Introduction. Research and clinical settings use the 10-meter walk test (10MWT) to measure locomotor capacity with considerable methodological diversity. Comparison between healthy and disabled children is frequent; however, the reproducibility of 10MWT using different methods is unknown. Objectives. This study analysed intrasubject, test-retest reliability, and agreement of four methods of 10MWT, exploring the influence of pace, acceleration-deceleration phases, and anthropometric measurements when calculating mean velocity. Methods. This cross-sectional study evaluated 120 typical children, both sexes, aged 6, 8, 10, and 12 (n=30 for each age). The mean times and velocities of the path (10 m) and middle path (6 m) obtained at a self-selected and fast pace were analysed. Initial assessment and another after seven days recorded three measurements per method (sV6 = self-selected pace and 6 m; sV10 = self-selected pace and 10 m; fV6 = fast pace and 6 m; fV10 = fast pace and 10 m). Interclass correlation coefficient (ICC), multiple regression, and Snedecor-F test (5% significance level) were used. Results. The fV10 method had high intrasubject reliability for all tested ages (0.70<ICC>0.89); sV10 exhibited high intrasubject reliability for ages 6, 8, and 12 (0.70<ICC>0.89) and moderate for age 10 (0.50<ICC<0.69).Test-retest reliability at sV6 and fV6 did not reach high ICC in any tested ages. The test-retest reliability at sV10 and fV10 was moderate for ages 6, 8, and 12 (0.50<ICC>0.69) and poor for age 10 (0.25<ICC>0.49). There was no agreement between methods: sV6 versus sV10 (mean difference=0.91 m/s; SEM=0.036); fV6 versus fV10 (mean difference=1.70; SEM=0.046). The fV6 method versus fV10 overestimated the velocity (bias=1.70 m/s). Conclusions. For typical children, the method that ensured the highest intrasubject reliability used fast pace and 10 m. Moreover, test-retest reliability increased when adopting 10 m at both self-selected and fast pace. The methods were not equivalent but were related, and those that did not compute the entire pathway overestimated the results.