We recruited 19 healthy young adults and 19 healthy older adults aged 55 or older (Table 1). All participants completed written informed consent prior to study participation. Sample size of 19 per group was determined based upon effect sizes from previous literature (Bruijn et al. 2012b) using F-test power analysis within G*Power 3.1, assuming β error probability of 0.8 and α error probability of 0.05. Participants were free from cardiovascular, pulmonary, renal, metabolic, vestibular, and neurologic disorders and reported no lower-extremity injuries or surgeries in the past 12 months that might limit their capacity to complete the protocol or alter their gait. All participants reported being able to walk unassisted without balance issues, and all were naïve to split-belt treadmill walking. The protocol was approved by the Auburn University Institutional Review Board before any subject enrollment and performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki.

Before treadmill testing, all participants completed cognitive tests including the Mini-Mental State Exam, Digit Span Forward and Backward, and Trails Making Test A and B (Blackburn and Benton 1957; Folstein et al. 1983; Bowie and Harvey 2006). All cognitive tests were performed in the same order and by the same administrator. Retroreflective markers were then placed according to the Vicon Plug-In Gait Full Body Ai Functional model, and kinematic data were collected at 100 Hz (VICON; Vicon Motion Systems Ltd, Oxford, United Kingdom). Next, participants began testing on the split-belt treadmill. Participants were instructed to hold on to the handrails for all trials. We first obtained each participant’s typical and fast walking speeds while they walked on the treadmill with both belts moving at the same speed. To calculate comfortable walking speed, the belts began at a relatively slow speed and then increased until the participants reported reaching their typical speed. This was repeated twice, and the average of the reported values was recorded as the participant’s typical walking speed. Similarly, to calculate the participant’s fast walking speed, the belts began at a relatively slow speed and increased in speed until the participants felt like they were walking at the fastest speed that they could comfortably walk at for ten minutes. Two trials were recorded, and the average value was reported at the participant’s fastest walking speed. The participant’s fastest walking speed was set as the fast belt speed, and half of the fastest comfortable speed was set as the slow belt speed. Previous studies have undertaken a similar methodology to determine fast and slow belt speeds (Dingwell and Marin 2006; Roemmich et al. 2014b). To determine leg dominance, we asked participants which leg they would use to kick a soccer ball. We placed their self-reported dominant leg on the slow belt of the treadmill.

Participants then walked for three minutes with both belts at their typical walking speed, followed by three minutes with both belts at their fastest and three minutes at slow walking speed (BASELINE). This three-minute walking trial at slow walking speed served as the baseline trial, as split-belt literature has recommended matching the speed of baseline trials to the speed at which the after-effects will be tested during the deadaptation phase. It has been shown that the largest treadmill after-effects occur when the tied-belt speed matches that of the slower belt during split-belt adaptation (Vasudevan and Bastian 2010; Vasudevan et al. 2017). Next, the belt under the participant’s non-dominant leg was increased to their fastest walking speed while the belt under the dominant leg remained at their slow walking speed. Participants walked under these conditions for ten minutes (ADAPT) and subsequently walked with both belts at the slow speed for three more minutes (DEADAPT). The belts were stopped temporarily between each treadmill condition before initiating the next condition.

All treadmill walking trials took place on a Bertec instrumented spit-belt treadmill (Bertec, Columbus, OH, USA). Kinematic data were filtered using a low-pass fourth-order Butterworth filter with a cutoff frequency of 6 Hz. Step length was calculated as the anteroposterior distance between the ankle markers at foot strike. SLA was calculated and normalized to stride length using the following equation:

$$Step length Asymmetry= \frac_-_ }_+_}$$

Step Lengthfast refers to when the leg in the fast belt strikes the belt, and Step Lengthslow refers to when the leg on the slow belt strikes the belt. Positive asymmetry indicates that the leg on the fast belt is taking a longer step, and negative asymmetry indicates that the leg on the slow belt is taking a longer step. An asymmetry value of zero indicates that the legs are taking steps of equal length. The calculation of SLA is the same equation as used in previous studies that have examined SLA during split-belt treadmill walking (Bruijn et al. 2012b; Roemmich et al. 2014b; Malone and Bastian 2016; Roper et al. 2019b). Previous studies have demonstrated that SLA can be decomposed into a sum of contributions related to space, timing, and velocity (Finley et al. 2015b; Long et al. 2016b). These measures encompass the location, timing, and velocity of the feet during walking. Accordingly, we can calculate SLA’s spatial, temporal, and velocity contributions.

$$Step length Asymmetry=\left(_- _\right) +\left[\frac_+_} X \left(_-_\right)\right] + \frac_+_} X (_-_)$$

The first term represents the spatial contribution to SLA based on where participants placed their feet relative to their bodies at heel strike. \(_\) refers to where the fast foot is placed relative to the previous slow foot placement, while \(_\) refers to where the slow foot is placed relative to the previous fast foot placement. The second term represents timing contributions to SLA based on step timing differences. \(_\) refers to the time between a slow foot heel strike and the previous fast foot heel strike while \(_\) is the time between a fast foot heel strike and the previous slow foot heel strike. Additionally, \(_\) is the average velocity of the slow ankle relative to the body during the slow step time, while \(_\) is the average velocity of the fast ankle relative to the body during the fast step time. The final term represents velocity contributions related to the differences in belt speeds. The spatial, temporal, and velocity contributions to SLA were derived from non-normalized SLA values.

SLA values and the contributions to SLA were calculated at five different epochs throughout the split-belt treadmill paradigm: (1) the average of the last five strides of baseline at slow walking speed (BASELINE), (2) the average of the first five strides of ADAPT (Early-ADAPT), (3) the average of the middle five strides of ADAPT (Mid-ADAPT), (4) the average of the last five strides of ADAPT (Late-ADAPT) and (5) the average of the first five strides of DEADAPT.

Rates of adaptation and deadaptation were calculated as the number of steps a participant took until five consecutive strides were within two standard deviations of the plateau during adaptation and deadaptation. Plateaus were defined as the mean of the last thirty strides of adaptation and deadaptation, as described previously (Finley et al. 2015b; Brinkerhoff et al. 2021b).

All statistical tests were conducted using SPSS Statistics version 26 (IBM, Armonk, New York). One-way ANOVA analyzed demographic variables (Table 1), and repeated measures ANOVA contrasted differences in spatial and temporal contributions to SLA by age group (old versus young adults) and across adaptation epochs (Early-ADAPT, Mid-ADAPT, Late-ADAPT). Second, repeated measures ANOVA contrasted differences in SLA by age group (young versus old adults) and adaptation epochs (BASELINE, Early-ADAPT, Mid-ADAPT, Late-ADAPT, and DEADAPT).

To examine adaptation and deadaptation rates, we used repeated measures ANOVA across age groups (old versus young adults) and epoch (adaptation versus deadaptation). All analyses accounted for differences in the fastest walking speed and baseline asymmetries. Analyses of subtracting each participant’s initial baseline asymmetry from their subsequent gait measurement did not alter the study outcomes, ensuring that any observed effects were specifically due to the experimental conditions. For each statistical procedure, the significance level was α = 0.05, and Tukey post-hoc adjustments were applied when appropriate. Partial eta-squared (ŋ2) values are also reported as an index of effect size.

Before conducting the analyses, we verified normality and visually inspecting histograms, assessed for skewness and kurtosis, and conducted Shapiro–Wilk tests. We also investigated extreme data points that exceeded two standard deviations from the mean. However, excluding these data points did not alter the statistical inference or interpretation of the results. To assess the assumption of sphericity, we utilized Mauchly’s test. If Mauchly’s test indicated a violation of the sphericity assumption, we relied on adjusted degrees of freedom based on the Greenhouse–Geisser correction.