OBJECTIVE To assess the sensitivity to change of two depression scales for stroke patients: the Aphasic Depression Rating Scale (ADRS), which is a 9-item external assessment, and the Visual Analog Mood Scale (VAMS), which is a visual self-assessment scale. METHODS Forty-nine stroke patients admitted to two rehabilitation units. METHODS Symptoms of depression were assessed twice at a one-month interval (D0-D30) using the ADRS, the VAMS, and by a trained psychologist (PSY). Sensitivity to change was assessed by effect size and standardized response mean. A one-way ANOVA on ranks was performed to determine if the scales distinguished between deteriorated, stable and improved patient status. Spearman's correlation coefficient (r) was used to assess the relationship between changes in PSY and changes in the ADRS and the VAMS between D0 and D30. RESULTS Mean depression scores at D0 and D30 were 2.6 +/- 1.8 and 2.4 +/- 1.7 (ADRS/10), 3.1 +/- 2.9 and 3.0 +/- 3.2 (VAMS/10), 2.9 +/- 2.3 and 2.4 +/- 2.4 (PSY/10). Percentages of deteriorated, stable and improved patient status were 22%, 45% and 33% (ADRS); 41%, 22% and 37% (VAMS); and 20%, 41% and 39% (PSY). Changes in PSY correlated with the ADRS (r = 0.72, P<10( -6)) and the VAMS (r = 0.41, P<10(- 2)). The ADRS was better than the VAMS in terms of effect size, standardized response mean and ability to discriminate between deteriorated, stable and improved patient status. This result was partly due to the difficulty encountered by nine (19%) patients in understanding directions to complete the VAMS. CONCLUSIONS The ADRS is more sensitive than the VAMS for detecting changes in post-stroke depression. The VAMS is less appropriate in very cognitively impaired stroke patients.