Objective: To evaluate construct validity and reproducibility of the Functional Gait Assessment (FGA) for measuring walking balance capacity in persons after stroke. Design: Cross-sectional study. Setting: Inpatient and outpatient rehabilitation center. Subjects: Fifty-two persons post-stroke (median (25% and 75% percentiles)) time post-stroke 6 (5–10) weeks) with independent walking ability (mean gait speed 1.1 ±.4 m/s). Methods: Subjects completed a standardized FGA twice within one to eight days by the same investigator. Validity was evaluated by testing hypotheses on the association with two timed walking tests, Berg Balance Scale, and the mobility domain of the Stroke Impact Scale using correlation coefficients (r), and with Functional Ambulation Categories using the Kruskal–Wallis test. Reproducibility of FGA scores was assessed with intraclass correlation coefficient and standard error of measurement. Results: Subjects scored a median of 22 out of 30 points at the first FGA. Moderate to high significant correlations (r.61–.83) and significant differences in FGA median scores between the Functional Ambulation Categories were found. Eight hypotheses (80%) could be confirmed. Inter-rater, intra-rater, and test–retest reliability of the total scores were excellent. The standard error of measurement and minimal detectable change were 2 and 6 points, respectively. No relevant ceiling effect was observed. Conclusion: The FGA demonstrated good measurement properties in persons after stroke and yielded no ceiling effect in contrast to other capacity measures. In clinical practice, a measurement error of 6 points should be taken into account in interpreting changes in walking balance.