A test-retest reproducibility study was performed to define a criterion for stability as opposed to change of motor function of the lower extremity in stroke patients. Forty-nine patients with stroke were examined twice by the same physiotherapist, using the Fugl-Meyer Assessment Scale. The interval between both measurements was three weeks. The mean differences between the first and the second measurement were small, with 0.04 points for the lower extremity scale and 0.92 points for the balance scale, respectively. Intraclass correlation coefficient for the lower extremity scale was 0.86, and 0.34 for the balance scale. The standard error of measurement for each scale was 1.76 and 1.17 points, respectively. The standard error of measurement can be transformed in an 'error threshold', which is a criterion to differentiate real changes from changes due to chance variation or measurement error. As the absence of real change is a parameter for stability, a change of less than 5 points for the lower extremity scale and of less than 4 points for balance confirms stability of motor function.