Objectives Reliability of clinical tests to evaluate ambulation in chronic hemiparesis may vary according to the testing condition. The 10-meter ambulation test (AT10) assesses walking speed and step length over 10 m, starting and ending in seated position. In the present study, we compared the intra- and inter-reliability of AT10 in chronic hemiparesis in four different conditions: with shoes and barefoot, at free and maximal safe speed.Methods Ten patients with hemiparesis, >1 year post-stroke (age 45 ± 12, time since stroke 16 ± 9 months, mean ± SD) participated in the reliability study (registration, ID-RCB-2017-A00090-53). All patients performed the AT10 twice, one week apart, in each of the four conditions. The number of steps and time to complete the task were manually recorded by four independent raters. The main outcome measurements were the intraclass correlation coefficients (ICC), coefficients of variation (CV), and mean raw differences (DIFF) of the three parameters of AT10 (speed, step length, and cadence) in each of the four conditions. Effects of wearing shoes and speed condition were explored using ANOVA.Results Across all conditions, mean intra- and inter-rater ICCs were, respectively, 98.5 ± 0.1 and 99.9 ± 0.1% for speed, 98.3 ± 0.1 and 99.7 ± 0.2% for step length, and 96.5 ± 0.1 and 98.9 ± 0.6% for cadence. Mean intra- and inter-rater CV for speed were 0.051 ± 0.016 and 0.022 ± 0.002, respectively. Intra-rater reliability of speed assessments was higher at maximal than at free speed (ICC, CV, DIFF, p < 0.05). At free speed, intra-rater ICCs were higher barefoot than with shoes (p < 0.05).Discussion Performing the 10-meter ambulation test barefoot at maximal speed optimizes its reliability.