The Rivermead Mobility Index is used to measure mobility in patients with head injury or stroke. The purpose of the study was to examine construct validity, predictive validity, and the responsiveness of the Rivermead Mobility Index in stroke patients. Thirty-eight stroke inpatients participated in the study. The Rivermead Mobility Index, the Barthel Index, and the Berg Balance Scale were administered at admission to the rehabilitation ward and at discharge. The results showed that the Rivermead Mobility Index fulfilled the Guttman scaling criteria (coefficients of reproducibility > 0.9, coefficients of scalability > 0.7). The Rivermead Mobility Index scores were highly correlated with the Barthel Index scores (Spearman rs > 0.6) and the Berg Balance Scale scores (Spearman rs > = 0.8, all ps < 0.001). The Rivermead Mobility Index score at admission was closely correlated with the Barthel Index score at discharge (Spearman r = 0.77, p < 0.001). About 76% (29) of the subjects improved by more than 3 Rivermead Mobility Index points (median = 5) during their stay. The relationship between the change in score of the Rivermead Mobility Index and the Barthel Index was fair (Spearman r = 0.6, p < 0.001). These results indicate that the Rivermead Mobility Index is valid and sensitive to change over time. It is therefore a useful scale for the assessment of mobility in stroke patients.