BackgroundThe common paradigm to study the adaptability of human gait is split-belt walking. Short-term savings (minutes to days) of split-belt adaptation have been widely studied to gain knowledge in locomotor learning but reports on long-term savings are limited. Here, we studied whether after a prolonged inter-exposure interval (three weeks), the newly acquired locomotor pattern is subject to forgetting or that the pattern is saved in long-term locomotor memory. Research QuestionCan savings of adaptation to split-belt walking remain after a prolonged inter-exposure interval of three weeks? MethodsFourteen healthy adults participated in a single ten-minute adaptation session to split-belt walking and five-minute washout to tied-belt walking. They received no training after the first exposure and returned to the laboratory exactly three weeks later for the second exposure. To identify the adaptation trends and quantify saving parameters we used Singular Spectrum Analysis, a non-parametric, data-driven approach. We identified trends in step length asymmetry and double support asymmetry, and calculated the adaptation volume (reduction in asymmetry over the course of adaptation), and the plateau time (time required for the trend to level off). ResultsAt the second exposure after three weeks, we found substantial savings in adaptation for step length asymmetry volume (61.6–67.6% decrease) and plateau time (76.3 % decrease). No differences were found during washout or in double support asymmetry. SignificanceThis study shows that able-bodied individuals retain savings of split-belt adaptation over a three-week period, which indicates that only naïve split-belt walkers should be included in split-belt adaptation studies, as previous experience to split-belt walking will not be washed out, even after a prolonged period. In future research, these results can be compared with long-term savings in patient groups, to gain insight into factors underlying (un)successful gait training in rehabilitation.