Limb-Specific Modulation of Muscle Synergies and Segmental Coordination During Curved Running

Limb-Specific Modulation of Muscle Synergies and Segmental Coordination During Curved Running | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Limb-Specific Modulation of Muscle Synergies and Segmental Coordination During Curved Running R. M. Mesquita, T. D. Toussaint, P. A. Willems, A. H. Dewolf This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8038490/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Curved running imposes asymmetrical mechanical demands on the lower-limbs and thus provides a model to test the adaptability of the locomotor system. This study investigated how muscle synergies reorganise during curved versus straight-line running. Using non-negative matrix factorisation of EMG signals, four synergies were extracted from both inner and outer limbs. While the overall modular structure was preserved, spatial patterns reorganised systematically while temporal patterns showed earlier onsets, particularly in synergies associated with push-off and late swing. Adaptations were limb-specific: the inner limb displayed greater reweighting and reduced complexity, while the outer limb showed more anticipatory shifts. Higuchi’s fractal dimension indicated reduced complexity in touchdown and late swing synergies but increased complexity in push-off, suggesting differential demands for robust versus finely tuned control. To corroborate this, kinematic analyses confirmed that curved running modified intersegmental coordination, with divergence of the covariation plane between inner and outer limbs, reflecting their distinct functional roles in redirecting versus propelling the body. Together, these findings indicate that curved running use modular locomotor control strategies which combine robust rhythm-generating spinal networks with asymmetric, limb-specific modulation. This coordination likely arises from the interaction of feedforward CPG activity with feedback-driven adjustments at mechanically critical phases of the gait cycle, providing new quantitative evidence for the flexible yet stable organisation of human locomotion. Curved running Muscle Synergies Muscular Coordination Kinematic Coordination Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Locomotion arises from complex, multi-layered interactions between central neural networks, reflex pathways and the musculoskeletal system, which interacts with the environment to produce effective movement (Nishikawa et al., 2007 ). The spinal cord plays a key role in integrating both feedforward, predictive signals, with feedback from sensory pathways to regulate locomotion (Ijspeert & Daley, 2023 ). Evidence from both animal and human studies show that dedicated spinal circuits, i.e. , central pattern generators (CPGs), organise locomotor activity (Grillner, 1985 ; Ijspeert & Daley, 2023 ; Minassian et al., 2017 ). These networks of spinal interneurons, generate basic rhythmic and patterned motor outputs which are modulated by a supraspinal neural “drive” to produce coordinated locomotion in different situations, reducing the computational demands of higher brain centres (Ivanenko et al., 2006a ; Lacquaniti et al., 2012 ; Minassian et al., 2017 ). This feedforward command is strongly coupled with limb sensory feedback (Dzeladini et al., 2014 ; Grillner & El Manira, 2020), which is essential for maintaining the timing and adaptability of CPGs and coordination between the limbs, particularly in unstable conditions and/or when limbs act asymmetrically (Fujiki et al., 2018 ; Pearson, 2008 ). In humans, indirect evidence suggests that CPGs are distributed across lumbar and sacral segments, with partially autonomous circuits dedicated to each limb (Danner et al., 2015 ; Dewolf et al., 2022 ; MacLellan et al., 2014 ). This limb-specific organisation allows for a flexible coordination of left–right activity and more robust adaptation to changing biomechanical demands. Behavioural goals dictate movement, and one must adapt to diverse spatial constraints to generate coordinated motion to match the required task such as walking and running, avoiding obstacles or changing directions (Alexander, 2002 ). In nature, situations where one moves only in a straight line are probably rather infrequent and, as such, adaptions due to changes terrain and direction require modulation of both the spinal motor output and structured limb coordination to adapt to navigational constraints (Courtine & Schieppati, 2004 ; Dewolf et al., 2019 ; Dewolf et al., 2020 a; Santuz et al., 2018 a). In a recent group of studies, we investigated how humans adapt when running on curves at different speeds (Mesquita et al., 2024 a, Mesquita et al., under review ). Our findings showed that the mechanical work required to sustain centre of mass (CoM) motion during turning increased by ~ 25% at the highest speeds, largely due to the additional work needed to manage lateral displacements. Furthermore, although the overall CoM work was balanced across limbs, functional asymmetries emerged: the inner limb contributed more to redirecting the CoM inward, while the outer limb primarily supported forward progression. To understand how these asymmetries were managed, we analysed joint kinetics in both sagittal and frontal planes (Mesquita et al., under review ). We observed that the work production shifted from proximal to distal joints, consistent, among others, with findings on unstable locomotion in birds (Daley et al., 2007 ). When grouping EMG data to observe the motoneuron (MN) activity at the spinal output level, using “spinal mapping” (Ivanenko et al., 2006b ), we observed a stronger increase in lumbar than sacral motoneuron activity. This matched inverse dynamics findings at the hip and suggested that lumbar networks play a dominant role in curved running, while sacral networks adapt more subtly to foot placement. Together, these kinematic, muscular, and spinal-level findings raise the question of how modular control of muscles, and the MN output itself, is shaped by both feedforward (CPG-driven) and asymmetric sensory feedback signals during curved running. Unsupervised machine learning techniques, such as non-negative matrix factorisation, applied to EMG data can capture the locomotor activity of large groups of muscles, thereby illustrating the integration between CPG-based patterns, sensory-driven MN adaption and supraspinal inputs (Ijspeert & Daley, 2023 ; Ivanenko et al., 2004 ). This approach allows one to investigate modularity and signal complexity of the MN output across spatial and temporal dimensions (Santuz et al., 2020 a; Santuz et al., 2018 a; Sylos-Labini et al., 2022 ). Building on the asymmetrical observations at the spinal output level in curved running, we expected muscular synergies themselves would follow similar modifications compared to straight line running and that inner and outer limb synergies would become asymmetric. Such modifications would indicate adjustments of the locomotor command, integrating both feedforward and feedback adaptions, and provide further support for the limb-specific organisation of MN output (Danner et al., 2015 ; Dzeladini et al., 2014 ; MacLellan et al., 2014 ; Yokoyama et al., 2018 ). To further understand these adaptions, we also investigated the ‘global’ kinematic coordination of the limb segments (the thigh, shank and foot motion), via inter-segmental coordination (Ivanenko et al., 2008 ). Various studies have shown that limb kinematics are similarly contained within the spinal circuitry and also originates from a feedforward central control loop coupled with peripheral sensory feedback to achieve the desired kinematic task (Lacquaniti, 2002 ; Lacquaniti et al., 1999 ; Poppele & Bosco, 2003 ). Indeed, the motion of the elevation angles in the lower limb segments covary along a plane, effectively simplifying the degrees of freedom of the locomotor system to two principle components : limb length and orientation (Bianchi et al., 1998 ; Ivanenko et al., 2008 ; Poppele & Bosco, 2003 ). The orientation of the covariation plane, along with the characteristics of the loop, defines a general kinematic pattern of locomotion. This pattern adapts to different locomotor conditions (Dewolf et al., 2018 ), adjusts with changes in speed (Ivanenko et al., 2007 ; Mesquita et al., 2024 b) and mechanical demands (Dewolf et al., 2018 ; Dewolf et al., 2020 a). Considering that plane orientation is limb function specific (Catavitello et al., 2018 ), when limbs have different functions we hypothesised that the plane describing the motion of the inner limb would diverge from that of the outer limb in curved running. We also expected that the hypothesised limb-specific shifts in muscle synergies to be mirrored by corresponding adjustments in intersegmental coordination changes. Materials & Methods Participants and experimental procedure The present experimental protocol description was adapted from Mesquita et al. ( 2024 a). Eleven recreational male runners (height: 1.81 ± 0.03m, mass: 74.91 ± 5.38 kg, age: 24.64 ± 2.91 years, mean ± SD) participated in the study. Informed written consent was obtained in accordance with the Declaration of Helsinki, and the study was approved by the UCLouvain Ethical Committee (B403201940199). A power analysis, conducted using GLIMMPSE 3.1.2 (Denver, Colorado, USA), confirmed that a sample size of 10 participants was required to achieve adequate statistical power (1- ß = 0.87) and avoid a type II error when considering the model for both kinematic and muscular activity variables. For this present analysis, participants ran at speeds ranging from 7 to 17 km h − 1 on a circular track with a 6 m radius of curvature. The track was placed on 16 1m x 1m instrumented force platforms (described at length with figures in Mesquita et al. 2024 a). After the curved session, they then ran on a treadmill, to allow for comparison with straight-line running (SLR). Each trial began with participants running along a 20 m straight corridor before entering the curved path, ensuring that they took at least one step on the curve prior to data collection. Participants were instructed to try and maintain a constant average running speed, firstly at intermediate speeds (~ 12–14 km h − 1 ), then at fast speeds (~ 16–18 km h − 1 ), followed by slow speeds (~ 8–10 km h − 1 ). After each trial, the runner’s average horizontal velocity and the validity of strides were verified and communicated to the participant. The number of trials per speed category depended on the how many trials it took to achieve ≥ 10 valid strides. Rest periods were provided as needed to prevent fatigue-related bias. A trial was considered valid if it included at least one stride (two consecutive steps) in which the horizontal velocity, calculated via the ground reaction forces, remained constant over a complete stride. A stride began when the vertical component of the ground reaction force became greater than 30 N on one foot and ended when the same foot touched the ground for the following stride: a total of 1021 strides were analysed. In the present study, ground reaction force data has only been used to divide strides. Experimental set-up The present study measured two types of variables. First, surface EMG activity in both lower limbs was recorded via 24 surface electrodes. Second, the kinematic movements of body limb segments were captured using high-speed infrared cameras. Two pairs of photocells were positioned ~ 6 m apart at each end of the running track, at level with the neck and were used to trigger and stop the data acquisition for all three systems simultaneously. Measurement of EMG variables The EMG activity of 12 bilateral lower limb muscles was recorded at 2.15 kHz with a Delsys Trigno surface electrode system of (Natick, MA, USA). Two EMG protocols (A and B) were established to increase the number of muscles recorded. Six subjects used protocol A and five B. In both protocols, 10 out of 12 muscles were the same; the gluteus maximus (Gmax), rectus femoris (ReFe), sartorius (Sart), gracilis (Grac), semi-tendinosum (SeTe), biceps femoris (BiFe), medial gastrocnemius (MeGa), lateral gastrocnemius (LaGa), tibialis anterior (TiAn), peroneus longues (PeLo) of both limbs were always recorded. The muscles that differed between protocols were the tensor fascia latae (TeFL) and vastus medialis (VaMe) for protocol A and gluteus medius (Gmed) and vastus lateralis (VaLa) for protocol B. Six subjects had protocol A applied and five for protocol B. After preparing the shaved skin with alcohol, we placed EMG electrodes based on the SENIAM protocol (seniam.org), the European project on surface EMG. To this end, we located the muscle bellies by means of palpation and oriented the electrodes along the main direction of the fibres (Winter, 2009 ). The raw EMG signals were down sampled to 1 kHz during post processing to match the other types of data. Measurement of kinematic variables Bilateral, full-body three-dimensional kinematics were recorded by means of a Qualisys system (Gothenburg, Sweden) equipped with 12 “Mocap OQUS 6+” cameras placed around the running track or the treadmill. Subjects were equipped with 43 retroreflective markers placed symmetrically on specific bony anatomical locations. The relevant markers for this study were those placed on the greater trochanters, lateral femoral epicondyles, lateral malleoli and fifth metatarsal-phalangeal joints. Kinematics were recorded at a sampling rate of 240 Hz and oversampled similarly during post-processing at 1 kHz. Data analysis All data were separated into the following speed classes based on the average horizontal velocity (in km h − 1 ): [7, 8[, [8, 10[, [10,12[, [12, 14[, [14, 16[, [16, 18[. As subjects had to freely match speeds according to vocal guides, in post-processing when less than half of the participants covered the speeds in the class, the group was removed from the analysis. Speed classes counted on average 9.6 subjects (range 7–11). The analysis was performed via custom-written programs using LABVIEW (National Instruments 2021, Austin, TX, USA) and Matlab software (MathWorks r2019a, Natick, MA, USA). EMG analysis The raw EMG signals were visually inspected for amplifier saturation and noise artefacts. The signals were then high-pass filtered (30 Hz), rectified and low-pass filtered (10 Hz) with a zero-lag fourth-order Butterworth filter to obtain each muscle’s linear envelope (Dewolf et al., 2020 b) (Fig. 1 ). Signals were then cut by stride and the geometric mean of all strides of the same speed and curvature or straight-line was calculated per subject. The geometric average was chosen as it is less sensitive to outliers and thus further reduces noise in the signal (Linssen et al., 1993 ). The signals were time-normalised to 400 points and to reduce residual baseline noise, the minimum signal for each envelope was subtracted (La Scaleia et al., 2014 ). This gave us a 28x400 matrix (muscles from both protocols x time-normalisation) per speed class, per radius class and per subject. The waveform amplitudes were also normalised based on the maximum values for each muscle per subject when running on the treadmill at all speeds. Muscle synergies for each limb were extracted from the muscular activity using a dimensionality-reduction non-negative matrix factorisation approach (NNMF), (Lee & Seung, 1999 ). For each subject, speed radius, and limb condition, the average muscle activation profiles were concatenated into an m x n matrix ( M ), where m is the number of muscles (28) and n is the number of time-normalised samples (400). The NNMF algorithm was applied to M to identify the underlying basic activation temporal patterns, P , and their associated time-invariant spatial coefficients, C , (Lee & Seung, 1999 ): \(\:M=P\:\text{X}\:C+\:\) e , (1) where e is the residual error matrix. The algorithm searches for an approximate solution to minimise the root-mean-square error between M and \(\:P\:\text{X}\:C\) . For a given number of motor components (see below), the factorisation uses an iterative method starting from randomly chosen initial conditions for P and C (Ivanenko, 2005 ; Sartori et al., 2013 ). Because the root-mean-square error may have local minima, the best solution was selected out of 100 iterations to find C and P from multiple random starting values. The number of spatial and temporal modules was fixed at 4, based on their temporal associations to specific mechanical events as observed in different SLR stable and unstable running conditions (Santuz et al., 2018 a; Santuz et al., 2020 b) (Fig. 2 ). The goodness of the pattern decomposition was assessed with the percentage of variability accounted (VAF; Torres-Oviedo et al., 2006 ), defined as: VAF = \(\:1-\frac{\text{S}\text{S}\text{E}}{\text{T}\text{S}\text{S}}\) , (2) where SSE is the sum of squared errors between the experimental and reconstructed data and TSS is the total sum of squares (Dewolf et al., 2019 ; Dominici et al., 2011 ; Sartori et al., 2013 ). The number of muscle synergies has been a topic of discussion (Cappellini et al., 2006 ; Cheung et al., 2020 ; Lacquaniti et al., 2012 ; Yokoyama et al., 2016 ), and several objective methods have been proposed to determine an appropriate VAF cut-off (Cheung et al., 2005 , 2009 ; Chvatal & Ting, 2013 ; d’Avella et al., 2003 ). When applying both a linear-fit (10⁻⁵) criterion based to observe when VAF increase between successive synergies becomes negligible (Cheung et al., 2005 ), or a ‘cusp method’ (Cheung et al., 2009 ), comparing the rate of VAF increase in real versus randomised EMG data, the median number of synergies per limb consistently converged to four. We therefore fixed the dimensionality at four synergies and compared differences between conditions within this common framework. The centre of activity (CoA) was computed for each temporal ( P) synergy pattern as in Martino et al., ( 2014 ) using circular statistics and was expressed in polar coordinates (polar direction denoted the instant t of the stride cycle expressed an angle ( \(\:\alpha\:\) ) going from 0 to 360°). The CoA of each synergy was calculated as the angle of the vector (1st trigonometric moment) that points to the centre of mass of that circular distribution using the following formula: $$\:\text{C}\text{o}\text{A}=\:{\text{tan}}^{-1}\left(\frac{\sum\:_{t=1}^{400}\text{sin}{{\alpha\:}}_{t}\:\text{x}\:{\text{s}\text{y}\text{n}}_{t}}{\sum\:_{t=1}^{400}\text{cos}{{\alpha\:}}_{t}\:\text{x}\:{\text{s}\text{y}\text{n}}_{t}}\right)$$ 3 . Higuchi’s Fractal Dimension analysis To assess the local complexity of synergy temporal patterns, Higuchi’s fractal dimension, HFD, was computed assuming that each time series exhibits scale-invariant self-similar structure properties (Gneiting & Schlather, 2004 ; Higuchi, 1988 ; Kesić & Spasić, 2016 ; Santuz et al., 2020 a; Santuz et al., 2020 b). For each synergy’s temporal activation pattern, P (t) [ P (1) , P (2) , ... P (n) ], k sets of new time series were constructed, where k was an integer interval between 2 < k < k max : $$\:{P}_{k}^{{t}_{0}}:\:{P}_{{t}_{0}},\:{P}_{{t}_{0}+k},\:{P}_{{t}_{0}+2k},\:\dots\:,\:\left[{t}_{0}+int\left(\frac{n-{t}_{0}}{k}\right)k\right]$$ 4 , t 0 is the first sample of ranges 1 to k , used to generate the subsampled series. Debate exists in the literature as to what value one should assign to k max (Kesić & Spasić, 2016 ). As in Santuz et al., ( 2020 b) we selected a k max which was the “most linear part of the log-log plot”. As such, we ran an HFD analysis for multiple values of k and selected the point where the plot was most linear for each synergy, the median value for all synergies, speeds and curves was k max = 40. Therefore, for each trial, the HFD of each muscle synergy was computed separately with the median k max and then compared across limbs and speed. The non-Euclidean length of each k subsample was then defined as: $$\:{L}_{{t}_{0}}\left(k\right)=\frac{1}{k}\left\{\frac{n-1}{int\left(\frac{n-{t}_{0}}{k}\right)k}\left[\sum\:_{i=1}^{int\left(\frac{n-{t}_{0}}{k}\right)}\left|{P}_{{t}_{0}+ik}-{P}_{{t}_{0}-\left(i-1\right)k}\right|\right]\right\}$$ 5 , and for every considered k step considered, the average of the k sets of lengths of the temporal pattern was defined as: $$\:L\left(k\right)=\frac{1}{k}\sum\:_{{t}_{0}}^{k}{L}_{{t}_{0}}\left(k\right)$$ . If L( k ) ∝ k − HFD , then the curve is fractal, exhibiting self-similar structure across scale, with dimension HFD explaining its complexity. HFD ranges from 1 ( e.g. , a smooth time series) to 2 ( e.g. , random white noise) and is independent on the amplitude of the signal. Dissimilarity index and weighted difference To assess the similarity between inner and outer limb temporal and spatial patterns, a simplified Procrustes analysis was used. Procrustes analysis quantifies the similarity between two matrices by optimally aligning them through translation, isotropic scaling, and rotation (Andreella et al., 2023 ; Zaidi & Harris-Love, 2023 ). Specifically, one matrix is considered as a reference, and the second matrix is transformed to minimise the Frobenius norm of their difference. Since we wanted to quantify similarities between specific synergies, each comparison was done between 1 x n spatial or temporal matrices and as such, no matrix rotation was employed. The resulting Procrustes error quantified dissimilarity between each. Given its novelty in synergy similarity calculation, a similarity index using the best-matching scalar product of temporal patterns and basic activation coefficients normalised to the Euclidean norm was computed as in Martino et al., ( 2015 ) and can be found in the supplementary material for comparison. These yielded similar results. Furthermore, since a dissimilarity index does not indicate which individual muscles drive the observed differences, we examined the individual muscle weighting differences of spatial patterns within each synergy between limbs via linear mixed-effects modelling. This allowed us to identify how changes in specific muscles partly contributed to the dissimilarity index. Handling Kinematic turning data: Range of motion and intersegmental coordination In our previous papers on this subject (Mesquita et al., 2024 a, & Mesquita et al., under review ) we presented a method for rotating ground reaction forces from a o-x-y-z reference frame, which was fixed to the lab to a reference frame O- X- Y- z that moves as the runner turns. To achieve this, we calculated a virtual position for the centre of the pelvis in the transverse plane, P c , as the centroid of a triangle between the right and left ASIS and S2 markers. We then ran a circular fit on the movement of P c in the transverse plane to determine the radius of the curve covered by the subject over the trial and rotated the transverse plane forces and kinematics by the angle ( \(\:\theta\:\) ) formed by the o-x axis, the position of P c and the centre of a circle. Here, we rotated the kinematic marker positions based on the instantaneous \(\:{\theta\:}_{j}\) -angle. In this way, the ( x,y ) positions of each marker ( i ) in the fixed reference frame were rotated at each instant ( j ) of the stride to give a lateral and fore-aft, respectively ( X,Y ), position relative a non-inertial reference frame: $$\:\left(\begin{array}{c}{X}_{i,j}\\\:{Y}_{i,j}\end{array}\right)=\:\left(\begin{array}{c}{x}_{i,j}\text{cos}{\theta\:}_{j}-{y}_{i,j}\text{sin}{\theta\:}_{j}\\\:{x}_{i,j}\text{sin}{\theta\:}_{j}+{y}_{i,j}\text{cos}{\theta\:}_{j}\end{array}\right)$$ 4 . This rotation is possible given the angular displacement covered by the outer and inner limb are symmetric over a step (Mesquita et al., 2024 a). Note, the vertical axis was not rotated compared to body pitch as to maintain a reference perpendicular to the ground. From these rotated marker positions, the elevation angles in the rotated sagittal ( O-Y-z ) plane, i.e. , the orientation of the limb segments of both lower limbs relative to vertical were computed. This was also similar to the rotation applied in a walking while turning kinematics study (Courtine & Schieppati, 2004 ). In each speed and radius class, the average waveforms of the elevation angles of the three lower-limb segments, thigh, shank, and foot, were computed across participants. The segment range of motion (ROM) over a stride was computed as the difference between the maxima and the minima elevation angle over the period and was used to analyse the amplitude of each segment. To evaluate the time-course components of the elevation angles over one stride, a first-harmonic Fourier fit was performed (Dewolf et al., 2018 ). The amplitude ( A ), phase shift ( P ) and variance accounted for by the first harmonic were computed. The amplitude ratio ( G pd ) and phase shift (F pd ) between two adjacent limb segments p and d were computed as G pd = A d / A p and as F pd = P d – P p . The inter-segmental coordination of the lower-limb segments was analysed using principal component analysis (PCA), a method previously described in detail (Borghese et al., 1996 ; Catavitello et al., 2018 ; Ivanenko et al., 2008 ). Briefly, when the thigh, shank, and foot, 400-point time-normalised, elevation angles are plotted in three-dimensional space, which form a closed-loop trajectory that lies near a plane. The best-fitting plane is defined by the first two principal components obtained from the PCA of the covariance matrix of the three elevation angle waveforms (Daffertshofer et al., 2004 ; Ivanenko et al., 2008 ) (Fig. 3 B). The PCA returns a matrix of eigenvectors u with associated eigenvalues (λ1, λ2, λ3) ordered from the largest to the smallest. The first two eigenvectors, u 1 and u 2 generate the best fitting plane while the third eigenvector, u 3 , orthogonal to the plane, defines its orientation. The components of u 3 : u 3 t , u 3 s and u 3 f correspond to the direction cosines with the positive semi-axis of the thigh, shank and foot angular coordinates. The percentage of variance associated to vectors u 1 , u 2 , u 3 are respectively PV1, PV2 and PV3 and are computed as the ratio of each eigenvalue to the total eigenvalue sum ( e.g. , PV3 = λ3/ (λ1 + λ2 + λ3)). PV3 was used as an index of planarity, with PV3 = 0% indicating maximum planarity. Statistics For each participant, the different parameters were first averaged for all inner and outer steps over one trial. Then descriptive statistical analysis was performed. A Shapiro-Wilk test was performed to verify normality on linear data. Then, a linear mixed effect model with pairwise comparison and Bonferroni post-hoc correction was used to compute the individual effects of radius, speed and limb on the calculated variables. For circular data, i.e. , CoA, the mean resultant length ( r ) was evaluated for each variable. For data which were concentrated appropriately ( r > 0.5) a Watson-Williams test was performed with pairwise comparison and a Bonferroni post-hoc correction was used to compute factor effects. Statistical tests were run on IBM SPSS Statistics (PASW Statistic 27, NY, USA) for linear data and on Matlab for circular data. The results of the statistical tests were considered significant for a p-value < 0.05. Results Muscular activity and coordination The overall average variance accounted for, VAF, across all subjects, speeds and radii was 94.91 ± 2.2%. VAF increased at the lowest speeds and levelled off at higher speeds (F = 3.66, p = 0.004). When considering both limbs, the VAF was higher in curved compared to SLR (F = 4.77, p = 0.031). This increase was largely due to the inner limb, in which the VAF was on average 0.9% greater than the outer limb effect (F = 12.243, p < 0.001). Indeed, the outer limb was not different when compared to SLR (Bonferroni post-hoc p = 0.4). Synergies were classified according to their temporal alignment with specific mechanical events, as previously described in SLR under stable and unstable running conditions (Santuz et al., 2018 a; Santuz et al., 2020 b). Synergy 1 was associated with initial ground contact and the braking phase of stance. It was dominated by proximal hip (Gmax, Gmed) and knee extensors (VaMe, VaLe, ReFe), muscles. Synergy 2 was linked to the propulsive phase of stance, with its CoA occurring around mid-stance. It was primarily explained by the ankle plantarflexors (MeGa, LaGa, PeLo), all acting on distal joints, and producing positive work to reaccelerate the CoM. Synergy 3 corresponded to early-mid swing and was dominated by hip flexors (TeFe, Sart, ReFe) and ankle dorsiflexors (TiAn), ensuring foot clearance to prevent ‘foot-catch’ during swing. Synergy 4 was associated with late swing and preparation for the next stride, dominated by both hip flexors (BiFe, SeTe, Grac) and ankle dorsiflexors (TiAn). When observing each temporal activation pattern individually, CoA was anticipated, i.e. , its principal burst of activity arrived earlier, in curved running compared to SLR for all synergies (synergy 1: F = 32.49, synergy 2: F = 29.36, synergy 3: F = 5.16 and synergy 4: F = 44.31, all synergies : p < 0.02) (Fig. 4 A). In synergies 2 and 4, the outer limb was more anticipated than its inner counterpart (respectively, F = 6.86, 12.43, p 0.1). However, both limbs were more anticipated in synergy 1 (Bonferroni post-hoc p < 0.01) whereas in synergy 3 only the outer limb differed to SLR (Bonferroni post-hoc p < 0.05). When observing the temporal dissimilarity index (Fig. 4 B), synergies 2 and 4 showed distinct shape differences in curved running compared to SLR (synergy 2: F = 20.60, p < 0.001, synergy 4: F = 5.38, p = 0.24) while the shape difference in synergies involved in the initial touchdown and beginning of swing phase, 1 and 3, did not differ between curved and SLR (synergy 1: F = 0.17, p = 0.685, synergy 3: F = 0.29, p = 0.5913). Higuchi’s fractal dimension of the temporal patterns, Fig. 4 C, varied depending on specific muscle synergies in curved running. Synergies 1 and 4, associated with the touchdown and the late swing phase, were both less complex, i.e. , smaller HFD, in curved running compared to SLR (synergy 1: F = 34.11, synergy 4: F = 40.39, p < 0.001). Synergy 1 showed less complexity for the inner compared to the outer limb (synergy 1: F = 4.43, p = 0.037). The second synergy was more complex in curved running compared to SLR (F = 39.82, p < 0.001) and increased with speed (F = 8.59, p < 0.001). Regarding the spatial activation patterns dissimilarity index (Fig. 4 B), all synergies had greater dissimilarity in curved running compared to SLR (synergy 1: F = 18.10, synergy 2: F = 20.50, synergy 3: F = 5.70 and F = 9.49; p < 0.02) and only synergy 2 also changed with speed (F = 7.72, p < 0.001). Furthermore, since a dissimilarity index does not indicate which muscle weighting and which limb contributes to these differences, a signed difference was computed of spatial patterns per limb and per synergy. Synergy 1 showed specific increases in the outer limb BiFe and PeLo muscle contributions (p < 0.001) whereas the inner limb had increased contribution of the Grac, SeTe and Sart (p < 0.01). The second synergy showed an increase in the inner PeLo (p < 0.001) and outer MeGa and SeTe (p < 0.008). The muscle patterns tied to the initial swing phase had increased contribution for the inner Grac and TiAn (p < 0.001) and increased contribution for the outer PeLo and Sart (p < 0.001). Finally, synergy 4 showed an increased contribution to the outer TiAn (p < 0.001) and inner BiFe and SeTe (p < 0.005). Lower limb segment coordination The ROM of each lower limb segment increased with speed (thigh: F = 750.01; shank: F = 514.60; foot: F = 280.70, p < 0.001) (Fig. 5 A). The ROM of the thigh and shank increased on curve as compared to SLR (thigh: F = 44.45, p < 0.001; shank: F = 8.18, p < 0.001). When comparing limbs, the inner thigh and shank had a greater ROM than their outer counterparts (thigh: F = 19.15, shank: F = 12.62, p < 0.005). The foot ROM, however, did not change significantly with radius nor between limbs (radius F = 0.49, p = 0.49, limb: F = 2.30, p = 0.13) (Fig. 5 A, 5 B). When analysing the time-course variables, the first harmonic accounted for a large portion of the accounted variance across all speeds and radius scenarios (overall average across limbs: 90.19 ± 5.48% mean ± SD). At the level of the thigh and shank, the phase between the two segment’s first harmonic decreased with speed (F = 5.10, p < 0.001) and decreased in curved running as compared to SLR (F = 10.95, p < 0.001). The outer limb phase shift was smaller than its inner counterpart (F = 25.77, p < 0.001). The latter was not different from the shift observed in SLR (rad*limb post-hoc comparison between the inner limb in SLR and curved running p = 0.244). The thigh-shank amplitude ratio was smaller in curved running as compared to SLR (F = 9.80, p < 0.002) but did not change with speed (F = 0.516, p < 0.796) nor between limbs (F = 2.87, p < 0.09). Between the shank and foot, the phase-shift increases with speed (F = 13.37, p < 0.001), and decreases in curved running (F = 19.55, p < 0.001). The inner limb shank and foot were more out of phase than in the outer limb (limb F = 22.78, p < 0.001), which remained similar to SLR (rad*limb post-hoc comparison between the outer limb in SLR and curved running p = 0.130). The shank-foot amplitude ratio changed with both speed (F = 38.29, p < 0.001) and curvature (F = 8.13, p < 0.005). The inner limb decreased its ratio in curved running (rad*limb post-hoc for the inner limb p 98% and decreased compared to SLR (F = 6.95, p < 0.009). PV1 decreased (F = 46.03, p < 0.001) as PV2 increased (F = 26.35, p < 0.001) in curved compared to SLR. Furthermore, particularly at high speeds, the rotation of the covariation plane differed between the inner and outer limbs (Fig. 3 B). This is further shown when observing the inner limb u 3 t component, orthogonal to the plane (Fig. 6 B, left), which progressively dissociated from the outer limb in curved running (limb effect: F = 52.95, p < 0.001, radius*limb effect: F = 17.74, p < 0.001). The inner limb relied less on PV1 (limb effect: F = 33.12, p < 0.001) and had a higher percentage of variance explained by PV2 (limb effect: F = 18.70, p < 0.001) compared to the outer limb (Fig. 6 B, right). Discussion This study aimed to investigate how curved running modifies one’s muscular and segmental programming patterns compared to straight-line running. Our results showed that muscle synergies were restructured during curved running, with earlier activation timings, altered complexity, and distinct spatial and temporal recruitment patterns relative to straight-line conditions. These adaptations were also limb-specific, as the inner limb generally displayed greater adjustments in muscle weighting and reduced complexity, whereas the outer limb showed more pronounced anticipations in certain synergies. Regarding kinematic coordination, curved running modified intersegmental phase relations and planar covariation of lower limb segments, again with asymmetric contributions from the inner and outer limbs. Muscular Coordination Results from muscular synergies aligned well with our previous observations at the spinal motor output level (Mesquita et al., under review ). Similar to curved walking (Courtine & Schieppati, 2003 ), overall the muscle activation patterns did not markedly differ from those in SLR during curved running, and, thus, the general structure of the muscle synergies was also largely preserved. Nevertheless, systematic temporal and spatial modifications emerged. Across all synergies, curved running showed anticipatory shifts in the CoA, with activations occurring earlier in the cycle compared with SLR (Fig. 4 A). These anticipations were also limb-dependent in synergies 2 and 4, where the outer limb anticipated more than the inner. We hypothesised that such anticipatory behaviour may result from heightened muscle spindle sensitivity, facilitating tendomuscular systems in tolerating and absorbing greater impact loads (Gollhofer & Kyröläinen, 1991 ; Kyröläinen et al., 2005 ). This explanation remains plausible, and suggests that reactive feedback adaptations alter muscle synergies differently between limbs (Donelan & Pearson, 2004 ). In parallel, evidence from postural control research indicates that anticipatory adjustments can reorganise muscle coordination before predictable perturbations or voluntary actions, reflecting proactive feedforward reconfiguration of synergies to stabilise task performance (Klous et al., 2011 ; Krishnan et al., 2012 ). Although demonstrated in postural contexts, a similar mechanism may underlie the anticipatory shifts observed here, with the locomotor system recalibrating modular activation in advance to meet the changing mechanical demands of curved running. A second point previously observed (Mesquita et al., under review ) was a proximal-to-distal redistribution of joint work during curved running, with both limbs producing more positive work at the ankle, similar to unstable locomotion in birds and uphill running in humans (Daley et al., 2007 ; Qiao et al., 2017 ). This distal shift required compensatory energy absorption in the proximal hip joint, consistent with greater lumbar than sacral MN output. These results support segment-specific spinal specialisation, with lumbar segments driving rhythm generation and sacral segments adapting to feedback and foot–ground interactions (Dewolf et al., 2019 ; Lacquaniti et al., 2012 ; Mesquita et al., 2023 ; Minassian et al., 2017 ; Selionov et al., 2009 ). Synergy 2, active during push-off and primarily driven by the MeGa, LaGa, and PeLo, was where distal muscles produced the most positive work. It also showed greater anticipation in the outer limb, along with larger spatial and temporal dissimilarities between limbs and a wider divergence from SLR patterns (Fig. 4 B). Notably, MeGa contributed more to the outer limb, while PeLo contributed more to the inner limb throughout push-off, consistent with sacral-level adaptations associated with foot placement. Importantly, synergy 2 represents the phase most tightly coupled to reflex-driven sensory feedback, as the period starts where the limb is maximally loaded and ends where the hip flexor tendons are maximally stretched, both critical sources of afferent input (af Klint et al., 2010 ; Donelan & Pearson, 2004 ; Ekeberg & Pearson, 2005 ; Pearson, 2008 ). These features make the second synergy a key moment for task-specific reactive control, integrating load- and stretch-sensitive feedback to stabilise and adapt to challenging locomotor conditions. In curved running, the inner and outer limbs have distinct mechanical functions, requiring partly independent control to redirect the centre of mass laterally and accommodate with increased lateral forces (Mesquita et al., 2024 a). The inner limb primarily contributed to lateral redirection, with greater involvement of hip muscles such as the tensor fasciae latae and gracilis (Mesquita et al., under revision ). Neural adaptations can be asymmetrical, as shown in split-belt treadmill walking where each limb develops distinct neuromuscular responses under different mechanical demands (MacLellan et al., 2014 ; Ogawa et al., 2012 ; Yokoyama et al., 2018 ). In line with this, we observed systematic limb differences in activation patterns across all synergies (Fig. 4 B), which likely reflect a global reorganisation at the pattern formation level. By contrast, temporal shifts confined to synergies 2 and 4 suggest more targeted, sensory-driven modulation (Dzeladini et al., 2014 ), consistent with their sacral motoneuron output dominance and modifications due to environmental interactions. This interpretation fits with CPG-circuit models that show a two-layer organisation, one for rhythm generation and a second for pattern formation, where both descending inputs and sensory feedback modulate motor output (Danner et al., 2015 ; Dzeladini et al., 2014 ). Lastly, four synergies explained over 94% of variance on average from our data using NNMF. The VAF for the inner limb muscular activity was on average ~ 1% greater than those of both SLR and the outer limb muscular activity, implying that the model better explained the inner compared to the outer limb data. This suggested that the inner limb possibly operated in a more constrained and less complex manner, similar to observations of running in challenging locomotor environments (Santuz et al., 2018 a; Santuz et al., 2020 b). To better analyse this complexity, Higuchi’s fractal dimension (Higuchi, 1988 ) was applied to the temporal activation patterns and showed that curved running was generally less complex than SLR in synergies linked to touch-down and late-swing, while synergy 2 became more complex. Limb-specific complexity differences were particularly evident in synergy 1, where the outer limb showed greater HFD than its inner counterpart. Although small (< 1% VAF), these differences indicate that inner-limb synergies, especially at touchdown, were less complex than their outer counterparts. This reduction likely reflects more robust control, with signals being expressed more as singular bursts of activity with less fragmented additional bursts. Previous accounts have explained this, in part, by a widening of the synergy temporal parameters which begin to overlap and lose in robustness (Santuz et al., 2020 a). In the present case, no statistically significant effects were found for synergy width increase, neither with speed nor curvature, however, trends consistently indicated that synergies were wider in both situations. In contrast, synergy 2, showed an increase in complexity. This likely reflected the need for finer control of distal musculature to accommodate stability demands during turning. Taken together, these findings suggest that while some synergies become more robust (less complex) under curved running, others can be selectively tuned, with increased complexity emerging when their functional role is critical for task stability. Limb coordination Regarding the kinematic coordination of the lower limb segments (Fig. 3 ), the ROM of the thigh and shank of both limbs increased in curved running compared to SLR (Fig. 4 ). In terms of the orientation of the covariance plane, when pooling both limbs together, the segment coordination did not differ from SLR. This emphasizes the robustness of the planar covariance law in the sagittal plane segment angular motion (Ivanenko et al., 2008 ; Lacquaniti et al., 2012 ). However, the amount of variance explained by either changed, PC1 decreased whereas PC2 increased, similar to curved walking observations (Courtine & Schieppati, 2004 ), which shows that kinematic modifications do occur to account for the specific task. By looking at the outer and inner segmental coordination, differences in the movement programming between limbs emerged. The coordination loops (Fig. 3 ) revealed that the outer limb followed the long axis more closely, which is better explained by the first eigenvector (increased PV1), whereas the inner limb aligned more with the wide axis, better explained by the second eigenvector (increased PV2). This suggested that the outer limb’s movement was primarily governed by orientation changes while the inner limb’s motion was more influenced by variations in limb length (Ivanenko et al., 2008 ), indicating distinct strategies in limbs for managing the curved running motion. This was consistent with observations that limb function affects the plane orientation (Catavitello et al., 2018 ). This can be further observed in the phase differences between segments and their relationship to an overall coordination. Previously, Courtine & Schieppati ( 2004 ) showed that in both linear and curved walking the thigh-shank phase relation was strongly related to loop shape, represented by u 1 t . Here, when we regress the F ts onto u 1 t we see distinct limb-specific patterns (Fig S2), where the slope formed by the outer limb relationship was similar to SLR, albeit with less variance explained (R 2 = 0.27), whereas the inner limb had explained variance higher (R 2 = 0.36) but with a different slope. This indicates that the inner limb’s thigh–shank timing better explained the loop’s main axis, i.e. , limb orientation variations and that the outer limb kept a similar behaviour to SLR (see supplementary material). In contrast, when regressing the thigh-shank phase difference onto u 2 t , the projection of PC2 (the loop width), the inner limb had a stronger relationship (R 2 = 0.67 for the inner limb, vs. 0.44 in SLR) with a steeper slope, while the outer limb loses nearly all association (R 2 = 0.14). These results suggest that during turning, the thigh – shank phase shifts are increasingly expressed along the orthogonal axis for the inner limb, reflecting in-plane adjustments of loop geometry, whereas the outer limb preserves its primary phase–orientation relationship. In the discussion on curved walking, Courtine & Schiepatti (2004) proposed that descending commands modulate spinal oscillators to adjust phase relationships between segments without disrupting the basic locomotor rhythm. They further argued, based on EMG differences (Courtine & Schieppati, 2003 ), that corticospinal drive enables inner–outer limb adjustments without interfering with rhythm or gait organisation. Two decades later, advances in muscle activity modelling using NNMF, amongst others, provide necessary tools to try and explain the modular organisation of muscle synergies more directly. Our results confirm these observations: all synergies showed systematically modified spatial patterns, while temporal patterns remained more robustly conserved during curved running. Adaptations to the latter were subtler but evident in amplitude and timing, likely reflecting both feedforward and sensory feedback adjustments at critical points in the step cycle. These findings provide quantitative support for the view that locomotor control integrates robust spinal rhythm generators with flexible, limb-specific modulation to accommodate for situations such as changes in trajectory. Limitations A limitation of this study is that the SLR condition was done on a treadmill and not overground. Thus, despite most outcomes between treadmill and overground running being largely comparable (Van Hooren et al., 2020 ), it does add some bias to our results. Declarations Data is available under https://doi.org/10.17605/OSF.IO/UGJDC Code can be made available upon reasonable request. Author Contributions • Conceptualization: [Raphael Mesquita, Patrick Willems Arthur Dewolf]; Methodology: [Raphael Mesquita, Arthur Dewolf]; Formal analysis and investigation: [Raphael Mesquita, Arthur Dewolf]; Writing - original draft preparation: [Raphael Mesquita, Thibaut Toussaint, Patrick Willems, Arthur Dewolf]; Writing - review and editing[Raphael Mesquita, Thibaut Toussaint, Patrick Willems, Arthur Dewolf]; Funding acquisition: [Patrick Willems]; Supervision: [Arthur Dewolf] Funding This study was funded by the Fonds de la Recherche Scientifique (F.N.R.S - CDR 40013847). The authors have no relevant financial or non-financial interests to disclose. Informed consent was obtained from all individual participants included in the study. References af Klint R, Mazzaro N, Nielsen JB, Sinkjaer T, Grey MJ (2010) Load Rather Than Length Sensitive Feedback Contributes to Soleus Muscle Activity During Human Treadmill Walking. J Neurophysiol 103(5):2747–2756. https://doi.org/10.1152/jn.00547.2009 Alexander RM (2002) Stability and Manoeuvrability of Terrestrial Vertebrates. Integr Comp Biol 42(1):158–164. https://doi.org/10.1093/icb/42.1.158 Andreella A, De Santis R, Vesely A, Finos L (2023) Procrustes-based distances for exploring between-matrices similarity. Stat Methods Appl 32(3):867–882. https://doi.org/10.1007/s10260-023-00689-y Bianchi L, Angelini D, Orani GP, Lacquaniti F (1998) Kinematic Coordination in Human Gait: Relation to Mechanical Energy Cost. J Neurophysiol 79(4):2155–2170. https://doi.org/10.1152/jn.1998.79.4.2155 Borghese NA, Bianchi L, Lacquaniti F (1996) Kinematic determinants of human locomotion. J Physiol 494(3):863–879. https://doi.org/10.1113/jphysiol.1996.sp021539 Cappellini G, Ivanenko YP, Poppele RE, Lacquaniti F (2006) Motor Patterns in Human Walking and Running. J Neurophysiol 95(6):3426–3437. https://doi.org/10.1152/jn.00081.2006 Catavitello G, Ivanenko Y, Lacquaniti F (2018) A kinematic synergy for terrestrial locomotion shared by mammals and birds. eLife 7:e38190. https://doi.org/10.7554/eLife.38190 Cheung VCK, Cheung BMF, Zhang JH, Chan ZYS, Ha SCW, Chen C-Y, Cheung RTH (2020) Plasticity of muscle synergies through fractionation and merging during development and training of human runners. Nat Commun 11(1):4356. https://doi.org/10.1038/s41467-020-18210-4 Cheung VCK, d’Avella A, Tresch MC, Bizzi E (2005) Central and Sensory Contributions to the Activation and Organization of Muscle Synergies during Natural Motor Behaviors. J Neurosci 25(27):6419–6434. https://doi.org/10.1523/JNEUROSCI.4904-04.2005 Cheung VCK, Piron L, Agostini M, Silvoni S, Turolla A, Bizzi E (2009) Stability of muscle synergies for voluntary actions after cortical stroke in humans. Proceedings of the National Academy of Sciences , 106 (46), 19563–19568. https://doi.org/10.1073/pnas.0910114106 Chvatal SA, Ting LH (2013) Common muscle synergies for balance and walking. Frontiers in Computational Neuroscience , 7 . https://doi.org/10.3389/fncom.2013.00048 Courtine G, Schieppati M (2003) Human walking along a curved path. II. Gait features and EMG patterns. Eur J Neurosci 18(1):191–205. https://doi.org/10.1046/j.1460-9568.2003.02737.x Courtine G, Schieppati M (2004) Tuning of a Basic Coordination Pattern Constructs Straight-Ahead and Curved Walking in Humans. J Neurophysiol 91(4):1524–1535. https://doi.org/10.1152/jn.00817.2003 d’Avella A, Saltiel P, Bizzi E (2003) Combinations of muscle synergies in the construction of a natural motor behavior. Nat Neurosci 6(3):300–308. https://doi.org/10.1038/nn1010 Daffertshofer A, Lamoth CJC, Meijer OG, Beek PJ (2004) PCA in studying coordination and variability: A tutorial. Clin Biomech Elsevier Ltd 19(4):415–428. https://doi.org/10.1016/j.clinbiomech.2004.01.005 Daley MA, Felix G, Biewener AA (2007) Running stability is enhanced by a proximo-distal gradient in joint neuromechanical control. J Exp Biol Danner SM, Hofstoetter US, Freundl B, Binder H, Mayr W, Rattay F, Minassian K (2015) Human spinal locomotor control is based on flexibly organized burst generators. Brain 138(3):577–588. https://doi.org/10.1093/brain/awu372 Dewolf AH, Ivanenko YP, Zelik KE, Lacquaniti F, Willems PA (2019) Differential activation of lumbar and sacral motor pools during walking at different speeds and slopes. J Neurophysiol 122(2):872–887. https://doi.org/10.1152/jn.00167.2019 Dewolf AH, Ivanenko Y, Zelik KE, Lacquaniti F, Willems PA (2018) Kinematic patterns while walking on a slope at different speeds. J Appl Physiol 125(2):642–653. https://doi.org/10.1152/japplphysiol.01020.2017 Dewolf AH, La Scaleia V, Fabiano A, Sylos-Labini F, Mondi V, Picone S, Di Paolo A, Paolillo P, Ivanenko Y, Lacquaniti F (2022) Left–Right Locomotor Coordination in Human Neonates. J Neurosci 42(34):6566–6580. https://doi.org/10.1523/JNEUROSCI.0612-22.2022 Dewolf AH, Mesquita RM, Willems PA (2020) Intra-limb and muscular coordination during walking on slopes. Eur J Appl Physiol 120(8):1841–1854. https://doi.org/10.1007/s00421-020-04415-4 Dewolf AH, Sylos-Labini F, Cappellini G, Lacquaniti F, Ivanenko Y (2020) Emergence of Different Gaits in Infancy: Relationship Between Developing Neural Circuitries and Changing Biomechanics. Front Bioeng Biotechnol 8:473. https://doi.org/10.3389/fbioe.2020.00473 Dominici N, Ivanenko YP, Cappellini G, d’Avella A, Mondì V, Cicchese M, Fabiano A, Silei T, Di Paolo A, Giannini C, Poppele RE, Lacquaniti F (2011) Locomotor Primitives in Newborn Babies and Their Development. Science 334(6058):997–999. https://doi.org/10.1126/science.1210617 Donelan JM, Pearson KG (2004) Contribution of Force Feedback to Ankle Extensor Activity in Decerebrate Walking Cats. J Neurophysiol 92(4):2093–2104. https://doi.org/10.1152/jn.00325.2004 Dzeladini F, Van Den Kieboom J, Ijspeert A (2014) The contribution of a central pattern generator in a reflex-based neuromuscular model. Frontiers in Human Neuroscience , 8 . https://doi.org/10.3389/fnhum.2014.00371 Ekeberg Ö, Pearson K (2005) Computer Simulation of Stepping in the Hind Legs of the Cat: An Examination of Mechanisms Regulating the Stance-to-Swing Transition. J Neurophysiol 94(6):4256–4268. https://doi.org/10.1152/jn.00065.2005 Fujiki S, Aoi S, Funato T, Sato Y, Tsuchiya K, Yanagihara D (2018) Adaptive hindlimb split-belt treadmill walking in rats by controlling basic muscle activation patterns via phase resetting. Sci Rep 8(1):17341. https://doi.org/10.1038/s41598-018-35714-8 Gneiting T, Schlather M (2004) Stochastic Models That Separate Fractal Dimension and the Hurst Effect. SIAM Rev 46(2):269–282. https://doi.org/10.1137/S0036144501394387 Gollhofer A, Kyröläinen H (1991) Neuromuscular Control of the Human Leg Extensor Muscles in Jump Exercises Under Various Stretch-Load Conditions. Int J Sports Med 12(01):34–40. https://doi.org/10.1055/s-2007-1024652 Grillner S (1985) Neurobiological Bases of Rhythmic Motor Acts in Vertebrates. Science 228(4696):143–149. https://doi.org/10.1126/science.3975635 Grillner S, Manira E, A (2020) Current Principles of Motor Control, with Special Reference to Vertebrate Locomotion. Physiol Rev 100(1):271–320. https://doi.org/10.1152/physrev.00015.2019 Higuchi T (1988) Approach to an irregular time series on the basis of the fractal theory. Physica D 31(2):277–283. https://doi.org/10.1016/0167-2789(88)90081-4 Ijspeert AJ, Daley MA (2023) Integration of feedforward and feedback control in the neuromechanics of vertebrate locomotion: A review of experimental, simulation and robotic studies. J Exp Biol 226(15):jeb245784. https://doi.org/10.1242/jeb.245784 Ivanenko YP (2005) Coordination of Locomotion with Voluntary Movements in Humans. J Neurosci 25(31):7238–7253. https://doi.org/10.1523/JNEUROSCI.1327-05.2005 Ivanenko YP, Cappellini G, Dominici N, Poppele RE, Lacquaniti F (2007) Modular Control of Limb Movements during Human Locomotion. J Neurosci 27(41):11149–11161. https://doi.org/10.1523/JNEUROSCI.2644-07.2007 Ivanenko YP, d’Avella A, Poppele RE, Lacquaniti F (2008) On the Origin of Planar Covariation of Elevation Angles During Human Locomotion. J Neurophysiol 99(4):1890–1898. https://doi.org/10.1152/jn.01308.2007 Ivanenko YP, Poppele RE, Lacquaniti F (2004) Five basic muscle activation patterns account for muscle activity during human locomotion: Basic muscle activation patterns. J Physiol 556(1):267–282. https://doi.org/10.1113/jphysiol.2003.057174 Ivanenko YP, Poppele RE, Lacquaniti F (2006a) Motor Control Programs and Walking. Neuroscientist 12(4):339–348. https://doi.org/10.1177/1073858406287987 Ivanenko YP, Poppele RE, Lacquaniti F (2006b) Spinal Cord Maps of Spatiotemporal Alpha-Motoneuron Activation in Humans Walking at Different Speeds. J Neurophysiol 95(2):602–618. https://doi.org/10.1152/jn.00767.2005 Kesić S, Spasić SZ (2016) Application of Higuchi’s fractal dimension from basic to clinical neurophysiology: A review. Comput Methods Programs Biomed 133:55–70. https://doi.org/10.1016/j.cmpb.2016.05.014 Klous M, Mikulic P, Latash ML (2011) Two aspects of feedforward postural control: Anticipatory postural adjustments and anticipatory synergy adjustments. J Neurophysiol 105(5):2275–2288. https://doi.org/10.1152/jn.00665.2010 Krishnan V, Latash ML, Aruin AS (2012) Early and late components of feed-forward postural adjustments to predictable perturbations. Clin Neurophysiol 123(5):1016–1026. https://doi.org/10.1016/j.clinph.2011.09.014 Kyröläinen H, Avela J, Komi PV (2005) Changes in muscle activity with increasing running speed. J Sports Sci 23(10):1101–1109. https://doi.org/10.1080/02640410400021575 La Scaleia V, Ivanenko YP, Zelik KE, Lacquaniti F (2014) Spinal motor outputs during step-to-step transitions of diverse human gaits. Frontiers in Human Neuroscience , 8 . https://doi.org/10.3389/fnhum.2014.00305 Lacquaniti F (2002) Kinematic Control of Walking. Arch Ital Biol 140:263–272 Lacquaniti F, Grasso R, Zago M (1999) Motor Patterns in Walking. Physiology 14(4):168–174. https://doi.org/10.1152/physiologyonline.1999.14.4.168 Lacquaniti F, Ivanenko YP, Zago M (2012) Patterned control of human locomotion: Control of human locomotion. J Physiol 590(10):2189–2199. https://doi.org/10.1113/jphysiol.2011.215137 Lee DD, Seung HS (1999) Learning the parts of objects by non-negative matrix factorization. Nature 401(6755):788–791. https://doi.org/10.1038/44565 Linssen WHJP, Stegeman DF, Joosten EMG, Notermans SLH, Van ’T, Hof MA, Binkhorst RA (1993) Variability and interrelationships of surface EMG parameters during local muscle fatigue. Muscle Nerve 16(8):849–856. https://doi.org/10.1002/mus.880160808 MacLellan MJ, Ivanenko YP, Massaad F, Bruijn SM, Duysens J, Lacquaniti F (2014) Muscle activation patterns are bilaterally linked during split-belt treadmill walking in humans. J Neurophysiol 111(8):1541–1552. https://doi.org/10.1152/jn.00437.2013 Martino G, Ivanenko YP, d’Avella A, Serrao M, Ranavolo A, Draicchio F, Cappellini G, Casali C, Lacquaniti F (2015) Neuromuscular adjustments of gait associated with unstable conditions. J Neurophysiol 114(5):2867–2882. https://doi.org/10.1152/jn.00029.2015 Martino G, Ivanenko YP, Serrao M, Ranavolo A, d’Avella A, Draicchio F, Conte C, Casali C, Lacquaniti F (2014) Locomotor patterns in cerebellar ataxia. J Neurophysiol 112(11):2810–2821. https://doi.org/10.1152/jn.00275.2014 Mesquita RM, Catavitello G, Willems PA, Dewolf AH (2023) Modification of the locomotor pattern when deviating from the characteristic heel-to-toe rolling pattern during walking. Eur J Appl Physiol. https://doi.org/10.1007/s00421-023-05169-5 Mesquita RM, Willems PA, Catavitello G, Dewolf AH (2024) Kinematics and mechanical changes with step frequency at different running speeds. Eur J Appl Physiol 124(2):607–622. https://doi.org/10.1007/s00421-023-05303-3 Mesquita RM, Willems PA, Dewolf AH, Catavitello G (2024) Kinetics and mechanical work done to move the body centre of mass along a curve. PLoS ONE 19(2):21 Minassian K, Hofstoetter US, Dzeladini F, Guertin PA, Ijspeert A (2017) The Human Central Pattern Generator for Locomotion: Does It Exist and Contribute to Walking? Neuroscientist 23(6):649–663. https://doi.org/10.1177/1073858417699790 Nishikawa K, Biewener AA, Aerts P, Ahn AN, Chiel HJ, Daley MA, Daniel TL, Full RJ, Hale ME, Hedrick TL, Lappin AK, Nichols TR, Quinn RD, Satterlie RA, Szymik B (2007) Neuromechanics: An integrative approach for understanding motor control. Integr Comp Biol 47(1):16–54. https://doi.org/10.1093/icb/icm024 Ogawa T, Kawashima N, Suzuki S, Nakazawa K (2012) Asymmetrical Neural Adaptation in Lower Leg Muscles as a Consequence of Stereotypical Motor Training. J Mot Behav 44(1):63–68. https://doi.org/10.1080/00222895.2011.645912 Pearson KG (2008) Role of sensory feedback in the control of stance duration in walking cats. Brain Res Rev 57(1):222–227. https://doi.org/10.1016/j.brainresrev.2007.06.014 Poppele R, Bosco G (2003) Sophisticated spinal contributions to motor control. Trends Neurosci 26(5):269–276. https://doi.org/10.1016/S0166-2236(03)00073-0 Qiao M, Abbas JJ, Jindrich DL (2017) A model for differential leg joint function during human running. Bioinspir Biomim 12(1):016015. https://doi.org/10.1088/1748-3190/aa50b0 Santuz A, Brüll L, Ekizos A, Schroll A, Eckardt N, Kibele A, Schwenk M, Arampatzis A (2020) Neuromotor Dynamics of Human Locomotion in Challenging Settings. iScience 23(1):100796. https://doi.org/10.1016/j.isci.2019.100796 Santuz A, Ekizos A, Eckardt N, Kibele A, Arampatzis A (2018) Challenging human locomotion: Stability and modular organisation in unsteady conditions. Sci Rep 8(1):2740. https://doi.org/10.1038/s41598-018-21018-4 Santuz A, Ekizos A, Janshen L, Mersmann F, Bohm S, Baltzopoulos V, Arampatzis A (2018) Modular Control of Human Movement During Running: An Open Access Data Set. Front Physiol 9:1509. https://doi.org/10.3389/fphys.2018.01509 Santuz A, Ekizos A, Kunimasa Y, Kijima K, Ishikawa M, Arampatzis A (2020) Lower complexity of motor primitives ensures robust control of high-speed human locomotion. Heliyon 6(10):e05377. https://doi.org/10.1016/j.heliyon.2020.e05377 Sartori M, Gizzi L, Lloyd DG, Farina D (2013) A musculoskeletal model of human locomotion driven by a low dimensional set of impulsive excitation primitives. Frontiers in Computational Neuroscience , 7 . https://doi.org/10.3389/fncom.2013.00079 Selionov VA, Ivanenko YP, Solopova IA, Gurfinkel VS (2009) Tonic Central and Sensory Stimuli Facilitate Involuntary Air-Stepping in Humans. J Neurophysiol 101(6):2847–2858. https://doi.org/10.1152/jn.90895.2008 Sylos-Labini F, La Scaleia V, Cappellini G, Dewolf A, Fabiano A, Solopova IA, Mondì V, Ivanenko Y, Lacquaniti F (2022) Complexity of modular neuromuscular control increases and variability decreases during human locomotor development. Commun Biology 5(1):1256. https://doi.org/10.1038/s42003-022-04225-8 Torres-Oviedo G, Macpherson JM, Ting LH (2006) Muscle Synergy Organization Is Robust Across a Variety of Postural Perturbations. J Neurophysiol 96(3):1530–1546. https://doi.org/10.1152/jn.00810.2005 Van Hooren B, Fuller JT, Buckley JD, Miller JR, Sewell K, Rao G, Barton C, Bishop C, Willy RW (2020) Is Motorized Treadmill Running Biomechanically Comparable to Overground Running? A Systematic Review and Meta-Analysis of Cross-Over Studies. Sports Med 50(4):785–813. https://doi.org/10.1007/s40279-019-01237-z Winter DA (2009) Biomechanics and motor control of human movement, 4th edn. Wiley Yokoyama H, Ogawa T, Kawashima N, Shinya M, Nakazawa K (2016) Distinct sets of locomotor modules control the speed and modes of human locomotion. Sci Rep 6(1):36275. https://doi.org/10.1038/srep36275 Yokoyama H, Sato K, Ogawa T, Yamamoto S-I, Nakazawa K, Kawashima N (2018) Characteristics of the gait adaptation process due to split-belt treadmill walking under a wide range of right-left speed ratios in humans. PLoS ONE 13(4):e0194875. https://doi.org/10.1371/journal.pone.0194875 Zaidi KF, Harris-Love M (2023) A Novel Procrustes Analysis Method to Quantify Multi-Joint Coordination of the Upper Extremity after Stroke. 2023 45th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC) , 1–4. https://doi.org/10.1109/EMBC40787.2023.10341023 Additional Declarations No competing interests reported. Supplementary Files SupplementarymaterialMesquita.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8038490","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":543431855,"identity":"42c7afdf-4ca3-4bbe-b956-c1595b62b57a","order_by":0,"name":"R. M. Mesquita","email":"data:image/png;base64,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","orcid":"","institution":"Université Catholique de Louvain","correspondingAuthor":true,"prefix":"","firstName":"R.","middleName":"M.","lastName":"Mesquita","suffix":""},{"id":543431856,"identity":"d2906801-e2ed-4515-8a52-446af3c3e1dd","order_by":1,"name":"T. D. Toussaint","email":"","orcid":"","institution":"Université Catholique de Louvain","correspondingAuthor":false,"prefix":"","firstName":"T.","middleName":"D.","lastName":"Toussaint","suffix":""},{"id":543431857,"identity":"7b93bc78-1e88-448f-b0ed-453415239f4e","order_by":2,"name":"P. A. Willems","email":"","orcid":"","institution":"Université Catholique de Louvain","correspondingAuthor":false,"prefix":"","firstName":"P.","middleName":"A.","lastName":"Willems","suffix":""},{"id":543431858,"identity":"8359f932-1cfd-475c-9e4a-2332603d9b0f","order_by":3,"name":"A. H. Dewolf","email":"","orcid":"","institution":"Université Catholique de Louvain","correspondingAuthor":false,"prefix":"","firstName":"A.","middleName":"H.","lastName":"Dewolf","suffix":""}],"badges":[],"createdAt":"2025-11-05 12:38:36","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8038490/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8038490/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":99862525,"identity":"d9e16046-3537-459e-87a6-be54acc4dca1","added_by":"auto","created_at":"2026-01-09 07:18:46","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":72992,"visible":true,"origin":"","legend":"\u003cp\u003eTime normalised typical trace of raw (unfiltered) electromyographic (EMG) rectified data of the 14 bilateral lower limb muscles (red = inner limb, blue = outer limb) when running on a 6\u0026nbsp;m radius of curvature at both 13 and 17\u0026nbsp;km\u0026nbsp;h\u003csup\u003e-\u0026nbsp;1\u003c/sup\u003e. Superimposed on each figure, are the normalised linear envelopes of each curve (see Methods for details). Traces begin at foot contact. Subjects were divided into two EMG groups, to expand the number of muscles recorded, and ran with 12 sensors per limb. Thus, the TeFe and VaMe are from a subject in the other protocol.\u003c/p\u003e","description":"","filename":"TurningcoordinationfiguresMesquita1.png","url":"https://assets-eu.researchsquare.com/files/rs-8038490/v1/4b9e47a9c47290930de2fe51.png"},{"id":100357817,"identity":"b3df00b9-3a0d-44a0-8538-96ddf06117c2","added_by":"auto","created_at":"2026-01-16 07:20:23","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":81027,"visible":true,"origin":"","legend":"\u003cp\u003eMuscular synergy time-normalised temporal (upper continuous lines) and spatial basic activation (bar graph) patterns presented during straight and curved running at 13 and 17\u0026nbsp;km\u0026nbsp;h\u003csup\u003e-1\u003c/sup\u003e. Each synergy is represented in numerical order from left to right. Synergy 1 appears at foot contact, synergy 2 corresponds to push-off phase, synergy 3 corresponds to the early-mid swing phase and the fourth synergy represents the late swing phase. Temporal patterns are extracted from muscular activity which is normalised to the maximal activity for each muscle during SLR. Each basic activation pattern represents the means ±SD contributions of muscles and was normalised such that the sum of its weights equals 1, allowing the pattern to be interpreted as the relative contribution of each muscle to the synergy, independent of overall amplitude. For both temporal and spatial patterns, the red represents the inner and blue the outer limb in curved running. The black dashed line and black bar plots represent straight line running.\u003c/p\u003e","description":"","filename":"TurningcoordinationfiguresMesquita2.png","url":"https://assets-eu.researchsquare.com/files/rs-8038490/v1/cccc3410bc63fc58b7e73edb.png"},{"id":99862526,"identity":"7ca1221a-55ed-4fdd-843a-d90908354d27","added_by":"auto","created_at":"2026-01-09 07:18:46","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":169124,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA)\u003c/strong\u003e Elevation angles of the thigh, shank, and foot for the inner and outer limbs over a step at 17\u0026nbsp;km\u0026nbsp;h\u003csup\u003e-\u0026nbsp;1\u003c/sup\u003e in SLR\u003cstrong\u003e \u003c/strong\u003eand curved running. Traces begin at foot contact. 3B) 3D-plots of the elevation angles of Figure 3A, along with the best-fitting planes for the inner and outer limbs. In SLR (black lines and cubes), only the right limb is presented. Colours are the same as in previous figures.\u003c/p\u003e","description":"","filename":"TurningcoordinationfiguresMesquita3.png","url":"https://assets-eu.researchsquare.com/files/rs-8038490/v1/ea41d000255e692a88500c1d.png"},{"id":99862529,"identity":"66d2f791-ebf3-4e15-88f6-974727e6debe","added_by":"auto","created_at":"2026-01-09 07:18:46","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":68140,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA) \u003c/strong\u003eCentre of activity (CoA) of temporal activation patterns for each muscle synergy as a function of speed for the inner limb (red) and outer limb (blue) during curved running, and the right limb (black) during straight-line running (SLR). The x-axis is presented in polar coordinates (0°–360°), representing relative time over a stride, where 0° corresponds to the stride onset for both limbs. Squares indicate the circular mean ± circular standard deviation across participants. Lines represent a non-linear second-order fit applied to all data points (performed in GraphPad Prism 10). Symbols indicate significant effects: # = effect of radius; $ = effect of speed; * = effect of limb (p \u0026lt; 0.05). \u003cstrong\u003e4B) \u003c/strong\u003eProcrustes dissimilarity index (mean ± SD) as a function of speed, comparing outer and inner limbs during curved running and left vs. right limbs during straight-line running. Curved running data are shown in colour, SLR in black. \u003cstrong\u003e4C)\u003c/strong\u003e Higuchi’s fractal dimension (mean ± SD) as a function of speed, for inner and outer limbs during curved running, and right limb during SLR. Symbols, lines, colour coding and notations are as in panel A.\u003c/p\u003e","description":"","filename":"TurningcoordinationfiguresMesquita4.png","url":"https://assets-eu.researchsquare.com/files/rs-8038490/v1/2b06b41cae91dc1ced259c67.png"},{"id":99862530,"identity":"4cfca931-0908-4945-a867-284df3af83cb","added_by":"auto","created_at":"2026-01-09 07:18:46","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":65477,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA) \u003c/strong\u003eRange of angular motion of the thigh, shank and foot segments. All other indications are as in figure 4A. \u003cstrong\u003eB) \u003c/strong\u003eThe projections of each 3D-loop in figure 3B on the thigh-shank (left) and thigh-foot (right) planes over a stride at 13 and 17\u0026nbsp;km\u0026nbsp;h\u003csup\u003e-1\u003c/sup\u003e.\u003c/p\u003e","description":"","filename":"TurningcoordinationfiguresMesquita5.png","url":"https://assets-eu.researchsquare.com/files/rs-8038490/v1/dadcaf1b662983dcb8ed7cea.png"},{"id":100358396,"identity":"72edd904-b092-4fe7-9898-97820b2a44c0","added_by":"auto","created_at":"2026-01-16 07:21:00","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":44720,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA) \u003c/strong\u003ePhase differences (F) and amplitude ratios (G) of the thigh-shank segments, top two figures, and shank-foot segments, bottom two figures. All other indications are as in figure 4A. \u003cstrong\u003e6B) \u003c/strong\u003eleft side: direction cosine for the thigh of the third eigenvector \u003cem\u003eu\u003c/em\u003e\u003csub\u003e3\u003c/sub\u003e\u003cem\u003et \u003c/em\u003eaverages and standard deviations in curved running as a function of speeds. Other indications are as in figure 4A. Right side: percentage of variance (average and standard deviations) explained by the first eigenvector (PV1) and percentage of variance explained by the first and second eigenvector (PV1+2) for inner (red) and outer (blue) limbs in curved running as a function of speeds. For SLR the right limb is presented.\u003c/p\u003e","description":"","filename":"TurningcoordinationfiguresMesquita6.png","url":"https://assets-eu.researchsquare.com/files/rs-8038490/v1/0baa585070025792108d3e68.png"},{"id":100377270,"identity":"d05ab012-e92a-4ede-9289-6828a2959566","added_by":"auto","created_at":"2026-01-16 08:47:38","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1291116,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8038490/v1/f291d2d2-1920-4077-a038-72c88fdb6d2d.pdf"},{"id":99862531,"identity":"96bfd47a-399c-4db5-8b63-1821a8463a7d","added_by":"auto","created_at":"2026-01-09 07:18:46","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":237912,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementarymaterialMesquita.docx","url":"https://assets-eu.researchsquare.com/files/rs-8038490/v1/80bda44963a2ac9d22acc57b.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Limb-Specific Modulation of Muscle Synergies and Segmental Coordination During Curved Running","fulltext":[{"header":"Introduction","content":"\u003cp\u003eLocomotion arises from complex, multi-layered interactions between central neural networks, reflex pathways and the musculoskeletal system, which interacts with the environment to produce effective movement (Nishikawa et al., \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2007\u003c/span\u003e). The spinal cord plays a key role in integrating both feedforward, predictive signals, with feedback from sensory pathways to regulate locomotion (Ijspeert \u0026amp; Daley, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Evidence from both animal and human studies show that dedicated spinal circuits, \u003cem\u003ei.e.\u003c/em\u003e, central pattern generators (CPGs), organise locomotor activity (Grillner, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e1985\u003c/span\u003e; Ijspeert \u0026amp; Daley, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Minassian et al., \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). These networks of spinal interneurons, generate basic rhythmic and patterned motor outputs which are modulated by a supraspinal neural \u0026ldquo;drive\u0026rdquo; to produce coordinated locomotion in different situations, reducing the computational demands of higher brain centres (Ivanenko et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2006a\u003c/span\u003e; Lacquaniti et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Minassian et al., \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). This feedforward command is strongly coupled with limb sensory feedback (Dzeladini et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Grillner \u0026amp; El Manira, 2020), which is essential for maintaining the timing and adaptability of CPGs and coordination between the limbs, particularly in unstable conditions and/or when limbs act asymmetrically (Fujiki et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Pearson, \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2008\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn humans, indirect evidence suggests that CPGs are distributed across lumbar and sacral segments, with partially autonomous circuits dedicated to each limb (Danner et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Dewolf et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; MacLellan et al., \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). This limb-specific organisation allows for a flexible coordination of left\u0026ndash;right activity and more robust adaptation to changing biomechanical demands. Behavioural goals dictate movement, and one must adapt to diverse spatial constraints to generate coordinated motion to match the required task such as walking and running, avoiding obstacles or changing directions (Alexander, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2002\u003c/span\u003e). In nature, situations where one moves only in a straight line are probably rather infrequent and, as such, adaptions due to changes terrain and direction require modulation of both the spinal motor output and structured limb coordination to adapt to navigational constraints (Courtine \u0026amp; Schieppati, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2004\u003c/span\u003e; Dewolf et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Dewolf et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2020\u003c/span\u003ea; Santuz et al., \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2018\u003c/span\u003ea).\u003c/p\u003e \u003cp\u003eIn a recent group of studies, we investigated how humans adapt when running on curves at different speeds (Mesquita et al., \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2024\u003c/span\u003ea, Mesquita et al., \u003cem\u003eunder review\u003c/em\u003e). Our findings showed that the mechanical work required to sustain centre of mass (CoM) motion during turning increased by ~\u0026thinsp;25% at the highest speeds, largely due to the additional work needed to manage lateral displacements. Furthermore, although the overall CoM work was balanced across limbs, functional asymmetries emerged: the inner limb contributed more to redirecting the CoM inward, while the outer limb primarily supported forward progression.\u003c/p\u003e \u003cp\u003eTo understand how these asymmetries were managed, we analysed joint kinetics in both sagittal and frontal planes (Mesquita et al., \u003cem\u003eunder review\u003c/em\u003e). We observed that the work production shifted from proximal to distal joints, consistent, among others, with findings on unstable locomotion in birds (Daley et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2007\u003c/span\u003e). When grouping EMG data to observe the motoneuron (MN) activity at the spinal output level, using \u0026ldquo;spinal mapping\u0026rdquo; (Ivanenko et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2006b\u003c/span\u003e), we observed a stronger increase in lumbar than sacral motoneuron activity. This matched inverse dynamics findings at the hip and suggested that lumbar networks play a dominant role in curved running, while sacral networks adapt more subtly to foot placement.\u003c/p\u003e \u003cp\u003eTogether, these kinematic, muscular, and spinal-level findings raise the question of how modular control of muscles, and the MN output itself, is shaped by both feedforward (CPG-driven) and asymmetric sensory feedback signals during curved running.\u003c/p\u003e \u003cp\u003eUnsupervised machine learning techniques, such as non-negative matrix factorisation, applied to EMG data can capture the locomotor activity of large groups of muscles, thereby illustrating the integration between CPG-based patterns, sensory-driven MN adaption and supraspinal inputs (Ijspeert \u0026amp; Daley, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Ivanenko et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2004\u003c/span\u003e). This approach allows one to investigate modularity and signal complexity of the MN output across spatial and temporal dimensions (Santuz et al., \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2020\u003c/span\u003ea; Santuz et al., \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2018\u003c/span\u003ea; Sylos-Labini et al., \u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Building on the asymmetrical observations at the spinal output level in curved running, we expected muscular synergies themselves would follow similar modifications compared to straight line running and that inner and outer limb synergies would become asymmetric. Such modifications would indicate adjustments of the locomotor command, integrating both feedforward and feedback adaptions, and provide further support for the limb-specific organisation of MN output (Danner et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Dzeladini et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; MacLellan et al., \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Yokoyama et al., \u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eTo further understand these adaptions, we also investigated the \u0026lsquo;global\u0026rsquo; kinematic coordination of the limb segments (the thigh, shank and foot motion), \u003cem\u003evia\u003c/em\u003e inter-segmental coordination (Ivanenko et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). Various studies have shown that limb kinematics are similarly contained within the spinal circuitry and also originates from a feedforward central control loop coupled with peripheral sensory feedback to achieve the desired kinematic task (Lacquaniti, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2002\u003c/span\u003e; Lacquaniti et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e1999\u003c/span\u003e; Poppele \u0026amp; Bosco, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2003\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIndeed, the motion of the elevation angles in the lower limb segments covary along a plane, effectively simplifying the degrees of freedom of the locomotor system to two principle components : limb length and orientation (Bianchi et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e1998\u003c/span\u003e; Ivanenko et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; Poppele \u0026amp; Bosco, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2003\u003c/span\u003e). The orientation of the covariation plane, along with the characteristics of the loop, defines a general kinematic pattern of locomotion. This pattern adapts to different locomotor conditions (Dewolf et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), adjusts with changes in speed (Ivanenko et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Mesquita et al., \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2024\u003c/span\u003eb) and mechanical demands (Dewolf et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Dewolf et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2020\u003c/span\u003ea). Considering that plane orientation is limb function specific (Catavitello et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), when limbs have different functions we hypothesised that the plane describing the motion of the inner limb would diverge from that of the outer limb in curved running. We also expected that the hypothesised limb-specific shifts in muscle synergies to be mirrored by corresponding adjustments in intersegmental coordination changes.\u003c/p\u003e"},{"header":"Materials \u0026 Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n\u003ch2\u003eParticipants and experimental procedure\u003c/h2\u003e\n\u003cp\u003eThe present experimental protocol description was adapted from Mesquita et al. (\u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003ea). Eleven recreational male runners (height: 1.81\u0026thinsp;\u0026plusmn;\u0026thinsp;0.03m, mass: 74.91\u0026thinsp;\u0026plusmn;\u0026thinsp;5.38 kg, age: 24.64\u0026thinsp;\u0026plusmn;\u0026thinsp;2.91 years, mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD) participated in the study. Informed written consent was obtained in accordance with the Declaration of Helsinki, and the study was approved by the UCLouvain Ethical Committee (B403201940199). A power analysis, conducted using GLIMMPSE 3.1.2 (Denver, Colorado, USA), confirmed that a sample size of 10 participants was required to achieve adequate statistical power (1- \u0026szlig; = 0.87) and avoid a type II error when considering the model for both kinematic and muscular activity variables.\u003c/p\u003e\n\u003cp\u003eFor this present analysis, participants ran at speeds ranging from 7 to 17 km h\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e on a circular track with a 6 m radius of curvature. The track was placed on 16 1m x 1m instrumented force platforms (described at length with figures in Mesquita et al. \u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003ea). After the curved session, they then ran on a treadmill, to allow for comparison with straight-line running (SLR).\u003c/p\u003e\n\u003cp\u003eEach trial began with participants running along a 20 m straight corridor before entering the curved path, ensuring that they took at least one step on the curve prior to data collection. Participants were instructed to try and maintain a constant average running speed, firstly at intermediate speeds (~\u0026thinsp;12\u0026ndash;14 km h\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e), then at fast speeds (~\u0026thinsp;16\u0026ndash;18 km h\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e), followed by slow speeds (~\u0026thinsp;8\u0026ndash;10 km h\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e). After each trial, the runner\u0026rsquo;s average horizontal velocity and the validity of strides were verified and communicated to the participant. The number of trials per speed category depended on the how many trials it took to achieve\u0026thinsp;\u0026ge;\u0026thinsp;10 valid strides. Rest periods were provided as needed to prevent fatigue-related bias.\u003c/p\u003e\n\u003cp\u003eA trial was considered valid if it included at least one stride (two consecutive steps) in which the horizontal velocity, calculated \u003cem\u003evia\u003c/em\u003e the ground reaction forces, remained constant over a complete stride. A stride began when the vertical component of the ground reaction force became greater than 30 N on one foot and ended when the same foot touched the ground for the following stride: a total of 1021 strides were analysed. In the present study, ground reaction force data has only been used to divide strides.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003eExperimental set-up\u003c/h3\u003e\n\u003cp\u003eThe present study measured two types of variables. First, surface EMG activity in both lower limbs was recorded via 24 surface electrodes. Second, the kinematic movements of body limb segments were captured using high-speed infrared cameras. Two pairs of photocells were positioned\u0026thinsp;~\u0026thinsp;6 m apart at each end of the running track, at level with the neck and were used to trigger and stop the data acquisition for all three systems simultaneously.\u003c/p\u003e\n\u003ch3\u003eMeasurement of EMG variables\u003c/h3\u003e\n\u003cp\u003eThe EMG activity of 12 bilateral lower limb muscles was recorded at 2.15 kHz with a Delsys Trigno surface electrode system of (Natick, MA, USA). Two EMG protocols (A and B) were established to increase the number of muscles recorded. Six subjects used protocol A and five B. In both protocols, 10 out of 12 muscles were the same; the gluteus maximus (Gmax), rectus femoris (ReFe), sartorius (Sart), gracilis (Grac), semi-tendinosum (SeTe), biceps femoris (BiFe), medial gastrocnemius (MeGa), lateral gastrocnemius (LaGa), tibialis anterior (TiAn), peroneus longues (PeLo) of both limbs were always recorded. The muscles that differed between protocols were the tensor fascia latae (TeFL) and vastus medialis (VaMe) for protocol A and gluteus medius (Gmed) and vastus lateralis (VaLa) for protocol B. Six subjects had protocol A applied and five for protocol B. After preparing the shaved skin with alcohol, we placed EMG electrodes based on the SENIAM protocol (seniam.org), the European project on surface EMG. To this end, we located the muscle bellies by means of palpation and oriented the electrodes along the main direction of the fibres (Winter, \u003cspan class=\"CitationRef\"\u003e2009\u003c/span\u003e). The raw EMG signals were down sampled to 1 kHz during post processing to match the other types of data.\u003c/p\u003e\n\u003ch3\u003eMeasurement of kinematic variables\u003c/h3\u003e\n\u003cp\u003eBilateral, full-body three-dimensional kinematics were recorded by means of a Qualisys system (Gothenburg, Sweden) equipped with 12 \u0026ldquo;Mocap OQUS 6+\u0026rdquo; cameras placed around the running track or the treadmill. Subjects were equipped with 43 retroreflective markers placed symmetrically on specific bony anatomical locations. The relevant markers for this study were those placed on the greater trochanters, lateral femoral epicondyles, lateral malleoli and fifth metatarsal-phalangeal joints. Kinematics were recorded at a sampling rate of 240 Hz and oversampled similarly during post-processing at 1 kHz.\u003c/p\u003e\n\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\n\u003ch2\u003eData analysis\u003c/h2\u003e\n\u003cp\u003eAll data were separated into the following speed classes based on the average horizontal velocity (in km h\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e): [7, 8[, [8, 10[, [10,12[, [12, 14[, [14, 16[, [16, 18[. As subjects had to freely match speeds according to vocal guides, in post-processing when less than half of the participants covered the speeds in the class, the group was removed from the analysis. Speed classes counted on average 9.6 subjects (range 7\u0026ndash;11). The analysis was performed \u003cem\u003evia\u003c/em\u003e custom-written programs using LABVIEW (National Instruments 2021, Austin, TX, USA) and Matlab software (MathWorks r2019a, Natick, MA, USA).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n\u003ch2\u003eEMG analysis\u003c/h2\u003e\n\u003cp\u003eThe raw EMG signals were visually inspected for amplifier saturation and noise artefacts. The signals were then high-pass filtered (30 Hz), rectified and low-pass filtered (10 Hz) with a zero-lag fourth-order Butterworth filter to obtain each muscle\u0026rsquo;s linear envelope (Dewolf et al., \u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003eb) (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e). Signals were then cut by stride and the geometric mean of all strides of the same speed and curvature or straight-line was calculated per subject. The geometric average was chosen as it is less sensitive to outliers and thus further reduces noise in the signal (Linssen et al., \u003cspan class=\"CitationRef\"\u003e1993\u003c/span\u003e). The signals were time-normalised to 400 points and to reduce residual baseline noise, the minimum signal for each envelope was subtracted (La Scaleia et al., \u003cspan class=\"CitationRef\"\u003e2014\u003c/span\u003e). This gave us a 28x400 matrix (muscles from both protocols x time-normalisation) per speed class, per radius class and per subject. The waveform amplitudes were also normalised based on the maximum values for each muscle per subject when running on the treadmill at all speeds.\u003c/p\u003e\n\u003cp\u003eMuscle synergies for each limb were extracted from the muscular activity using a dimensionality-reduction non-negative matrix factorisation approach (NNMF), (Lee \u0026amp; Seung, \u003cspan class=\"CitationRef\"\u003e1999\u003c/span\u003e). For each subject, speed radius, and limb condition, the average muscle activation profiles were concatenated into an \u003cem\u003em\u003c/em\u003e x \u003cem\u003en\u003c/em\u003e matrix (\u003cem\u003eM\u003c/em\u003e), where \u003cem\u003em\u003c/em\u003e is the number of muscles (28) and \u003cem\u003en\u003c/em\u003e is the number of time-normalised samples (400). The NNMF algorithm was applied to \u003cem\u003eM\u003c/em\u003e to identify the underlying basic activation temporal patterns, \u003cem\u003eP\u003c/em\u003e, and their associated time-invariant spatial coefficients, \u003cem\u003eC\u003c/em\u003e, (Lee \u0026amp; Seung, \u003cspan class=\"CitationRef\"\u003e1999\u003c/span\u003e):\u003c/p\u003e\n\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e \u003cspan class=\"mathinline\"\u003e\\(\\:M=P\\:\\text{X}\\:C+\\:\\)\u003c/span\u003e \u003c/span\u003e \u003cem\u003ee\u003c/em\u003e, (1)\u003c/p\u003e\n\u003cp\u003ewhere \u003cem\u003ee\u003c/em\u003e is the residual error matrix. The algorithm searches for an approximate solution to minimise the root-mean-square error between \u003cem\u003eM\u003c/em\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:P\\:\\text{X}\\:C\\)\u003c/span\u003e\u003c/span\u003e. For a given number of motor components (see below), the factorisation uses an iterative method starting from randomly chosen initial conditions for \u003cem\u003eP\u003c/em\u003e and \u003cem\u003eC\u003c/em\u003e (Ivanenko, \u003cspan class=\"CitationRef\"\u003e2005\u003c/span\u003e; Sartori et al., \u003cspan class=\"CitationRef\"\u003e2013\u003c/span\u003e). Because the root-mean-square error may have local minima, the best solution was selected out of 100 iterations to find \u003cem\u003eC\u003c/em\u003e and \u003cem\u003eP\u003c/em\u003e from multiple random starting values.\u003c/p\u003e\n\u003cp\u003eThe number of spatial and temporal modules was fixed at 4, based on their temporal associations to specific mechanical events as observed in different SLR stable and unstable running conditions (Santuz et al., \u003cspan class=\"CitationRef\"\u003e2018\u003c/span\u003ea; Santuz et al., \u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003eb) (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e). The goodness of the pattern decomposition was assessed with the percentage of variability accounted (VAF; Torres-Oviedo et al., \u003cspan class=\"CitationRef\"\u003e2006\u003c/span\u003e), defined as:\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv class=\"Heading\"\u003eVAF = \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:1-\\frac{\\text{S}\\text{S}\\text{E}}{\\text{T}\\text{S}\\text{S}}\\)\u003c/span\u003e\u003c/span\u003e, (2)\u003c/div\u003e\n\u003cp\u003ewhere SSE is the sum of squared errors between the experimental and reconstructed data and TSS is the total sum of squares (Dewolf et al., \u003cspan class=\"CitationRef\"\u003e2019\u003c/span\u003e; Dominici et al., \u003cspan class=\"CitationRef\"\u003e2011\u003c/span\u003e; Sartori et al., \u003cspan class=\"CitationRef\"\u003e2013\u003c/span\u003e). The number of muscle synergies has been a topic of discussion (Cappellini et al., \u003cspan class=\"CitationRef\"\u003e2006\u003c/span\u003e; Cheung et al., \u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003e; Lacquaniti et al., \u003cspan class=\"CitationRef\"\u003e2012\u003c/span\u003e; Yokoyama et al., \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e), and several objective methods have been proposed to determine an appropriate VAF cut-off (Cheung et al., \u003cspan class=\"CitationRef\"\u003e2005\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e2009\u003c/span\u003e; Chvatal \u0026amp; Ting, \u003cspan class=\"CitationRef\"\u003e2013\u003c/span\u003e; d\u0026rsquo;Avella et al., \u003cspan class=\"CitationRef\"\u003e2003\u003c/span\u003e). When applying both a linear-fit (10⁻⁵) criterion based to observe when VAF increase between successive synergies becomes negligible (Cheung et al., \u003cspan class=\"CitationRef\"\u003e2005\u003c/span\u003e), or a \u0026lsquo;cusp method\u0026rsquo; (Cheung et al., \u003cspan class=\"CitationRef\"\u003e2009\u003c/span\u003e), comparing the rate of VAF increase in real versus randomised EMG data, the median number of synergies per limb consistently converged to four. We therefore fixed the dimensionality at four synergies and compared differences between conditions within this common framework.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe centre of activity (CoA) was computed for each temporal (\u003cem\u003eP)\u003c/em\u003e synergy pattern as in Martino et al., (\u003cspan class=\"CitationRef\"\u003e2014\u003c/span\u003e) using circular statistics and was expressed in polar coordinates (polar direction denoted the instant \u003cem\u003et\u003c/em\u003e of the stride cycle expressed an angle (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\alpha\\:\\)\u003c/span\u003e\u003c/span\u003e) going from 0 to 360\u0026deg;). The CoA of each synergy was calculated as the angle of the vector (1st trigonometric moment) that points to the centre of mass of that circular distribution using the following formula:\u003c/p\u003e\n\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\n\u003cdiv id=\"FileID_Equ1\" class=\"mathdisplay\"\u003e$$\\:\\text{C}\\text{o}\\text{A}=\\:{\\text{tan}}^{-1}\\left(\\frac{\\sum\\:_{t=1}^{400}\\text{sin}{{\\alpha\\:}}_{t}\\:\\text{x}\\:{\\text{s}\\text{y}\\text{n}}_{t}}{\\sum\\:_{t=1}^{400}\\text{cos}{{\\alpha\\:}}_{t}\\:\\text{x}\\:{\\text{s}\\text{y}\\text{n}}_{t}}\\right)$$\u003c/div\u003e\n\u003cdiv class=\"EquationNumber\"\u003e3\u003c/div\u003e\n\u003c/div\u003e\n\u003cp\u003e.\u003c/p\u003e\n\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\n\u003ch2\u003eHiguchi\u0026rsquo;s Fractal Dimension analysis\u003c/h2\u003e\n\u003cp\u003eTo assess the local complexity of synergy temporal patterns, Higuchi\u0026rsquo;s fractal dimension, HFD, was computed assuming that each time series exhibits scale-invariant self-similar structure properties (Gneiting \u0026amp; Schlather, \u003cspan class=\"CitationRef\"\u003e2004\u003c/span\u003e; Higuchi, \u003cspan class=\"CitationRef\"\u003e1988\u003c/span\u003e; Kesić \u0026amp; Spasić, \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e; Santuz et al., \u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003ea; Santuz et al., \u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003eb). For each synergy\u0026rsquo;s temporal activation pattern, \u003cem\u003eP\u003c/em\u003e\u003csub\u003e(t)\u003c/sub\u003e [\u003cem\u003eP\u003c/em\u003e\u003csub\u003e(1)\u003c/sub\u003e, \u003cem\u003eP\u003c/em\u003e\u003csub\u003e(2)\u003c/sub\u003e, \u003cem\u003e... P\u003c/em\u003e\u003csub\u003e(n)\u003c/sub\u003e], \u003cem\u003ek\u003c/em\u003e sets of new time series were constructed, where \u003cem\u003ek\u003c/em\u003e was an integer interval between 2\u0026thinsp;\u0026lt;\u0026thinsp;\u003cem\u003ek\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;\u003cem\u003ek\u003c/em\u003e\u003csub\u003emax\u003c/sub\u003e:\u003c/p\u003e\n\u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\n\u003cdiv id=\"FileID_Equ2\" class=\"mathdisplay\"\u003e$$\\:{P}_{k}^{{t}_{0}}:\\:{P}_{{t}_{0}},\\:{P}_{{t}_{0}+k},\\:{P}_{{t}_{0}+2k},\\:\\dots\\:,\\:\\left[{t}_{0}+int\\left(\\frac{n-{t}_{0}}{k}\\right)k\\right]$$\u003c/div\u003e\n\u003cdiv class=\"EquationNumber\"\u003e4\u003c/div\u003e\n\u003c/div\u003e\n\u003cp\u003e,\u003c/p\u003e\n\u003cp\u003e\u003cem\u003et\u003c/em\u003e \u003csub\u003e0\u003c/sub\u003e is the first sample of ranges 1 to \u003cem\u003ek\u003c/em\u003e, used to generate the subsampled series. Debate exists in the literature as to what value one should assign to \u003cem\u003ek\u003c/em\u003e\u003csub\u003emax\u003c/sub\u003e (Kesić \u0026amp; Spasić, \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e). As in Santuz et al., (\u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003eb) we selected a \u003cem\u003ek\u003c/em\u003e\u003csub\u003emax\u003c/sub\u003e which was the \u0026ldquo;most linear part of the log-log plot\u0026rdquo;. As such, we ran an HFD analysis for multiple values of \u003cem\u003ek\u003c/em\u003e and selected the point where the plot was most linear for each synergy, the median value for all synergies, speeds and curves was \u003cem\u003ek\u003c/em\u003e\u003csub\u003emax\u003c/sub\u003e = 40. Therefore, for each trial, the HFD of each muscle synergy was computed separately with the median \u003cem\u003ek\u003c/em\u003e\u003csub\u003emax\u003c/sub\u003e and then compared across limbs and speed. The non-Euclidean length of each \u003cem\u003ek\u003c/em\u003e subsample was then defined as:\u003c/p\u003e\n\u003cdiv id=\"Equ3\" class=\"Equation\"\u003e\n\u003cdiv id=\"FileID_Equ3\" class=\"mathdisplay\"\u003e$$\\:{L}_{{t}_{0}}\\left(k\\right)=\\frac{1}{k}\\left\\{\\frac{n-1}{int\\left(\\frac{n-{t}_{0}}{k}\\right)k}\\left[\\sum\\:_{i=1}^{int\\left(\\frac{n-{t}_{0}}{k}\\right)}\\left|{P}_{{t}_{0}+ik}-{P}_{{t}_{0}-\\left(i-1\\right)k}\\right|\\right]\\right\\}$$\u003c/div\u003e\n\u003cdiv class=\"EquationNumber\"\u003e5\u003c/div\u003e\n\u003c/div\u003e\n\u003cp\u003e,\u003c/p\u003e\n\u003cp\u003eand for every considered \u003cem\u003ek\u003c/em\u003e step considered, the average of the \u003cem\u003ek\u003c/em\u003e sets of lengths of the temporal pattern was defined as:\u003c/p\u003e\n\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\n\u003cdiv id=\"FileID_Equa\" class=\"mathdisplay\"\u003e$$\\:L\\left(k\\right)=\\frac{1}{k}\\sum\\:_{{t}_{0}}^{k}{L}_{{t}_{0}}\\left(k\\right)$$\u003c/div\u003e\n\u003c/div\u003e\n\u003cp\u003e.\u003c/p\u003e\n\u003cp\u003eIf L(\u003cem\u003ek\u003c/em\u003e) \u0026prop; \u003cem\u003ek\u003c/em\u003e\u003csup\u003e\u0026minus;\u0026thinsp;HFD\u003c/sup\u003e, then the curve is fractal, exhibiting self-similar structure across scale, with dimension HFD explaining its complexity. HFD ranges from 1 (\u003cem\u003ee.g.\u003c/em\u003e, a smooth time series) to 2 (\u003cem\u003ee.g.\u003c/em\u003e, random white noise) and is independent on the amplitude of the signal.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n\u003ch2\u003e\u003cem\u003eDissimilarity\u003c/em\u003e index and weighted difference\u003c/h2\u003e\n\u003cp\u003eTo assess the similarity between inner and outer limb temporal and spatial patterns, a simplified Procrustes analysis was used. Procrustes analysis quantifies the similarity between two matrices by optimally aligning them through translation, isotropic scaling, and rotation (Andreella et al., \u003cspan class=\"CitationRef\"\u003e2023\u003c/span\u003e; Zaidi \u0026amp; Harris-Love, \u003cspan class=\"CitationRef\"\u003e2023\u003c/span\u003e). Specifically, one matrix is considered as a reference, and the second matrix is transformed to minimise the Frobenius norm of their difference. Since we wanted to quantify similarities between specific synergies, each comparison was done between 1 x \u003cem\u003en\u003c/em\u003e spatial or temporal matrices and as such, no matrix rotation was employed. The resulting Procrustes error quantified dissimilarity between each. Given its novelty in synergy similarity calculation, a similarity index using the best-matching scalar product of temporal patterns and basic activation coefficients normalised to the Euclidean norm was computed as in Martino et al., (\u003cspan class=\"CitationRef\"\u003e2015\u003c/span\u003e) and can be found in the supplementary material for comparison. These yielded similar results.\u003c/p\u003e\n\u003cp\u003eFurthermore, since a dissimilarity index does not indicate which individual muscles drive the observed differences, we examined the individual muscle weighting differences of spatial patterns within each synergy between limbs \u003cem\u003evia\u003c/em\u003e linear mixed-effects modelling. This allowed us to identify how changes in specific muscles partly contributed to the dissimilarity index.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n\u003ch2\u003eHandling Kinematic turning data: Range of motion and intersegmental coordination\u003c/h2\u003e\n\u003cp\u003eIn our previous papers on this subject (Mesquita et al., \u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003ea, \u0026amp; Mesquita et al., \u003cem\u003eunder review\u003c/em\u003e) we presented a method for rotating ground reaction forces from a \u003cem\u003eo-x-y-z\u003c/em\u003e reference frame, which was fixed to the lab to a reference frame \u003cem\u003eO- X- Y- z\u003c/em\u003e that moves as the runner turns. To achieve this, we calculated a virtual position for the centre of the pelvis in the transverse plane, \u003cem\u003eP\u003c/em\u003e\u003csub\u003ec\u003c/sub\u003e, as the centroid of a triangle between the right and left ASIS and S2 markers. We then ran a circular fit on the movement of \u003cem\u003eP\u003c/em\u003e\u003csub\u003ec\u003c/sub\u003e in the transverse plane to determine the radius of the curve covered by the subject over the trial and rotated the transverse plane forces and kinematics by the angle (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\theta\\:\\)\u003c/span\u003e\u003c/span\u003e) formed by the \u003cem\u003eo-x\u003c/em\u003e axis, the position of \u003cem\u003eP\u003c/em\u003e\u003csub\u003ec\u003c/sub\u003e and the centre of a circle.\u003c/p\u003e\n\u003cp\u003eHere, we rotated the kinematic marker positions based on the instantaneous \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\theta\\:}_{j}\\)\u003c/span\u003e\u003c/span\u003e-angle. In this way, the (\u003cem\u003ex,y\u003c/em\u003e) positions of each marker (\u003cem\u003ei\u003c/em\u003e) in the fixed reference frame were rotated at each instant (\u003cem\u003ej\u003c/em\u003e) of the stride to give a lateral and fore-aft, respectively (\u003cem\u003eX,Y\u003c/em\u003e), position relative a non-inertial reference frame:\u003c/p\u003e\n\u003cdiv id=\"Equ4\" class=\"Equation\"\u003e\n\u003cdiv id=\"FileID_Equ4\" class=\"mathdisplay\"\u003e$$\\:\\left(\\begin{array}{c}{X}_{i,j}\\\\\\:{Y}_{i,j}\\end{array}\\right)=\\:\\left(\\begin{array}{c}{x}_{i,j}\\text{cos}{\\theta\\:}_{j}-{y}_{i,j}\\text{sin}{\\theta\\:}_{j}\\\\\\:{x}_{i,j}\\text{sin}{\\theta\\:}_{j}+{y}_{i,j}\\text{cos}{\\theta\\:}_{j}\\end{array}\\right)$$\u003c/div\u003e\n\u003cdiv class=\"EquationNumber\"\u003e4\u003c/div\u003e\n\u003c/div\u003e\n\u003cp\u003e.\u003c/p\u003e\n\u003cp\u003eThis rotation is possible given the angular displacement covered by the outer and inner limb are symmetric over a step (Mesquita et al., \u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003ea). Note, the vertical axis was not rotated compared to body pitch as to maintain a reference perpendicular to the ground. From these rotated marker positions, the elevation angles in the rotated sagittal (\u003cem\u003eO-Y-z\u003c/em\u003e) plane, \u003cem\u003ei.e.\u003c/em\u003e, the orientation of the limb segments of both lower limbs relative to vertical were computed. This was also similar to the rotation applied in a walking while turning kinematics study (Courtine \u0026amp; Schieppati, \u003cspan class=\"CitationRef\"\u003e2004\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eIn each speed and radius class, the average waveforms of the elevation angles of the three lower-limb segments, thigh, shank, and foot, were computed across participants. The segment range of motion (ROM) over a stride was computed as the difference between the maxima and the minima elevation angle over the period and was used to analyse the amplitude of each segment. To evaluate the time-course components of the elevation angles over one stride, a first-harmonic Fourier fit was performed (Dewolf et al., \u003cspan class=\"CitationRef\"\u003e2018\u003c/span\u003e). The amplitude (\u003cem\u003eA\u003c/em\u003e), phase shift (\u003cem\u003eP\u003c/em\u003e) and variance accounted for by the first harmonic were computed. The amplitude ratio (\u003cem\u003eG\u003c/em\u003e\u003csub\u003epd\u003c/sub\u003e) and phase shift (F\u003csub\u003epd\u003c/sub\u003e) between two adjacent limb segments \u003cem\u003ep\u003c/em\u003e and \u003cem\u003ed\u003c/em\u003e were computed as \u003cem\u003eG\u003c/em\u003e\u003csub\u003epd\u003c/sub\u003e = \u003cem\u003eA\u003c/em\u003e\u003csub\u003ed\u003c/sub\u003e/\u003cem\u003eA\u003c/em\u003e\u003csub\u003ep\u003c/sub\u003e and as F\u003csub\u003epd\u003c/sub\u003e = \u003cem\u003eP\u003c/em\u003e\u003csub\u003ed\u003c/sub\u003e \u0026ndash; \u003cem\u003eP\u003c/em\u003e\u003csub\u003ep\u003c/sub\u003e.\u003c/p\u003e\n\u003cp\u003eThe inter-segmental coordination of the lower-limb segments was analysed using principal component analysis (PCA), a method previously described in detail (Borghese et al., \u003cspan class=\"CitationRef\"\u003e1996\u003c/span\u003e; Catavitello et al., \u003cspan class=\"CitationRef\"\u003e2018\u003c/span\u003e; Ivanenko et al., \u003cspan class=\"CitationRef\"\u003e2008\u003c/span\u003e). Briefly, when the thigh, shank, and foot, 400-point time-normalised, elevation angles are plotted in three-dimensional space, which form a closed-loop trajectory that lies near a plane. The best-fitting plane is defined by the first two principal components obtained from the PCA of the covariance matrix of the three elevation angle waveforms (Daffertshofer et al., \u003cspan class=\"CitationRef\"\u003e2004\u003c/span\u003e; Ivanenko et al., \u003cspan class=\"CitationRef\"\u003e2008\u003c/span\u003e) (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003eB). The PCA returns a matrix of eigenvectors \u003cem\u003eu\u003c/em\u003e with associated eigenvalues (\u0026lambda;1, \u0026lambda;2, \u0026lambda;3) ordered from the largest to the smallest. The first two eigenvectors, \u003cem\u003eu\u003c/em\u003e\u003csub\u003e1\u003c/sub\u003e and \u003cem\u003eu\u003c/em\u003e\u003csub\u003e2\u003c/sub\u003e generate the best fitting plane while the third eigenvector, \u003cem\u003eu\u003c/em\u003e\u003csub\u003e3\u003c/sub\u003e, orthogonal to the plane, defines its orientation. The components of \u003cem\u003eu\u003c/em\u003e\u003csub\u003e3\u003c/sub\u003e: \u003cem\u003eu\u003c/em\u003e\u003csub\u003e3\u003c/sub\u003e\u003cem\u003et\u003c/em\u003e, \u003cem\u003eu\u003c/em\u003e\u003csub\u003e3\u003c/sub\u003e\u003cem\u003es\u003c/em\u003e and \u003cem\u003eu\u003c/em\u003e\u003csub\u003e3\u003c/sub\u003e\u003cem\u003ef\u003c/em\u003e correspond to the direction cosines with the positive semi-axis of the thigh, shank and foot angular coordinates. The percentage of variance associated to vectors \u003cem\u003eu\u003c/em\u003e\u003csub\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sub\u003e, \u003cem\u003eu\u003c/em\u003e\u003csub\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sub\u003e, \u003cem\u003eu\u003c/em\u003e\u003csub\u003e\u003cem\u003e3\u003c/em\u003e\u003c/sub\u003e are respectively PV1, PV2 and PV3 and are computed as the ratio of each eigenvalue to the total eigenvalue sum (\u003cem\u003ee.g.\u003c/em\u003e, PV3\u0026thinsp;=\u0026thinsp;\u0026lambda;3/ (\u0026lambda;1\u0026thinsp;+\u0026thinsp;\u0026lambda;2\u0026thinsp;+\u0026thinsp;\u0026lambda;3)). PV3 was used as an index of planarity, with PV3\u0026thinsp;=\u0026thinsp;0% indicating maximum planarity.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n\u003ch2\u003eStatistics\u003c/h2\u003e\n\u003cp\u003eFor each participant, the different parameters were first averaged for all inner and outer steps over one trial. Then descriptive statistical analysis was performed. A Shapiro-Wilk test was performed to verify normality on linear data. Then, a linear mixed effect model with pairwise comparison and Bonferroni post-hoc correction was used to compute the individual effects of radius, speed and limb on the calculated variables.\u003c/p\u003e\n\u003cp\u003eFor circular data, \u003cem\u003ei.e.\u003c/em\u003e, CoA, the mean resultant length (\u003cem\u003er\u003c/em\u003e) was evaluated for each variable. For data which were concentrated appropriately (\u003cem\u003er\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.5) a Watson-Williams test was performed with pairwise comparison and a Bonferroni post-hoc correction was used to compute factor effects. Statistical tests were run on IBM SPSS Statistics (PASW Statistic 27, NY, USA) for linear data and on Matlab for circular data. The results of the statistical tests were considered significant for a p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eMuscular activity and coordination\u003c/h2\u003e \u003cp\u003eThe overall average variance accounted for, VAF, across all subjects, speeds and radii was 94.91\u0026thinsp;\u0026plusmn;\u0026thinsp;2.2%. VAF increased at the lowest speeds and levelled off at higher speeds (F\u0026thinsp;=\u0026thinsp;3.66, p\u0026thinsp;=\u0026thinsp;0.004). When considering both limbs, the VAF was higher in curved compared to SLR (F\u0026thinsp;=\u0026thinsp;4.77, p\u0026thinsp;=\u0026thinsp;0.031). This increase was largely due to the inner limb, in which the VAF was on average 0.9% greater than the outer limb effect (F\u0026thinsp;=\u0026thinsp;12.243, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Indeed, the outer limb was not different when compared to SLR (Bonferroni \u003cem\u003epost-hoc\u003c/em\u003e p\u0026thinsp;=\u0026thinsp;0.4).\u003c/p\u003e \u003cp\u003eSynergies were classified according to their temporal alignment with specific mechanical events, as previously described in SLR under stable and unstable running conditions (Santuz et al., \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2018\u003c/span\u003ea; Santuz et al., \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2020\u003c/span\u003eb). Synergy 1 was associated with initial ground contact and the braking phase of stance. It was dominated by proximal hip (Gmax, Gmed) and knee extensors (VaMe, VaLe, ReFe), muscles. Synergy 2 was linked to the propulsive phase of stance, with its CoA occurring around mid-stance. It was primarily explained by the ankle plantarflexors (MeGa, LaGa, PeLo), all acting on distal joints, and producing positive work to reaccelerate the CoM. Synergy 3 corresponded to early-mid swing and was dominated by hip flexors (TeFe, Sart, ReFe) and ankle dorsiflexors (TiAn), ensuring foot clearance to prevent \u0026lsquo;foot-catch\u0026rsquo; during swing. Synergy 4 was associated with late swing and preparation for the next stride, dominated by both hip flexors (BiFe, SeTe, Grac) and ankle dorsiflexors (TiAn).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eWhen observing each temporal activation pattern individually, CoA was anticipated, \u003cem\u003ei.e.\u003c/em\u003e, its principal burst of activity arrived earlier, in curved running compared to SLR for all synergies (synergy 1: F\u0026thinsp;=\u0026thinsp;32.49, synergy 2: F\u0026thinsp;=\u0026thinsp;29.36, synergy 3: F\u0026thinsp;=\u0026thinsp;5.16 and synergy 4: F\u0026thinsp;=\u0026thinsp;44.31, all synergies : p\u0026thinsp;\u0026lt;\u0026thinsp;0.02) (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e4\u003c/span\u003eA). In synergies 2 and 4, the outer limb was more anticipated than its inner counterpart (respectively, F\u0026thinsp;=\u0026thinsp;6.86, 12.43, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01). In the first and third synergies, limbs did not differ in their CoA during curved running (synergy 1: F\u0026thinsp;=\u0026thinsp;2.76, synergy 3: 0.32, p\u0026thinsp;\u0026gt;\u0026thinsp;0.1). However, both limbs were more anticipated in synergy 1 (Bonferroni \u003cem\u003epost-hoc\u003c/em\u003e p\u0026thinsp;\u0026lt;\u0026thinsp;0.01) whereas in synergy 3 only the outer limb differed to SLR (Bonferroni \u003cem\u003epost-hoc\u003c/em\u003e p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). When observing the temporal dissimilarity index (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e4\u003c/span\u003eB), synergies 2 and 4 showed distinct shape differences in curved running compared to SLR (synergy 2: F\u0026thinsp;=\u0026thinsp;20.60, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001, synergy 4: F\u0026thinsp;=\u0026thinsp;5.38, p\u0026thinsp;=\u0026thinsp;0.24) while the shape difference in synergies involved in the initial touchdown and beginning of swing phase, 1 and 3, did not differ between curved and SLR (synergy 1: F\u0026thinsp;=\u0026thinsp;0.17, p\u0026thinsp;=\u0026thinsp;0.685, synergy 3: F\u0026thinsp;=\u0026thinsp;0.29, p\u0026thinsp;=\u0026thinsp;0.5913).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eHiguchi\u0026rsquo;s fractal dimension of the temporal patterns, Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e4\u003c/span\u003eC, varied depending on specific muscle synergies in curved running. Synergies 1 and 4, associated with the touchdown and the late swing phase, were both less complex, \u003cem\u003ei.e.\u003c/em\u003e, smaller HFD, in curved running compared to SLR (synergy 1: F\u0026thinsp;=\u0026thinsp;34.11, synergy 4: F\u0026thinsp;=\u0026thinsp;40.39, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Synergy 1 showed less complexity for the inner compared to the outer limb (synergy 1: F\u0026thinsp;=\u0026thinsp;4.43, p\u0026thinsp;=\u0026thinsp;0.037). The second synergy was more complex in curved running compared to SLR (F\u0026thinsp;=\u0026thinsp;39.82, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and increased with speed (F\u0026thinsp;=\u0026thinsp;8.59, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001).\u003c/p\u003e \u003cp\u003eRegarding the spatial activation patterns dissimilarity index (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e4\u003c/span\u003eB), all synergies had greater dissimilarity in curved running compared to SLR (synergy 1: F\u0026thinsp;=\u0026thinsp;18.10, synergy 2: F\u0026thinsp;=\u0026thinsp;20.50, synergy 3: F\u0026thinsp;=\u0026thinsp;5.70 and F\u0026thinsp;=\u0026thinsp;9.49; p\u0026thinsp;\u0026lt;\u0026thinsp;0.02) and only synergy 2 also changed with speed (F\u0026thinsp;=\u0026thinsp;7.72, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Furthermore, since a dissimilarity index does not indicate which muscle weighting and which limb contributes to these differences, a signed difference was computed of spatial patterns per limb and per synergy. Synergy 1 showed specific increases in the outer limb BiFe and PeLo muscle contributions (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) whereas the inner limb had increased contribution of the Grac, SeTe and Sart (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01). The second synergy showed an increase in the inner PeLo (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and outer MeGa and SeTe (p\u0026thinsp;\u0026lt;\u0026thinsp;0.008). The muscle patterns tied to the initial swing phase had increased contribution for the inner Grac and TiAn (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and increased contribution for the outer PeLo and Sart (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Finally, synergy 4 showed an increased contribution to the outer TiAn (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and inner BiFe and SeTe (p\u0026thinsp;\u0026lt;\u0026thinsp;0.005).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eLower limb segment coordination\u003c/h2\u003e \u003cp\u003eThe ROM of each lower limb segment increased with speed (thigh: F\u0026thinsp;=\u0026thinsp;750.01; shank: F\u0026thinsp;=\u0026thinsp;514.60; foot: F\u0026thinsp;=\u0026thinsp;280.70, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e5\u003c/span\u003eA). The ROM of the thigh and shank increased on curve as compared to SLR (thigh: F\u0026thinsp;=\u0026thinsp;44.45, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; shank: F\u0026thinsp;=\u0026thinsp;8.18, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). When comparing limbs, the inner thigh and shank had a greater ROM than their outer counterparts (thigh: F\u0026thinsp;=\u0026thinsp;19.15, shank: F\u0026thinsp;=\u0026thinsp;12.62, p\u0026thinsp;\u0026lt;\u0026thinsp;0.005). The foot ROM, however, did not change significantly with radius nor between limbs (radius F\u0026thinsp;=\u0026thinsp;0.49, p\u0026thinsp;=\u0026thinsp;0.49, limb: F\u0026thinsp;=\u0026thinsp;2.30, p\u0026thinsp;=\u0026thinsp;0.13) (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e5\u003c/span\u003eA, \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e5\u003c/span\u003eB).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eWhen analysing the time-course variables, the first harmonic accounted for a large portion of the accounted variance across all speeds and radius scenarios (overall average across limbs: 90.19\u0026thinsp;\u0026plusmn;\u0026thinsp;5.48% mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD). At the level of the thigh and shank, the phase between the two segment\u0026rsquo;s first harmonic decreased with speed (F\u0026thinsp;=\u0026thinsp;5.10, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and decreased in curved running as compared to SLR (F\u0026thinsp;=\u0026thinsp;10.95, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). The outer limb phase shift was smaller than its inner counterpart (F\u0026thinsp;=\u0026thinsp;25.77, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). The latter was not different from the shift observed in SLR (rad*limb \u003cem\u003epost-hoc\u003c/em\u003e comparison between the inner limb in SLR and curved running p\u0026thinsp;=\u0026thinsp;0.244). The thigh-shank amplitude ratio was smaller in curved running as compared to SLR (F\u0026thinsp;=\u0026thinsp;9.80, p\u0026thinsp;\u0026lt;\u0026thinsp;0.002) but did not change with speed (F\u0026thinsp;=\u0026thinsp;0.516, p\u0026thinsp;\u0026lt;\u0026thinsp;0.796) nor between limbs (F\u0026thinsp;=\u0026thinsp;2.87, p\u0026thinsp;\u0026lt;\u0026thinsp;0.09).\u003c/p\u003e \u003cp\u003eBetween the shank and foot, the phase-shift increases with speed (F\u0026thinsp;=\u0026thinsp;13.37, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and decreases in curved running (F\u0026thinsp;=\u0026thinsp;19.55, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). The inner limb shank and foot were more out of phase than in the outer limb (limb F\u0026thinsp;=\u0026thinsp;22.78, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), which remained similar to SLR (rad*limb \u003cem\u003epost-hoc\u003c/em\u003e comparison between the outer limb in SLR and curved running p\u0026thinsp;=\u0026thinsp;0.130). The shank-foot amplitude ratio changed with both speed (F\u0026thinsp;=\u0026thinsp;38.29, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and curvature (F\u0026thinsp;=\u0026thinsp;8.13, p\u0026thinsp;\u0026lt;\u0026thinsp;0.005). The inner limb decreased its ratio in curved running (rad*limb \u003cem\u003epost-hoc\u003c/em\u003e for the inner limb p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) whereas the outer limb maintained a similar behaviour in both curved and SLR.\u003c/p\u003e \u003cp\u003eRegarding the planar covariation, planarity was respected at all speeds in curved running since PV1\u0026thinsp;+\u0026thinsp;PV2\u0026thinsp;\u0026gt;\u0026thinsp;98% and decreased compared to SLR (F\u0026thinsp;=\u0026thinsp;6.95, p\u0026thinsp;\u0026lt;\u0026thinsp;0.009). PV1 decreased (F\u0026thinsp;=\u0026thinsp;46.03, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) as PV2 increased (F\u0026thinsp;=\u0026thinsp;26.35, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) in curved compared to SLR. Furthermore, particularly at high speeds, the rotation of the covariation plane differed between the inner and outer limbs (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e3\u003c/span\u003eB). This is further shown when observing the inner limb \u003cem\u003eu\u003c/em\u003e\u003csub\u003e3\u003c/sub\u003e\u003cem\u003et\u003c/em\u003e component, orthogonal to the plane (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e6\u003c/span\u003eB, left), which progressively dissociated from the outer limb in curved running (limb effect: F\u0026thinsp;=\u0026thinsp;52.95, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001, radius*limb effect: F\u0026thinsp;=\u0026thinsp;17.74, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). The inner limb relied less on PV1 (limb effect: F\u0026thinsp;=\u0026thinsp;33.12, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and had a higher percentage of variance explained by PV2 (limb effect: F\u0026thinsp;=\u0026thinsp;18.70, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) compared to the outer limb (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e6\u003c/span\u003eB, right).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study aimed to investigate how curved running modifies one\u0026rsquo;s muscular and segmental programming patterns compared to straight-line running. Our results showed that muscle synergies were restructured during curved running, with earlier activation timings, altered complexity, and distinct spatial and temporal recruitment patterns relative to straight-line conditions. These adaptations were also limb-specific, as the inner limb generally displayed greater adjustments in muscle weighting and reduced complexity, whereas the outer limb showed more pronounced anticipations in certain synergies. Regarding kinematic coordination, curved running modified intersegmental phase relations and planar covariation of lower limb segments, again with asymmetric contributions from the inner and outer limbs.\u003c/p\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eMuscular Coordination\u003c/h2\u003e \u003cp\u003eResults from muscular synergies aligned well with our previous observations at the spinal motor output level (Mesquita et al., \u003cem\u003eunder review\u003c/em\u003e). Similar to curved walking (Courtine \u0026amp; Schieppati, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2003\u003c/span\u003e), overall the muscle activation patterns did not markedly differ from those in SLR during curved running, and, thus, the general structure of the muscle synergies was also largely preserved. Nevertheless, systematic temporal and spatial modifications emerged.\u003c/p\u003e \u003cp\u003eAcross all synergies, curved running showed anticipatory shifts in the CoA, with activations occurring earlier in the cycle compared with SLR (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e4\u003c/span\u003eA). These anticipations were also limb-dependent in synergies 2 and 4, where the outer limb anticipated more than the inner. We hypothesised that such anticipatory behaviour may result from heightened muscle spindle sensitivity, facilitating tendomuscular systems in tolerating and absorbing greater impact loads (Gollhofer \u0026amp; Kyr\u0026ouml;l\u0026auml;inen, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e1991\u003c/span\u003e; Kyr\u0026ouml;l\u0026auml;inen et al., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2005\u003c/span\u003e). This explanation remains plausible, and suggests that reactive feedback adaptations alter muscle synergies differently between limbs (Donelan \u0026amp; Pearson, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2004\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn parallel, evidence from postural control research indicates that anticipatory adjustments can reorganise muscle coordination before predictable perturbations or voluntary actions, reflecting proactive feedforward reconfiguration of synergies to stabilise task performance (Klous et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Krishnan et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). Although demonstrated in postural contexts, a similar mechanism may underlie the anticipatory shifts observed here, with the locomotor system recalibrating modular activation in advance to meet the changing mechanical demands of curved running.\u003c/p\u003e \u003cp\u003eA second point previously observed (Mesquita et al., \u003cem\u003eunder review\u003c/em\u003e) was a proximal-to-distal redistribution of joint work during curved running, with both limbs producing more positive work at the ankle, similar to unstable locomotion in birds and uphill running in humans (Daley et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Qiao et al., \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). This distal shift required compensatory energy absorption in the proximal hip joint, consistent with greater lumbar than sacral MN output. These results support segment-specific spinal specialisation, with lumbar segments driving rhythm generation and sacral segments adapting to feedback and foot\u0026ndash;ground interactions (Dewolf et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Lacquaniti et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Mesquita et al., \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Minassian et al., \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Selionov et al., \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2009\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eSynergy 2, active during push-off and primarily driven by the MeGa, LaGa, and PeLo, was where distal muscles produced the most positive work. It also showed greater anticipation in the outer limb, along with larger spatial and temporal dissimilarities between limbs and a wider divergence from SLR patterns (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e4\u003c/span\u003eB). Notably, MeGa contributed more to the outer limb, while PeLo contributed more to the inner limb throughout push-off, consistent with sacral-level adaptations associated with foot placement. Importantly, synergy 2 represents the phase most tightly coupled to reflex-driven sensory feedback, as the period starts where the limb is maximally loaded and ends where the hip flexor tendons are maximally stretched, both critical sources of afferent input (af Klint et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Donelan \u0026amp; Pearson, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2004\u003c/span\u003e; Ekeberg \u0026amp; Pearson, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Pearson, \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). These features make the second synergy a key moment for task-specific reactive control, integrating load- and stretch-sensitive feedback to stabilise and adapt to challenging locomotor conditions.\u003c/p\u003e \u003cp\u003eIn curved running, the inner and outer limbs have distinct mechanical functions, requiring partly independent control to redirect the centre of mass laterally and accommodate with increased lateral forces (Mesquita et al., \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2024\u003c/span\u003ea). The inner limb primarily contributed to lateral redirection, with greater involvement of hip muscles such as the tensor fasciae latae and gracilis (Mesquita et al., \u003cem\u003eunder revision\u003c/em\u003e). Neural adaptations can be asymmetrical, as shown in split-belt treadmill walking where each limb develops distinct neuromuscular responses under different mechanical demands (MacLellan et al., \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Ogawa et al., \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Yokoyama et al., \u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn line with this, we observed systematic limb differences in activation patterns across all synergies (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e4\u003c/span\u003eB), which likely reflect a global reorganisation at the pattern formation level. By contrast, temporal shifts confined to synergies 2 and 4 suggest more targeted, sensory-driven modulation (Dzeladini et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2014\u003c/span\u003e), consistent with their sacral motoneuron output dominance and modifications due to environmental interactions. This interpretation fits with CPG-circuit models that show a two-layer organisation, one for rhythm generation and a second for pattern formation, where both descending inputs and sensory feedback modulate motor output (Danner et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Dzeladini et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eLastly, four synergies explained over 94% of variance on average from our data using NNMF. The VAF for the inner limb muscular activity was on average\u0026thinsp;~\u0026thinsp;1% greater than those of both SLR and the outer limb muscular activity, implying that the model better explained the inner compared to the outer limb data. This suggested that the inner limb possibly operated in a more constrained and less complex manner, similar to observations of running in challenging locomotor environments (Santuz et al., \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2018\u003c/span\u003ea; Santuz et al., \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2020\u003c/span\u003eb).\u003c/p\u003e \u003cp\u003eTo better analyse this complexity, Higuchi\u0026rsquo;s fractal dimension (Higuchi, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e1988\u003c/span\u003e) was applied to the temporal activation patterns and showed that curved running was generally less complex than SLR in synergies linked to touch-down and late-swing, while synergy 2 became more complex. Limb-specific complexity differences were particularly evident in synergy 1, where the outer limb showed greater HFD than its inner counterpart. Although small (\u0026lt;\u0026thinsp;1% VAF), these differences indicate that inner-limb synergies, especially at touchdown, were less complex than their outer counterparts. This reduction likely reflects more robust control, with signals being expressed more as singular bursts of activity with less fragmented additional bursts. Previous accounts have explained this, in part, by a widening of the synergy temporal parameters which begin to overlap and lose in robustness (Santuz et al., \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2020\u003c/span\u003ea). In the present case, no statistically significant effects were found for synergy width increase, neither with speed nor curvature, however, trends consistently indicated that synergies were wider in both situations.\u003c/p\u003e \u003cp\u003eIn contrast, synergy 2, showed an increase in complexity. This likely reflected the need for finer control of distal musculature to accommodate stability demands during turning. Taken together, these findings suggest that while some synergies become more robust (less complex) under curved running, others can be selectively tuned, with increased complexity emerging when their functional role is critical for task stability.\u003c/p\u003e \u003cdiv id=\"Sec19\" class=\"Section3\"\u003e \u003ch2\u003eLimb coordination\u003c/h2\u003e \u003cp\u003eRegarding the kinematic coordination of the lower limb segments (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e3\u003c/span\u003e), the ROM of the thigh and shank of both limbs increased in curved running compared to SLR (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e4\u003c/span\u003e). In terms of the orientation of the covariance plane, when pooling both limbs together, the segment coordination did not differ from SLR. This emphasizes the robustness of the planar covariance law in the sagittal plane segment angular motion (Ivanenko et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; Lacquaniti et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). However, the amount of variance explained by either changed, PC1 decreased whereas PC2 increased, similar to curved walking observations (Courtine \u0026amp; Schieppati, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2004\u003c/span\u003e), which shows that kinematic modifications do occur to account for the specific task.\u003c/p\u003e \u003cp\u003eBy looking at the outer and inner segmental coordination, differences in the movement programming between limbs emerged. The coordination loops (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e3\u003c/span\u003e) revealed that the outer limb followed the long axis more closely, which is better explained by the first eigenvector (increased PV1), whereas the inner limb aligned more with the wide axis, better explained by the second eigenvector (increased PV2). This suggested that the outer limb\u0026rsquo;s movement was primarily governed by orientation changes while the inner limb\u0026rsquo;s motion was more influenced by variations in limb length (Ivanenko et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2008\u003c/span\u003e), indicating distinct strategies in limbs for managing the curved running motion. This was consistent with observations that limb function affects the plane orientation (Catavitello et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThis can be further observed in the phase differences between segments and their relationship to an overall coordination. Previously, Courtine \u0026amp; Schieppati (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2004\u003c/span\u003e) showed that in both linear and curved walking the thigh-shank phase relation was strongly related to loop shape, represented by \u003cem\u003eu\u003c/em\u003e\u003csub\u003e1\u003c/sub\u003e\u003cem\u003et\u003c/em\u003e. Here, when we regress the F\u003csub\u003ets\u003c/sub\u003e onto \u003cem\u003eu\u003c/em\u003e\u003csub\u003e1\u003c/sub\u003e\u003cem\u003et\u003c/em\u003e we see distinct limb-specific patterns (Fig S2), where the slope formed by the outer limb relationship was similar to SLR, albeit with less variance explained (R\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.27), whereas the inner limb had explained variance higher (R\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.36) but with a different slope. This indicates that the inner limb\u0026rsquo;s thigh\u0026ndash;shank timing better explained the loop\u0026rsquo;s main axis, \u003cem\u003ei.e.\u003c/em\u003e, limb orientation variations and that the outer limb kept a similar behaviour to SLR (see supplementary material).\u003c/p\u003e \u003cp\u003eIn contrast, when regressing the thigh-shank phase difference onto \u003cem\u003eu\u003c/em\u003e\u003csub\u003e2\u003c/sub\u003e\u003cem\u003et\u003c/em\u003e, the projection of PC2 (the loop width), the inner limb had a stronger relationship (R\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.67 for the inner limb, \u003cem\u003evs.\u003c/em\u003e 0.44 in SLR) with a steeper slope, while the outer limb loses nearly all association (R\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.14). These results suggest that during turning, the thigh \u0026ndash; shank phase shifts are increasingly expressed along the orthogonal axis for the inner limb, reflecting in-plane adjustments of loop geometry, whereas the outer limb preserves its primary phase\u0026ndash;orientation relationship.\u003c/p\u003e \u003cp\u003eIn the discussion on curved walking, Courtine \u0026amp; Schiepatti (2004) proposed that descending commands modulate spinal oscillators to adjust phase relationships between segments without disrupting the basic locomotor rhythm. They further argued, based on EMG differences (Courtine \u0026amp; Schieppati, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2003\u003c/span\u003e), that corticospinal drive enables inner\u0026ndash;outer limb adjustments without interfering with rhythm or gait organisation. Two decades later, advances in muscle activity modelling using NNMF, amongst others, provide necessary tools to try and explain the modular organisation of muscle synergies more directly. Our results confirm these observations: all synergies showed systematically modified spatial patterns, while temporal patterns remained more robustly conserved during curved running. Adaptations to the latter were subtler but evident in amplitude and timing, likely reflecting both feedforward and sensory feedback adjustments at critical points in the step cycle. These findings provide quantitative support for the view that locomotor control integrates robust spinal rhythm generators with flexible, limb-specific modulation to accommodate for situations such as changes in trajectory.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003eLimitations\u003c/h2\u003e \u003cp\u003eA limitation of this study is that the SLR condition was done on a treadmill and not overground. Thus, despite most outcomes between treadmill and overground running being largely comparable (Van Hooren et al., \u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), it does add some bias to our results.\u003c/p\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003eData is available under \u003cstrong\u003ehttps://doi.org/10.17605/OSF.IO/UGJDC\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCode can be made available upon reasonable request.\u003c/p\u003e\n\u003ch3\u003eAuthor Contributions\u003c/h3\u003e\n\u003cp\u003e\u0026bull;\u0026nbsp;Conceptualization: [Raphael Mesquita, Patrick Willems Arthur Dewolf]; Methodology: [Raphael Mesquita, Arthur Dewolf]; Formal analysis and investigation: [Raphael Mesquita, Arthur Dewolf]; Writing - original draft preparation: [Raphael Mesquita, Thibaut Toussaint, Patrick Willems, Arthur Dewolf]; Writing - review and editing[Raphael Mesquita, Thibaut Toussaint, Patrick Willems, Arthur Dewolf]; Funding acquisition: [Patrick Willems]; Supervision: [Arthur Dewolf]\u003c/p\u003e\n\u003ch2\u003eFunding\u003c/h2\u003e\n\u003cp\u003eThis study was funded by the Fonds de la Recherche Scientifique (F.N.R.S - CDR 40013847).\u003c/p\u003e\n\u003cp\u003eThe authors have no relevant financial or non-financial interests to disclose.\u003c/p\u003e\n\u003cp\u003eInformed consent was obtained from all individual participants included in the study.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eaf Klint R, Mazzaro N, Nielsen JB, Sinkjaer T, Grey MJ (2010) Load Rather Than Length Sensitive Feedback Contributes to Soleus Muscle Activity During Human Treadmill Walking. J Neurophysiol 103(5):2747\u0026ndash;2756. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1152/jn.00547.2009\u003c/span\u003e\u003cspan address=\"10.1152/jn.00547.2009\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlexander RM (2002) Stability and Manoeuvrability of Terrestrial Vertebrates. Integr Comp Biol 42(1):158\u0026ndash;164. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/icb/42.1.158\u003c/span\u003e\u003cspan address=\"10.1093/icb/42.1.158\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAndreella A, De Santis R, Vesely A, Finos L (2023) Procrustes-based distances for exploring between-matrices similarity. Stat Methods Appl 32(3):867\u0026ndash;882. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10260-023-00689-y\u003c/span\u003e\u003cspan address=\"10.1007/s10260-023-00689-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBianchi L, Angelini D, Orani GP, Lacquaniti F (1998) Kinematic Coordination in Human Gait: Relation to Mechanical Energy Cost. J Neurophysiol 79(4):2155\u0026ndash;2170. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1152/jn.1998.79.4.2155\u003c/span\u003e\u003cspan address=\"10.1152/jn.1998.79.4.2155\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBorghese NA, Bianchi L, Lacquaniti F (1996) Kinematic determinants of human locomotion. J Physiol 494(3):863\u0026ndash;879. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1113/jphysiol.1996.sp021539\u003c/span\u003e\u003cspan address=\"10.1113/jphysiol.1996.sp021539\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCappellini G, Ivanenko YP, Poppele RE, Lacquaniti F (2006) Motor Patterns in Human Walking and Running. J Neurophysiol 95(6):3426\u0026ndash;3437. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1152/jn.00081.2006\u003c/span\u003e\u003cspan address=\"10.1152/jn.00081.2006\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCatavitello G, Ivanenko Y, Lacquaniti F (2018) A kinematic synergy for terrestrial locomotion shared by mammals and birds. eLife 7:e38190. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.7554/eLife.38190\u003c/span\u003e\u003cspan address=\"10.7554/eLife.38190\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCheung VCK, Cheung BMF, Zhang JH, Chan ZYS, Ha SCW, Chen C-Y, Cheung RTH (2020) Plasticity of muscle synergies through fractionation and merging during development and training of human runners. Nat Commun 11(1):4356. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41467-020-18210-4\u003c/span\u003e\u003cspan address=\"10.1038/s41467-020-18210-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCheung VCK, d\u0026rsquo;Avella A, Tresch MC, Bizzi E (2005) Central and Sensory Contributions to the Activation and Organization of Muscle Synergies during Natural Motor Behaviors. J Neurosci 25(27):6419\u0026ndash;6434. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1523/JNEUROSCI.4904-04.2005\u003c/span\u003e\u003cspan address=\"10.1523/JNEUROSCI.4904-04.2005\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCheung VCK, Piron L, Agostini M, Silvoni S, Turolla A, Bizzi E (2009) Stability of muscle synergies for voluntary actions after cortical stroke in humans. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e, \u003cem\u003e106\u003c/em\u003e(46), 19563\u0026ndash;19568. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1073/pnas.0910114106\u003c/span\u003e\u003cspan address=\"10.1073/pnas.0910114106\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChvatal SA, Ting LH (2013) Common muscle synergies for balance and walking. \u003cem\u003eFrontiers in Computational Neuroscience\u003c/em\u003e, \u003cem\u003e7\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fncom.2013.00048\u003c/span\u003e\u003cspan address=\"10.3389/fncom.2013.00048\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCourtine G, Schieppati M (2003) Human walking along a curved path. II. Gait features and EMG patterns. Eur J Neurosci 18(1):191\u0026ndash;205. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1046/j.1460-9568.2003.02737.x\u003c/span\u003e\u003cspan address=\"10.1046/j.1460-9568.2003.02737.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCourtine G, Schieppati M (2004) Tuning of a Basic Coordination Pattern Constructs Straight-Ahead and Curved Walking in Humans. J Neurophysiol 91(4):1524\u0026ndash;1535. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1152/jn.00817.2003\u003c/span\u003e\u003cspan address=\"10.1152/jn.00817.2003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ed\u0026rsquo;Avella A, Saltiel P, Bizzi E (2003) Combinations of muscle synergies in the construction of a natural motor behavior. Nat Neurosci 6(3):300\u0026ndash;308. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/nn1010\u003c/span\u003e\u003cspan address=\"10.1038/nn1010\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDaffertshofer A, Lamoth CJC, Meijer OG, Beek PJ (2004) PCA in studying coordination and variability: A tutorial. Clin Biomech Elsevier Ltd 19(4):415\u0026ndash;428. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.clinbiomech.2004.01.005\u003c/span\u003e\u003cspan address=\"10.1016/j.clinbiomech.2004.01.005\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDaley MA, Felix G, Biewener AA (2007) Running stability is enhanced by a proximo-distal gradient in joint neuromechanical control. J Exp Biol\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDanner SM, Hofstoetter US, Freundl B, Binder H, Mayr W, Rattay F, Minassian K (2015) Human spinal locomotor control is based on flexibly organized burst generators. Brain 138(3):577\u0026ndash;588. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/brain/awu372\u003c/span\u003e\u003cspan address=\"10.1093/brain/awu372\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDewolf AH, Ivanenko YP, Zelik KE, Lacquaniti F, Willems PA (2019) Differential activation of lumbar and sacral motor pools during walking at different speeds and slopes. J Neurophysiol 122(2):872\u0026ndash;887. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1152/jn.00167.2019\u003c/span\u003e\u003cspan address=\"10.1152/jn.00167.2019\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDewolf AH, Ivanenko Y, Zelik KE, Lacquaniti F, Willems PA (2018) Kinematic patterns while walking on a slope at different speeds. J Appl Physiol 125(2):642\u0026ndash;653. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1152/japplphysiol.01020.2017\u003c/span\u003e\u003cspan address=\"10.1152/japplphysiol.01020.2017\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDewolf AH, La Scaleia V, Fabiano A, Sylos-Labini F, Mondi V, Picone S, Di Paolo A, Paolillo P, Ivanenko Y, Lacquaniti F (2022) Left\u0026ndash;Right Locomotor Coordination in Human Neonates. J Neurosci 42(34):6566\u0026ndash;6580. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1523/JNEUROSCI.0612-22.2022\u003c/span\u003e\u003cspan address=\"10.1523/JNEUROSCI.0612-22.2022\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDewolf AH, Mesquita RM, Willems PA (2020) Intra-limb and muscular coordination during walking on slopes. Eur J Appl Physiol 120(8):1841\u0026ndash;1854. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00421-020-04415-4\u003c/span\u003e\u003cspan address=\"10.1007/s00421-020-04415-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDewolf AH, Sylos-Labini F, Cappellini G, Lacquaniti F, Ivanenko Y (2020) Emergence of Different Gaits in Infancy: Relationship Between Developing Neural Circuitries and Changing Biomechanics. Front Bioeng Biotechnol 8:473. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fbioe.2020.00473\u003c/span\u003e\u003cspan address=\"10.3389/fbioe.2020.00473\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDominici N, Ivanenko YP, Cappellini G, d\u0026rsquo;Avella A, Mond\u0026igrave; V, Cicchese M, Fabiano A, Silei T, Di Paolo A, Giannini C, Poppele RE, Lacquaniti F (2011) Locomotor Primitives in Newborn Babies and Their Development. Science 334(6058):997\u0026ndash;999. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1126/science.1210617\u003c/span\u003e\u003cspan address=\"10.1126/science.1210617\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDonelan JM, Pearson KG (2004) Contribution of Force Feedback to Ankle Extensor Activity in Decerebrate Walking Cats. J Neurophysiol 92(4):2093\u0026ndash;2104. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1152/jn.00325.2004\u003c/span\u003e\u003cspan address=\"10.1152/jn.00325.2004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDzeladini F, Van Den Kieboom J, Ijspeert A (2014) The contribution of a central pattern generator in a reflex-based neuromuscular model. \u003cem\u003eFrontiers in Human Neuroscience\u003c/em\u003e, \u003cem\u003e8\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fnhum.2014.00371\u003c/span\u003e\u003cspan address=\"10.3389/fnhum.2014.00371\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEkeberg \u0026Ouml;, Pearson K (2005) Computer Simulation of Stepping in the Hind Legs of the Cat: An Examination of Mechanisms Regulating the Stance-to-Swing Transition. J Neurophysiol 94(6):4256\u0026ndash;4268. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1152/jn.00065.2005\u003c/span\u003e\u003cspan address=\"10.1152/jn.00065.2005\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFujiki S, Aoi S, Funato T, Sato Y, Tsuchiya K, Yanagihara D (2018) Adaptive hindlimb split-belt treadmill walking in rats by controlling basic muscle activation patterns via phase resetting. Sci Rep 8(1):17341. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-018-35714-8\u003c/span\u003e\u003cspan address=\"10.1038/s41598-018-35714-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGneiting T, Schlather M (2004) Stochastic Models That Separate Fractal Dimension and the Hurst Effect. SIAM Rev 46(2):269\u0026ndash;282. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1137/S0036144501394387\u003c/span\u003e\u003cspan address=\"10.1137/S0036144501394387\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGollhofer A, Kyr\u0026ouml;l\u0026auml;inen H (1991) Neuromuscular Control of the Human Leg Extensor Muscles in Jump Exercises Under Various Stretch-Load Conditions. Int J Sports Med 12(01):34\u0026ndash;40. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1055/s-2007-1024652\u003c/span\u003e\u003cspan address=\"10.1055/s-2007-1024652\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrillner S (1985) Neurobiological Bases of Rhythmic Motor Acts in Vertebrates. Science 228(4696):143\u0026ndash;149. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1126/science.3975635\u003c/span\u003e\u003cspan address=\"10.1126/science.3975635\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrillner S, Manira E, A (2020) Current Principles of Motor Control, with Special Reference to Vertebrate Locomotion. Physiol Rev 100(1):271\u0026ndash;320. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1152/physrev.00015.2019\u003c/span\u003e\u003cspan address=\"10.1152/physrev.00015.2019\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHiguchi T (1988) Approach to an irregular time series on the basis of the fractal theory. Physica D 31(2):277\u0026ndash;283. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/0167-2789(88)90081-4\u003c/span\u003e\u003cspan address=\"10.1016/0167-2789(88)90081-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIjspeert AJ, Daley MA (2023) Integration of feedforward and feedback control in the neuromechanics of vertebrate locomotion: A review of experimental, simulation and robotic studies. J Exp Biol 226(15):jeb245784. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1242/jeb.245784\u003c/span\u003e\u003cspan address=\"10.1242/jeb.245784\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIvanenko YP (2005) Coordination of Locomotion with Voluntary Movements in Humans. J Neurosci 25(31):7238\u0026ndash;7253. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1523/JNEUROSCI.1327-05.2005\u003c/span\u003e\u003cspan address=\"10.1523/JNEUROSCI.1327-05.2005\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIvanenko YP, Cappellini G, Dominici N, Poppele RE, Lacquaniti F (2007) Modular Control of Limb Movements during Human Locomotion. J Neurosci 27(41):11149\u0026ndash;11161. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1523/JNEUROSCI.2644-07.2007\u003c/span\u003e\u003cspan address=\"10.1523/JNEUROSCI.2644-07.2007\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIvanenko YP, d\u0026rsquo;Avella A, Poppele RE, Lacquaniti F (2008) On the Origin of Planar Covariation of Elevation Angles During Human Locomotion. J Neurophysiol 99(4):1890\u0026ndash;1898. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1152/jn.01308.2007\u003c/span\u003e\u003cspan address=\"10.1152/jn.01308.2007\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIvanenko YP, Poppele RE, Lacquaniti F (2004) Five basic muscle activation patterns account for muscle activity during human locomotion: Basic muscle activation patterns. J Physiol 556(1):267\u0026ndash;282. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1113/jphysiol.2003.057174\u003c/span\u003e\u003cspan address=\"10.1113/jphysiol.2003.057174\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIvanenko YP, Poppele RE, Lacquaniti F (2006a) Motor Control Programs and Walking. Neuroscientist 12(4):339\u0026ndash;348. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/1073858406287987\u003c/span\u003e\u003cspan address=\"10.1177/1073858406287987\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIvanenko YP, Poppele RE, Lacquaniti F (2006b) Spinal Cord Maps of Spatiotemporal Alpha-Motoneuron Activation in Humans Walking at Different Speeds. J Neurophysiol 95(2):602\u0026ndash;618. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1152/jn.00767.2005\u003c/span\u003e\u003cspan address=\"10.1152/jn.00767.2005\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKesić S, Spasić SZ (2016) Application of Higuchi\u0026rsquo;s fractal dimension from basic to clinical neurophysiology: A review. Comput Methods Programs Biomed 133:55\u0026ndash;70. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.cmpb.2016.05.014\u003c/span\u003e\u003cspan address=\"10.1016/j.cmpb.2016.05.014\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKlous M, Mikulic P, Latash ML (2011) Two aspects of feedforward postural control: Anticipatory postural adjustments and anticipatory synergy adjustments. J Neurophysiol 105(5):2275\u0026ndash;2288. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1152/jn.00665.2010\u003c/span\u003e\u003cspan address=\"10.1152/jn.00665.2010\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKrishnan V, Latash ML, Aruin AS (2012) Early and late components of feed-forward postural adjustments to predictable perturbations. Clin Neurophysiol 123(5):1016\u0026ndash;1026. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.clinph.2011.09.014\u003c/span\u003e\u003cspan address=\"10.1016/j.clinph.2011.09.014\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKyr\u0026ouml;l\u0026auml;inen H, Avela J, Komi PV (2005) Changes in muscle activity with increasing running speed. J Sports Sci 23(10):1101\u0026ndash;1109. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/02640410400021575\u003c/span\u003e\u003cspan address=\"10.1080/02640410400021575\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLa Scaleia V, Ivanenko YP, Zelik KE, Lacquaniti F (2014) Spinal motor outputs during step-to-step transitions of diverse human gaits. \u003cem\u003eFrontiers in Human Neuroscience\u003c/em\u003e, \u003cem\u003e8\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fnhum.2014.00305\u003c/span\u003e\u003cspan address=\"10.3389/fnhum.2014.00305\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLacquaniti F (2002) Kinematic Control of Walking. Arch Ital Biol 140:263\u0026ndash;272\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLacquaniti F, Grasso R, Zago M (1999) Motor Patterns in Walking. Physiology 14(4):168\u0026ndash;174. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1152/physiologyonline.1999.14.4.168\u003c/span\u003e\u003cspan address=\"10.1152/physiologyonline.1999.14.4.168\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLacquaniti F, Ivanenko YP, Zago M (2012) Patterned control of human locomotion: Control of human locomotion. J Physiol 590(10):2189\u0026ndash;2199. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1113/jphysiol.2011.215137\u003c/span\u003e\u003cspan address=\"10.1113/jphysiol.2011.215137\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee DD, Seung HS (1999) Learning the parts of objects by non-negative matrix factorization. Nature 401(6755):788\u0026ndash;791. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/44565\u003c/span\u003e\u003cspan address=\"10.1038/44565\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLinssen WHJP, Stegeman DF, Joosten EMG, Notermans SLH, Van \u0026rsquo;T, Hof MA, Binkhorst RA (1993) Variability and interrelationships of surface EMG parameters during local muscle fatigue. Muscle Nerve 16(8):849\u0026ndash;856. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/mus.880160808\u003c/span\u003e\u003cspan address=\"10.1002/mus.880160808\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMacLellan MJ, Ivanenko YP, Massaad F, Bruijn SM, Duysens J, Lacquaniti F (2014) Muscle activation patterns are bilaterally linked during split-belt treadmill walking in humans. J Neurophysiol 111(8):1541\u0026ndash;1552. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1152/jn.00437.2013\u003c/span\u003e\u003cspan address=\"10.1152/jn.00437.2013\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMartino G, Ivanenko YP, d\u0026rsquo;Avella A, Serrao M, Ranavolo A, Draicchio F, Cappellini G, Casali C, Lacquaniti F (2015) Neuromuscular adjustments of gait associated with unstable conditions. J Neurophysiol 114(5):2867\u0026ndash;2882. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1152/jn.00029.2015\u003c/span\u003e\u003cspan address=\"10.1152/jn.00029.2015\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMartino G, Ivanenko YP, Serrao M, Ranavolo A, d\u0026rsquo;Avella A, Draicchio F, Conte C, Casali C, Lacquaniti F (2014) Locomotor patterns in cerebellar ataxia. J Neurophysiol 112(11):2810\u0026ndash;2821. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1152/jn.00275.2014\u003c/span\u003e\u003cspan address=\"10.1152/jn.00275.2014\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMesquita RM, Catavitello G, Willems PA, Dewolf AH (2023) Modification of the locomotor pattern when deviating from the characteristic heel-to-toe rolling pattern during walking. Eur J Appl Physiol. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00421-023-05169-5\u003c/span\u003e\u003cspan address=\"10.1007/s00421-023-05169-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMesquita RM, Willems PA, Catavitello G, Dewolf AH (2024) Kinematics and mechanical changes with step frequency at different running speeds. Eur J Appl Physiol 124(2):607\u0026ndash;622. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00421-023-05303-3\u003c/span\u003e\u003cspan address=\"10.1007/s00421-023-05303-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMesquita RM, Willems PA, Dewolf AH, Catavitello G (2024) Kinetics and mechanical work done to move the body centre of mass along a curve. PLoS ONE 19(2):21\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMinassian K, Hofstoetter US, Dzeladini F, Guertin PA, Ijspeert A (2017) The Human Central Pattern Generator for Locomotion: Does It Exist and Contribute to Walking? Neuroscientist 23(6):649\u0026ndash;663. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/1073858417699790\u003c/span\u003e\u003cspan address=\"10.1177/1073858417699790\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNishikawa K, Biewener AA, Aerts P, Ahn AN, Chiel HJ, Daley MA, Daniel TL, Full RJ, Hale ME, Hedrick TL, Lappin AK, Nichols TR, Quinn RD, Satterlie RA, Szymik B (2007) Neuromechanics: An integrative approach for understanding motor control. Integr Comp Biol 47(1):16\u0026ndash;54. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/icb/icm024\u003c/span\u003e\u003cspan address=\"10.1093/icb/icm024\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOgawa T, Kawashima N, Suzuki S, Nakazawa K (2012) Asymmetrical Neural Adaptation in Lower Leg Muscles as a Consequence of Stereotypical Motor Training. J Mot Behav 44(1):63\u0026ndash;68. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/00222895.2011.645912\u003c/span\u003e\u003cspan address=\"10.1080/00222895.2011.645912\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePearson KG (2008) Role of sensory feedback in the control of stance duration in walking cats. Brain Res Rev 57(1):222\u0026ndash;227. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.brainresrev.2007.06.014\u003c/span\u003e\u003cspan address=\"10.1016/j.brainresrev.2007.06.014\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePoppele R, Bosco G (2003) Sophisticated spinal contributions to motor control. Trends Neurosci 26(5):269\u0026ndash;276. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/S0166-2236(03)00073-0\u003c/span\u003e\u003cspan address=\"10.1016/S0166-2236(03)00073-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQiao M, Abbas JJ, Jindrich DL (2017) A model for differential leg joint function during human running. Bioinspir Biomim 12(1):016015. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1088/1748-3190/aa50b0\u003c/span\u003e\u003cspan address=\"10.1088/1748-3190/aa50b0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSantuz A, Br\u0026uuml;ll L, Ekizos A, Schroll A, Eckardt N, Kibele A, Schwenk M, Arampatzis A (2020) Neuromotor Dynamics of Human Locomotion in Challenging Settings. iScience 23(1):100796. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.isci.2019.100796\u003c/span\u003e\u003cspan address=\"10.1016/j.isci.2019.100796\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSantuz A, Ekizos A, Eckardt N, Kibele A, Arampatzis A (2018) Challenging human locomotion: Stability and modular organisation in unsteady conditions. Sci Rep 8(1):2740. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-018-21018-4\u003c/span\u003e\u003cspan address=\"10.1038/s41598-018-21018-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSantuz A, Ekizos A, Janshen L, Mersmann F, Bohm S, Baltzopoulos V, Arampatzis A (2018) Modular Control of Human Movement During Running: An Open Access Data Set. Front Physiol 9:1509. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fphys.2018.01509\u003c/span\u003e\u003cspan address=\"10.3389/fphys.2018.01509\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSantuz A, Ekizos A, Kunimasa Y, Kijima K, Ishikawa M, Arampatzis A (2020) Lower complexity of motor primitives ensures robust control of high-speed human locomotion. Heliyon 6(10):e05377. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.heliyon.2020.e05377\u003c/span\u003e\u003cspan address=\"10.1016/j.heliyon.2020.e05377\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSartori M, Gizzi L, Lloyd DG, Farina D (2013) A musculoskeletal model of human locomotion driven by a low dimensional set of impulsive excitation primitives. \u003cem\u003eFrontiers in Computational Neuroscience\u003c/em\u003e, \u003cem\u003e7\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fncom.2013.00079\u003c/span\u003e\u003cspan address=\"10.3389/fncom.2013.00079\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSelionov VA, Ivanenko YP, Solopova IA, Gurfinkel VS (2009) Tonic Central and Sensory Stimuli Facilitate Involuntary Air-Stepping in Humans. J Neurophysiol 101(6):2847\u0026ndash;2858. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1152/jn.90895.2008\u003c/span\u003e\u003cspan address=\"10.1152/jn.90895.2008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSylos-Labini F, La Scaleia V, Cappellini G, Dewolf A, Fabiano A, Solopova IA, Mond\u0026igrave; V, Ivanenko Y, Lacquaniti F (2022) Complexity of modular neuromuscular control increases and variability decreases during human locomotor development. Commun Biology 5(1):1256. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s42003-022-04225-8\u003c/span\u003e\u003cspan address=\"10.1038/s42003-022-04225-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTorres-Oviedo G, Macpherson JM, Ting LH (2006) Muscle Synergy Organization Is Robust Across a Variety of Postural Perturbations. J Neurophysiol 96(3):1530\u0026ndash;1546. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1152/jn.00810.2005\u003c/span\u003e\u003cspan address=\"10.1152/jn.00810.2005\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVan Hooren B, Fuller JT, Buckley JD, Miller JR, Sewell K, Rao G, Barton C, Bishop C, Willy RW (2020) Is Motorized Treadmill Running Biomechanically Comparable to Overground Running? A Systematic Review and Meta-Analysis of Cross-Over Studies. Sports Med 50(4):785\u0026ndash;813. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s40279-019-01237-z\u003c/span\u003e\u003cspan address=\"10.1007/s40279-019-01237-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWinter DA (2009) Biomechanics and motor control of human movement, 4th edn. Wiley\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYokoyama H, Ogawa T, Kawashima N, Shinya M, Nakazawa K (2016) Distinct sets of locomotor modules control the speed and modes of human locomotion. Sci Rep 6(1):36275. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/srep36275\u003c/span\u003e\u003cspan address=\"10.1038/srep36275\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYokoyama H, Sato K, Ogawa T, Yamamoto S-I, Nakazawa K, Kawashima N (2018) Characteristics of the gait adaptation process due to split-belt treadmill walking under a wide range of right-left speed ratios in humans. PLoS ONE 13(4):e0194875. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1371/journal.pone.0194875\u003c/span\u003e\u003cspan address=\"10.1371/journal.pone.0194875\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZaidi KF, Harris-Love M (2023) A Novel Procrustes Analysis Method to Quantify Multi-Joint Coordination of the Upper Extremity after Stroke. \u003cem\u003e2023 45th Annual International Conference of the IEEE Engineering in Medicine \u0026amp; Biology Society (EMBC)\u003c/em\u003e, 1\u0026ndash;4. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/EMBC40787.2023.10341023\u003c/span\u003e\u003cspan address=\"10.1109/EMBC40787.2023.10341023\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Curved running, Muscle Synergies, Muscular Coordination, Kinematic Coordination","lastPublishedDoi":"10.21203/rs.3.rs-8038490/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8038490/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eCurved running imposes asymmetrical mechanical demands on the lower-limbs and thus provides a model to test the adaptability of the locomotor system. This study investigated how muscle synergies reorganise during curved versus straight-line running. Using non-negative matrix factorisation of EMG signals, four synergies were extracted from both inner and outer limbs. While the overall modular structure was preserved, spatial patterns reorganised systematically while temporal patterns showed earlier onsets, particularly in synergies associated with push-off and late swing. Adaptations were limb-specific: the inner limb displayed greater reweighting and reduced complexity, while the outer limb showed more anticipatory shifts. Higuchi\u0026rsquo;s fractal dimension indicated reduced complexity in touchdown and late swing synergies but increased complexity in push-off, suggesting differential demands for robust versus finely tuned control. To corroborate this, kinematic analyses confirmed that curved running modified intersegmental coordination, with divergence of the covariation plane between inner and outer limbs, reflecting their distinct functional roles in redirecting versus propelling the body. Together, these findings indicate that curved running use modular locomotor control strategies which combine robust rhythm-generating spinal networks with asymmetric, limb-specific modulation. This coordination likely arises from the interaction of feedforward CPG activity with feedback-driven adjustments at mechanically critical phases of the gait cycle, providing new quantitative evidence for the flexible yet stable organisation of human locomotion.\u003c/p\u003e","manuscriptTitle":"Limb-Specific Modulation of Muscle Synergies and Segmental Coordination During Curved Running","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-09 07:18:41","doi":"10.21203/rs.3.rs-8038490/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"0b9b3de4-d3dc-4627-b51c-031711c393cd","owner":[],"postedDate":"January 9th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-01-09T07:18:41+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-09 07:18:41","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8038490","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8038490","identity":"rs-8038490","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}