Feedback-generated muscle activity reveals age-dependent decline of motor learning

Recent studies suggest that planned (feedforward-mediated) actions and corrective (feedback-mediated) movements are recalibrated during motor adaptation. However, feedback commands in locomotion are thought to simply reflect sudden environmental changes upon introduction or removal of novel situations independently from sensorimotor recalibration. Thus, we asked if human feedback activity can indicate the motor system’s adapted state in walking. We recorded electromyographic (EMG) signals bilaterally on 15 muscles before, during, and after split-belt walking that induces sensorimotor adaptation by moving the legs at different speeds. We exploited the temporal dynamics of feedback commands to isolate them from the overall motor output. We found that EMG aftereffects were dominated by feedback responses, which were structurally different after short vs. long adaptation periods. Only after a long adaptation duration, the structure of each leg’s feedback activity resembled the balance-like responses elicited in the contralateral side when the split condition was first introduced. This mirrored feedback activity was a proxy of motor learning at an individual level since it indicated the extent to which each subject adopted the split pattern as their new “reference” gait, such that deviations from it were processed as an opposing perturbation to the one originally experienced. Interestingly, this mirroring decayed with age, but steady state changes during split-belt walking did not, suggesting potentially different neural mechanisms underlying these motor patterns. Taken together our results show that feedback-commands reflect the adapted state of the motor system, which is less flexible as we age. Significance statement We showed that feedback activity elicited by sudden environmental transitions is revealing of the underlying recalibration process during sensorimotor adaptation, and hence indicates individual learning capacity. This allowed us to identify age-related decline in motor learning that was not discernible from kinematic measures conventionally used in motor adaptation studies. These findings suggest that older populations may have limited potential to correct their movements through error-based protocols simply given their age. Moreover, we found that split-belt walking recruits distinct motor patterns during and immediately after the split condition, informing our understanding of the therapeutic effect of this task. Therefore, our detailed EMG characterization provided valuable normative data of muscle activity that could be reinforced with repeated exposure of split-belt walking.

The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint 2 Materials and Methods 72 2.1 Subjects 73 A group of 16 healthy subjects of ages ranging between 46 to 78 years old (61+-9.9 y.o., 10 female) participated 74 in the study (see Table 2). Their movements and muscle activity were recorded before, during, and after walking 75 on a split-belt. Subject S01 was excluded from our analyses because this individual behaved strangely during 76 the split-belt walking period, which likely compromised this subject's motor adaptation (see Figure 1-1). All 77 subjects provided written informed consent prior to participating in the study, which was approved by the 78 Institutional Review Board at our Institution, and was in accordance to the declaration of Helsinski. 80 We assessed the adaptation and de-adaptation of muscle activity through the protocol illustrated in Figure   81 1A. The protocol consisted of treadmill walking in six different conditions presented in the following order: 82 Slow Walking (50 strides), Mid Walking (50 strides), Short Exposure (10 strides), Baseline Mid Walking (150 83 strides), Adaptation (900 strides, in three blocks of 300) and Post-Adaptation (600 strides, in two blocks of 84 300 or a single 600 block). Slow and mid walking were used to quantify speed-dependent modulation of 85 muscle activity during regular treadmill walking. Short Exposure and Adaptation were used to differentiate 86 learning-dependent changes in muscle activity that required multiple steps in the split environment from those 87 solely due to unexpected introduction or removal of the split perturbation. Baseline walking was used as a 88 reference for steady state walking on the treadmill prior to motor adaptation. Finally, the Post-Adaptation 89 condition was used to evaluate adaptation aftereffects in muscle activity following split-belt walking. The 90 Adaptation and Post-Adaptation conditions were designed to have enough strides to examine the evolution 91 of muscle activity from a transient to a steady state when the perturbation was either introduced or removed, 92 respectively. These conditions were divided into blocks to minimize fatigue and subjects were instructed not to 93 step between blocks to prevent deadaptation due to unrecorded steps. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint the dominant leg (self-reported leg used to kick a ball) walking faster. We refer to the dominant leg as the fast 97 leg throughout the text and the non-dominant one as the slow leg. In all other conditions, belts moved at the 98 same speed relative to one another (i.e., 'tied' mode) ( Figure 1A). The treadmill was started and stopped at the 99 beginning and at the end of each condition and speeds did not change while each condition was ongoing. Belts 100 moved at one of three possible speeds throughout the experiment: a self-selected ('mid') walking speed, a 'slow' 101 speed of 66.6% of the mid speed, and a 'fast' speed of 133.3% of the mid speed. We used self-selected speeds to 102 ensure subjects from all ages could complete the entire protocol. The self-selected speed was obtained by first 103 averaging each subject's speed when walking over ground in a 50-meter hallway during 6 minutes (i.e., 6-min 104 walking test (Rikli and Jones, 1998)) and then subtracting 0.35m/s, which resulted in a comfortable walking 105 speed on a treadmill based on pilot data in older adults (> 65 yrs). Mid speed was 0.72 ± 0.26m/s (mean ± 106 standard deviation) across the population. Speed values used for all subjects during baseline walking are shown 107 in Table 2. 108 109 Safety measures were designed such that participants from older ages could complete the study. First, all 110 subjects wore a harness that only provided weight support in the event of falling but not during walking or 111 standing. Also, subjects were told a few seconds in advance that they could hold on to a handrail (directly 112 located in front of them) whenever a condition or block started or finished. Subjects were encouraged to let go 113 of the handrail as soon as they felt comfortable doing so to minimize the effect of external support on muscle 114 recordings. Finally, we monitored subjects' heart-rate continuously and blood-pressure during the rest breaks 115 to prevent over exertion in any of the participants. Data collection 118 We collected electromyographic (EMG) signals, kinematics, and kinetic data to characterize subjects behavior.

119
Surface EMG signals from 15 muscles on each leg were recorded for all subjects (see Table 1 for full list and 120 abbreviations) at 2000Hz using a Delsys Trigno System (Delsys Inc., Natick, Massachussets). Signals were 121 6 . CC-BY-NC-ND 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint highpass-filtered to remove undesired movement artifacts and then rectified. We used a 2nd order Butterworth 122 filter (dual-pass) with a cutoff frequency of 30Hz, which resulted in 80 dB/dec attenuation and zero-lag (Merletti 123 and Parker, 2004). Unlike other studies (e.g., Torres-Oviedo and Ting 2007), we did not apply a subsequent 124 lowpass filter following rectification as we did not require the EMG envelope for our analysis (see below).

125
Kinematic data was collected at 100 Hz with a passive motion analysis system (Vicon Motion Systems, Oxford, 126 UK). Movements were recorded by placing reflective markers bilaterally on bony landmarks at the ankle 127 (i.e., lateral malleolus) and the hip (i.e., greater trochanter). Ground reaction forces were recorded with an 128 instrumented split-belt treadmill (Bertec Corporation, Columbus, Ohio) and sampled at 1000Hz. Forces along 129 the axis of gravity (Fz) were used to determine when the foot landed (i.e., heel-strike: Fz > 10N) or was lifted 130 off the ground (i.e., toe-off: Fz < 10N).

132
Data Analysis 133 EMG parameters: We characterized the changes in EMG during and after split-belt walking by first generating 134 muscle activity variables that discretized each muscle's activity over the gait cycle ( Figure 1B). Filtered EMG 135 activity was divided in sub-intervals of the gait cycle aligned to gait events to focus on changes in muscle 136 activity within the gait cycle, rather than on activation changes due to differences in timing of the gait cycle 137 across the distinct walking conditions (Dietz et al., 1994;Reisman et al., 2005). For example, in Figure 1B 138 we present sample EMG traces for Baseline walking at medium speed (gray trace) and late Adaptation (green 139 traces) for the leg walking slow (top panel) and the one walking fast (bottom panel). More specifically, we 140 divided the gait cycle of each leg into 6 intervals according to well defined gait phases (Perry and Burnfield, 141 2010): first double-support (from ipsilateral heel-strike to contralateral toe-off), early single-stance (first half 142 of single stance), late single-stance (second half of single stance), second double support (from contralateral 143 heel-strike to ipsilateral toe-off), early swing (first half of swing phase) and late swing (second half of swing 144 phase). In order to achieve better temporal resolution, each of these 6 intervals were further divided into two 145 equal sub-intervals, yielding 12 intervals for each gait cycle. Note that the normalized gait cycles appear to 146 be the same duration because of our normalization procedure. However, each of these sub-intervals' duration 147 7 . CC-BY-NC-ND 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint ranged between 75 and 120 ms. The precise timing for each sub-interval throughout the different epochs of the 148 study are presented in Table 3. Muscle activity amplitude was averaged in time for each of these sub-intervals, 149 for every stride and muscle resulting in 360 muscle activity variables per subject: 12 intervals x 30 muscles.

150
Sample of these variables for one muscle in one gait cycle for Baseline (B raw ) (gray row) and late Adaptation 151 (LateA raw ) (green row) are presented in the rows below the EMG traces in Figure 1B.

152
These EMG variables were normalized using the activity during Baseline walking as a reference. To this end, 153 we first computed the mean activity for each sub-interval across the last 40 strides (i.e., steady state) of Baseline 154 walking. Then each muscle's activity was scaled such that the least and most active phases for every muscle 155 during baseline had the value of 0 and 1, respectively. We specifically subtracted the mean value for the least 156 active phase in Baseline from the activity in all sub-intervals for all strides in all conditions and we subsequently 157 divided the shifted values by the mean baseline value for the most active phase. Following this normalization, 158 the units of EMG activity for each muscle of a given subject were represented with respect to the average 159 maximum value recorded in that muscle during Baseline walking of the same subject. This scaling allowed us 160 to aggregate subjects and compare effect sizes across muscles even when recorded EMG amplitudes were very 161 different because of sensor placement or underlying tissue properties ( Figure 1D).

163
Kinematic parameters: The adaptation of movements was characterized with step-length asymmetry, which is 164 a metric known to change during and after split-belt walking (Reisman et al., 2005). We computed step-length 165 asymmetry on each stride cycle by calculating the difference in step lengths (i.e., ankle to ankle distance at 166 foot landing) for two consecutive steps taken with the fast and slow leg. This difference was normalized by the 167 sum of step lengths to obtain a measure that was a proportion of each subjects' step sizes. A zero step-length 168 asymmetry value indicated that steps lengths were even, negative values indicated that the (non-dominant) leg  We also computed body displacement with respect to the foot in contact with the ground during the stance phase 172 for each leg. This was done to interpret the changes in muscle activity upon transitions between tied and split The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint conditions. Body displacement was computed as the anterior-posterior distance between the middle of the hip 174 markers (greater trochanter) and the ankle from ipsilateral heel-strike to contra-lateral heel-strike. To focus on 175 displacement changes within this period of the stance phase, we aligned and divided the body displacement in 176 two sub-intervals: double support (from ipsilateral heel-strike to contra-lateral toe off) and single stance (from   We compared muscle activity for a given epoch with that of Baseline walking ( Figure 1D). This allowed us to 189 characterize how motor commands differed from normal walking along the adaptation process. The comparison 190 was done by calculating differences in muscle activation variables between an epoch of interest (e.g. LateA raw ) 191 and those during Baseline (B raw ) walking (e.g., EarlyA = EarlyA raw − B raw ). We also computed differences 192 in muscle activation variables between consecutive epochs in which subjects transitioned from the split-to-tied 193 (e.g., FBK split−to−tied = EarlyP − LateA) or tied-to-split conditions (e.g., FBK tied−to−split = EarlyA − B = 194 EarlyA). The magnitude of these was calculated by computing the euclidean norm of the 360-dimensional 195 vectors (e.g., FBK tied−to−split ). This offered a quantification of overall changes in muscle activity with one 196 single metric, which was particularly relevant for EarlyP since it indicated the aftereffect size in muscle space. and Shadmehr, 2016), which is not the case for walking. Thus, we propose to dissociate the adaptation of feedforward and feedback components of muscle activity through an operational definition that exploits the distinct temporal dynamics of these processes. We specifically assume that feedforward components change slowly upon experiencing a novel situation because they are generated purely through internal models that are updated gradually. On the other hand, feedback components can change rapidly after a transition in the walking environment since sensory information is available immediately following a change in the walking context.
Note that our definition of feedback-generated motor commands includes modulation of muscle activity that occurs at different latencies after a movement disturbance (i.e., short-and long-latency reflex responses and subsequent voluntary responses) (Horak et al., 1990). In sum, we quantified feedback-generated activity as the changes in EMG after a sudden environmental transition (i.e., EMG a f ter ) with respect to EMG activity before the transition (i.e., EMG be f ore ). In equation form: Where EMG is the measured muscle activity whereas FF and FBK are the feedforward-and feedback-generated muscle activity, respectively. We assume that feedforward components change slowly upon sudden transitions: Thus: ∆EMG = EMG a f ter − EMG be f ore = ∆FBK + ∆FF ≈ ∆FBK = FEEDBACK-GENERATED ACTIVITY Therefore, even if the change in EMG is strictly equal to the change in both feedback-and feedforward-generated 199 activity, we approximated the latter one to be negligible because it occurs at a slower rate compared to changes The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint Where LateA and EarlyP are the unbiased muscle activity before and after the split-to-tied transition, respectively.

203
This equation simply formalizes the idea that feedback-generated activity might change immediately after 204 removing the split-belts environment. Importantly, the muscle activity aftereffects (EarlyP) are not the same 205 as our definition of the feedback-generated activity upon removal of the split condition (Eq. 1), but contribute 206 to it. Namely, we anticipate that aftereffects will be composed of the learned motor pattern carried over from  One possibility is that muscle activity is mostly controlled in a feedforward manner when the split perturbation is removed. In other words, feedback responses are small in magnitude when compared to the feedforward components. Thus, muscle activity before (LateA) and after (EarlyP) the split-to-tied transition are roughly the same, as illustrated in the schematic of c1 in Figure 1C. This possibility suggests that feedback components are present during the tied-to-split transition (EarlyA), but are greatly reduced during the split-to-tied transition because the return to a familiar environment (i.e., tied condition) may not require the same extent of feedback responses as when a novel condition (i.e., split) is experienced. We formulated this case on the basis that aftereffects result from the continuation of the motor output updated during the Adaptation period (Malone et al., 2012). Formally expressed: The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint 2.4.2 Case 2 (C2): Feedback-generated motor commands are environment-dependent 216 Another possibility is that feedback responses are not negligible during post-adaptation, but they are not adaptive in nature. That is, the feedback processes generating them are not updated during split-belt walking and produce corrective muscle patterns solely as a function of the sensed environment. There are two expected observations if feedback motor commands are purely environment-dependent. First, feedback-generated activity would change similarly after removing the split perturbation following a long or short exposure to the split condition because the environment transitions identically in these two situations. Second, we would expect that changes in muscle activity originated by the tied-to-split transition (i.e., FBK tied−to−split ) would be numerically opposite from those from the split-to-tied transition ( Figure 1C, c2 schematic). In other words, any muscle whose activity increased when the split perturbation was introduced, should decrease by the same amount when it is removed (and vice versa). We reasoned that this feedback case would be plausible given that split and tied walking require distinct motor patterns and subjects might switch between them. Thus, the immediate changes in activity would be numerically opposite, as expected when switching back and forth between two motor patterns (A to B vs. B to A). Formally expressed: Where EarlyA represents muscle activity during the Early Adaptation with respect to Baseline. Finally, we considered the possibility that substantial feedback responses are not only present following transitions in walking speeds, but they are indicative of the motor system's recalibration during sensorimotor adaptation.
In this case, changes in activity from the split-to-tied conditions would be different after a short vs. a long exposure to the split environment. We specifically considered that following the long exposure, the motor system would learn that the split (perturbed) condition is the "new normal". Consequently, removal of the split condition would be processed as the opposite perturbation to the one experienced when the novel environment was first introduced. We based this possibility on previous work reporting that the removal of the novel condition 12 . CC-BY-NC-ND 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint is in-and-of itself a perturbation (Herzfeld et al., 2014b), and that feedback responses are adapted through experience to an altered environment even if there is no feedback-specific learning opportunities (Wagner and Smith, 2008;Yousif and Diedrichsen, 2012). During the split-to-tied transition in our experiment the leg that was 'fast' moves down in speed, while the one that was 'slow' moves up. This is the mirrored version of the initial tied-to-split transition where the 'fast' leg went up and the 'slow' leg went down. Accordingly, we expect the feedback-generated muscle activity to reflect this mirror-symmetry between belt-speed transitions when the split condition is introduced or removed. Formally expressed:  In summary, we tested the extent to which each one of these three possibilities shaped the structure of feedback-generated 222 components of aftereffects with one single regression model: Where the parameters β S and β M are respectively interpreted as the extent of feedback-generated activity that 224 is solely environment-dependent (β S ) or adaptive (β M ). Thus, the three cases presented in the previous section The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint patterns of activity. It is worth pointing out that our regression analyses rely on the assumption that initial muscle 235 responses during early adaptation EarlyA are not anti-symmetric (i.e., same group of muscles increasing in a leg, 236 decrease in the other one). Under this assumption, the mirrored EarlyA * and opposite −EarlyA muscle vectors 237 are different from each other in muscle space (i.e., not colinear). This assumption was confirmed empirically, 238 as the cosine of the angle formed by the two vectors was −0.13 ± 0.38 (median ± inter-quartile range across 239 subjects). were determined by using the Wilcoxon signed-rank test (i.e., non-parametric analogue of a paired t-test) on 243 each of the 360 EMG parameters. Effect sizes were computed using median values across subjects since this is a 244 measure less susceptible to outliers. All tests were two-tailed and the null hypothesis was that the (normalized) 245 absolute effect size was smaller than 0.1 (i.e., 10% of the maximum baseline activity for that muscle). This was 246 done to avoid finding significant but small, and presumably meaningless, differences. We corrected for multiple  given that our regression model did not include intercept terms. We compared the regressors obtained for data 255 following the Short Exposure and Adaptation epochs using a two-tailed paired t-test. We report p-values, as 256 well as mean changes and well as Cohen's d for effect size. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint walking speed could explain the large inter-subject variability that we observed in the regression coefficients. 260 We focused on these subject-specific features because they exhibited large ranges in our cohort that could 261 have impacted our results. We also studied the association between aftereffects sizes in muscle activity and 262 step-length asymmetry. For all these analyses we applied Spearman's correlations as a non-parametric alternative 263 to Pearson's correlation because it is more robust to potential outliers. The correlation value (lowercase r, to 264 distinguish from previous use of R) and the corresponding p-value were presented.   (Figure 2A). In other words, subjects were either "falling" forward or backward 275 during early Adaptation immediately after hitting the ground (heel-strike) with the fast or slow legs, respectively.

276
This was likely due to the discrepancy between subjects' expected and actual walking speed ( Figure 2B).

277
Consistently, the corresponding muscle activity for both legs during 100-400ms (colored rectangles) after   Figure 3B). Namely, we found 301 that if a group of muscles increased activity on one side, the same group decreased activity on the other one.

302
Interestingly, this opposing modulation across legs was not merely determined by ipsilateral walking speed. 303 Should this have been the case, one would expect reduced muscle activity of the slow leg with respect to  The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint 3.3 Feedback responses are mirrored between the legs when returning to tied-belts after prolonged exposure 317 to split-belt walking 318 We proposed that aftereffects (EarlyP) would be generated by a combination of feedforward and feedback 319 processes. In particular, we considered the possibility that feedforward-generated motor commands would 320 dominate EMG aftereffects (C1), such that muscle activity during Early Post-Adaptation would be similar to that 321 of Late Adaptation (LateA). Qualitatively, we found little resemblance in the muscle activity during these two 322 epochs, as shown by the few significant changes that were similar between them (black outlines in Figure 3B and 323 3C). This indicated that feedback-generated motor commands also contributed to muscle patterns during Early The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint median activity (cosine of 0.778) and individual data (median ± inter-quartile range: 0.528 ± 0.232) than 344 FBK split−to−tied and −EarlyA for group (cosine of 0.367) and individual data (median ± inter-quartile range:  Our regression and cosine-based results are qualitatively supported by the remarkable similarity between the 360 slow leg's pattern at tied-to-split transition (Fig. 4C, top half) and the one of the fast leg at split-to-tied transition 361 (Fig. 4D, bottom half). This mirroring was also observed when comparing the contralateral sides (Fig. 4C 362 bottom half to 4D top half) at these transitions, except for the reduced activity in muscles RF, VL, VM, GLU, 363 TFL during early stance only observed in FBK split−to−tied (Fig. 4D). We also found that the body motion 364 exhibited mirroring between the legs' stance phases. For example, the change in body position with respect to 365 the fast stance leg in the tied-to-split transition (Fig. 4F, green trace) closely resembled the one during slow 366 stance at the split-to-tied transition (magenta trace). Thus, both EMG and kinematics exhibited mirroring when 367 we consider changes with respect to the previous steady-state condition. Further, note that these body changes 368 were distinct from the actual location of the body with respect to baseline walking illustrated in gray. In sum, 369 these results suggest that the nervous system (specifically, the feedback response mechanisms) adopted the The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint steady-state behavior as a new reference gait pattern. Therefore, kinematic perturbations away from this new 371 reference, rather than perturbations away from the subjects' baseline behavior may drive feedback-generated 372 muscle activity. 373 Lastly, to confirm that the mirroring between legs is a learning-dependent process, and not simply due to 374 removal of the split perturbation, we ran the same regressions on the split-to-tied transition following the 375 Short Exposure condition (i.e., we used the exact same regression factors as in the analysis for the (long) 376 Adaptation period). In this condition subjects did not have time to adapt, so we expected to observe solely 377 environment-dependent changes in EMG activity (C2), rather than mirroring (C3). Indeed, that is what we  The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint 3.4 Healthy aging reduces the adaptation of feedback-generated muscle activity contributing to aftereffects 396 Analysis of intersubject variability revealed that older adults exhibited less adaptation of feedback-generated 397 activity. Namely, β M and β S were associated to subjects' age (r = −0.63, p = 0.014 and r = 0.69, p = 0.0058 398 respectively, Figure 5A) with older subjects showing smaller β M and larger β S . This indicates that feedback 399 processes in older adults was less adapted (and more environment-driven) compared to younger adults. We also 400 noticed that the magnitude of feedback activity were possibly smaller in older adults ( FBK split−to−tied : r = 401 −0.69, p = 0.0055; FBK tied−to−split : r = −0.52, p = 0.049; Figure 5B). These smaller feedback responses 402 in older individuals could make it more difficult to identify the structure of feedback activity because of the 403 reduced signal to noise ratio, leading to biased or noisy estimates of β M and β S . To discard this possibility, we 404 correlated the R 2 of the fitted models with age. We found no effects of age (r = −0.43, p = 0.11, not shown), 405 meaning that the regression model applied to individual data captured comparable levels of variance regardless The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint learning. This suggests that the adaptation of muscle activity may be more sensitive to individual differences in 422 learning than kinematics. Alternatively, there might be partially different mechanisms involved in the adaptation 423 of muscle activity and step length asymmetry. In agreement, muscle activity aftereffects and kinematic aftereffects 424 were not correlated (r = 0.346, p = 0.206, not shown). Based on these latter regression results one could 425 consider that there are two different learning mechanisms: one that underlies the association between age and 426 EMG aftereffects, and a separate one captured by kinematics aftereffects. In sum, we observed diminished 427 mirroring of feedback activity post-adaptation and smaller EMG aftereffects in older adults, suggesting that 428 healthy aging has a negative impact on learning process updating feedback-generated muscle activity. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint 4 Discussion 430 We showed that EMG aftereffects were not merely a continuation of motor patterns learned during split-belt 431 walking, but exhibited a large contribution of feedback-generated motor commands. Importantly, the structure 432 of feedback activity post-adaptation resembled balance-like responses that were mirrored between the legs when 433 compared to those elicited in response to the split perturbation. This mirrored feedback activity indicated 434 that departures from an updated "reference" gait induced corrective responses similar to those when a novel 435 environment was introduced. Therefore, we interpret the mirrored feedback commands as a proxy of motor   The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint The learned (steady-state) pattern continued in only a few muscles when the split condition was removed, 457 which was to a certain degree unexpected given that kinematic aftereffects are thought to reflect a continuation The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint Second, feedback policies transforming the sensed state into corrective actions may rely on internal models that 484 are updated through exposure to an alternative environment (Bhushan and Shadmehr, 1999). Lastly, the state 485 estimation following split-belt walking may be inaccurate due to sensory shifts (Vazquez et al., 2015) leading 486 to altered feedback commands even if the feedback policy remains unchanged. In conclusion, the structure 487 of feedback activity post-adaptation reflects the recalibration of the motor system during split-belt walking, 488 suggesting that the feedback loop generating feedback responses is also recalibrated. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint in internal model recalibration because they are a closer correlate of neural activity than kinematics. Further, 511 the magnitude of muscle patterns during late adaptation, which is mostly feedforward-generated activity, was 512 not affected with age. This dichotomy in age effects on feedback vs. feedforward activity supports that the 513 adaptation of these two processes is partially dissociated (Yousif and Diedrichsen, 2012). We conclude that 514 age-related sensory decline might contribute to motor learning deficits in older adults, which is observed in the 515 adaptation of feedback responses, but not in the steady state motor pattern.

667
Full description and rationale are presented in the Methods (section 2.4). In brief, One possibility (C1) is that 668 FBK split−to−tied is much smaller than feedforward activity, which carries over from the split to tied condition. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint For example, TA on the slow leg is active during stance, which does not occur during normal walking. This activity is consistent with pulling the COM forward, closer to its expected position.

37
. CC-BY-NC-ND 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint for both late Adaptation and early Post-Adaptation. We propose these patterns of activity (aftereffects) are generated by a superposition of feedforward-and feedback-generated activity. P-value threshold: p = 0.035.

38
. CC-BY-NC-ND 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint  Figure 2C. P-value threshold: p = 0.041. (D) Split-to-tied change in EMG data (early Post-Adaptation with respect to late Adaptation). We observe that changes in activity in the fast (slow) leg during this transition resemble those of the slow (fast) leg during the tied-to-split transition (mirroring). For example, TA is not active during stance in steady-state walking (see Figures 1D and 4A,B), but it is active following the tied-to-split transition in the slow leg, and the split-to-tied transition in the fast leg. P-value threshold: p = 0.035. (E) Quantification of environment-dependent vs. adaptive feedback activity in split-to-tied transitions, after the short exposure (SE, gray) and long exposure (LE, magenta) to the split-belts environment (group averaged data). Black dots represent the results expected from two of the cases discussed in the Methods (C2, C3). (F) Changes in hip position (with respect to previous steady-state) after the two transitions (tied-to-split in green, split-to-tied in magenta). Line represents mean across subjects, shaded area is standard error. We observe these the kinematics of these two transitions are approximately mirror images of one another, possibly causing the observed mirroring of EMG responses. For comparison, we also present early Post-Adaptation with respect to Baseline (EarlyP, gray).

39
. CC-BY-NC-ND 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint age, but when returning to tied-belts ( FBK split−to−tied ) there is a significant age effect. (C) Size of muscle activity modulation during late Adaptation. (steady-state). No correlation to age was found. This confirms that older subjects are able to modulate muscle activity as much as healthy subjects. (D) Step-length asymmetry aftereffects are also not correlated with age. (E) Size of muscle activity modulation during early Post-adaptation (aftereffects). Aftereffects are correlated with age. This shows EMG-based measures of learning are more sensitive than kinematic-based ones.

40
. CC-BY-NC-ND 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint Step-length asymmetry The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/372359 doi: bioRxiv preprint