OBJECTIVE: To quantify the initial deficit, change, and outcome in gait velocity during inpatient rehabilitation following stroke. DESIGN: The initial deficit on admission to rehabilitation was quantified by comparing 42 stroke patients with 42 controls matched by gender and age. The change in the stroke patients during the next 8 weeks was quantified and gait outcome was compared with functional and normal criteria. SETTING: Patients were referred from four inpatient rehabilitation centers at the time of admission following a median of 16.5 days in the acute hospital. PATIENTS: SELECTION CRITERIA: ability to give informed consent; unilateral first stroke; ability to walk 10 meters. INTERVENTION: Patients participated in a median of 17.38 hours of individual physical therapy including a median of 6.92 hours of gait training during the 8 weeks. MAIN OUTCOME MEASURE: Gait velocity. RESULTS: Gait velocity was initially 38.6% (26.7m/min SD = 14.9) of the performance of controls and improved to 55.1% (38.1m/min). At outcome only 24% exceeded the 5th percentile of controls (48.1m/min) or the velocity required to cross the typical signalled intersection (46.2m/min). The change was only 26% of the initial deficit. Fifty-five percent of the patients improved beyond the 95% confidence intervals surrounding the error of measuring change. Indices of responsiveness indicated that there was a high signal-to-noise ratio and a robust effect size. CONCLUSION: Gait velocity discriminated the effect of stroke and the change during rehabilitation.