritized problems was moderate. The reproducibility of the mean performance and satisfaction scores was moderate, but it was poor for the scores of the separate problems. Therefore, the mean scores should be used for individual assessment.