1. Introduction
Gait is a complex locomotor behavior driven by the coupled interactions of the nervous, muscular, skeletal, and cardiorespiratory systems (Sutherland, 1997). In the early phase of independent walking, toddlers exhibit immature gait features such as a wide base of support, high cadence, prolonged double support, and flat-foot or unstable heel strike (Liu et al., 2022; Van de Walle et al., 2018); with increasing walking experience and neural maturation, spatiotemporal and kinematic parameters improve rapidly—marked by reductions in double-support duration and gait variability, narrowing of the base of support, and a shift toward heel-to-forefoot rollover—while energy economy, interlimb symmetry, and fine coordination continue to develop through the school years, ultimately yielding an efficient and stable adult gait pattern (Bach et al., 2021; Douglas et al., 2022; Hospodar & Adolph, 2024; Liu et al., 2022; Van de Walle et al., 2018; Van Hamme et al., 2015; Yoshimoto et al., 2023).
In healthy adolescents, mature gait is typically characterized by a high degree of bilateral consistency, reflected in the synchronized timing of joint kinematics and kinetics (angles, angular velocities, moments, and powers). However, minor left–right discrepancies are frequently observed in real-world movement, and their extent varies with both the specific parameter assessed and the walking or running speed (Bergamini et al., 2024; Van Der Velde et al., 2025; Vincent et al., 2025). When amplified or accumulated over time, such differences may increase metabolic cost and alter control strategies. At the same time, recent large-scale runner data indicate that not all asymmetry metrics are directly associated with a higher injury risk, underscoring the need for task- and metric-specific interpretation (Grimmitt et al., 2025; Malisoux et al., 2024; McAllister et al., 2025).
Compared with straight walking or running, turning tasks require rapid reorientation of the center of mass (COM), swift support exchange, and redistribution of plantar loading within very short time windows, imposing greater demands on online motor control. Compared with straight-line gait, turning more readily induces asymmetry in the temporal coordination of joint motion as well as in plantar loading patterns, with the magnitude influenced by movement strategy and foot-strike characteristics (Kawabata et al., 2025; Lautzenheiser & Kramer, 2025; Yi et al., 2024; Zhang et al., 2024). Accordingly, turning is regarded as a sensitive context for probing dynamic stability and interlimb coordination, capable of revealing latent control bottlenecks in healthy individuals under increased task demands (Earhart, 2013; Netukova et al., 2024; Reimann & Bruijn, 2024).
Considering the background, the present study systematically compares left–right differences in lower-limb kinematics and kinetics across walking, running, running-with-turning, and walking-with-turning in healthy adolescents, and identifies the key temporal windows of asymmetry. We hypothesize that: (1) both repetitive (straight-line) and non-repetitive (turning) tasks exhibit significant bilateral asymmetries; (2) asymmetry is more pronounced in non-repetitive turning tasks; and (3) asymmetry shows phase-specific distributions across joints and kinematic/kinetic variables, reflecting key mechanisms of lower-limb motor control. These results provide biomechanical evidence for understanding gait control in healthy populations, optimizing performance, and informing asymmetry surveillance.
2. Methods
2.1 Participant
Fifteen physically active adolescent males participated in the study (height: 160.62 ± 6.13 cm; age: 15.4 ± 1.05 years; body mass: 50.82 ± 5.28 kg). All participants regularly engaged in sports activities no fewer than 3 times per week, and each session lasting at least one hour. Limb dominance was determined according to the preferred leg used for kicking, and all subjects were classified as right-leg dominant (Chapman et al., 1987). Participants reported no history of surgery or musculoskeletal injury within the previous six months and had no known lower-limb disorders that could influence normal gait performance. All experimental procedures followed the ethical Declaration of Helsinki principles. Before participation, both the participants and their legal guardians received detailed information about the study procedures, and written consent was provided. This study received approval from the Ethics Committee of Ningbo University (Protocol No. TY2024022).
2.2 Biomechanics Parameters Collection
Three-dimensional motion data were captured by Vicon motion analysis system (Oxford Metrics, UK) composed of eight infrared cameras and Nexus software (Xu et al., 2022; Zhou et al., 2023). GRFs were measured by Kistler force plate system (Winterthur, Switzerland) (Xu et al., 2022). Motion and force measurements were recorded in a synchronized manner through the Vicon Nexus interface. Figure 1C shows the marker locations (39 reflective markers) and the corresponding experimental setup (Xu et al., 2024).
Figure 1
Experimental procedures for the four locomotor conditions: straight walking, straight running, walking with turning, and running with turning.
Before testing, participants performed a warm-up that included 10 minutes of treadmill running (8 km/h), and a lower-limb stretching exercises. To familiarize themselves with the experimental procedures, each participant completed three practice trials prior to the formal testing session. For the static trial, participants were stand upright on the force platform aligned with the laboratory Y-axis, with their arms folded across the chest and their gaze directed forward until the static calibration data were obtained. Testing setup and task procedures are shown in Figure 1A, 1B, 1D and 1E. Subsequently, participants performed four locomotor conditions at a self-selected speed: straight walking, straight running, walking with turning (45°), and running with turning, during which the stance phase was recorded for analysis (Zhou et al., 2021).
For each condition, seven valid trials were collected, with a one-minute rest period provided between consecutive trials. Initial ground contact was identified when the vertical GRF exceeded 10 N (Zhou & Ugbolue, 2019). To minimize the potential influence of fatigue, the four experimental conditions, straight walking, straight running, walking with turning, and running with turning, were performed on separate testing days, with only one condition completed per day.
2.3 Data processing
Marker trajectory and GRF data were labeled and reconstructed in Vicon Nexus software (version 1.8.6). Subsequent data processing and analysis were conducted in MATLAB R2019a (The MathWorks, Natick, MA, USA), including coordinate trans-formation, signal filtering, parameter extraction, and conversion to the formats re-quired for further analysis.
Data processing was conducted in several steps. First, marker trajectory and GRF signals were converted to the coordinate system used for musculoskeletal simulation. Both trajectory and force data were then filtered using 4-order 0-lag Butter-worth low-pass filters, with cut-off frequencies set at 6 Hz for kinematic data and 30 Hz for force data. Analyses were limited to the stance phase of each locomotor condition (straight walking, straight running, walking with turning, and running with turning). In this research, biomechanical parameters were processed and calculated using OpenSim (Stanford University, Stanford, CA, USA). Muscle and tendon actuators were employed in an OpenSim model (gait 2392) with 10 rigid bodies and 23 degrees of freedom. The following procedures were used to measure muscle activation and muscular force output. Firstly, open the OpenSim 4.2 software and import the static model. The anthropometric model of each participant may then be obtained using the scale tool. Determine the muscle’s beginning and ending sites, and make sure the moment arms are in line with the participants’ limb lengths. Secondly: Create a motion file (mot) by calculating the joint angle throughout the stance phases of running using the inverse kinematics (IK) tool in the OpenSim 4.2 software. Import the running marker files as well as the external force files into OpenSim simultaneously using the inverse dynamics (ID) tool, and then compute the joint kinetics of each participant.
2.4 Statistical Analysis
For the statistical parametric mapping analysis, stance-phase waveforms were extracted for each locomotor condition (straight walking, straight running, walking with turning, and running with turning). Each stance trajectory was time-normalized to 101 samples using a custom MATLAB routine (Zhou et al., 2025). One-dimensional SPM was then performed in SPM1d (joint angle, velocity, moment, power and force), applying independent-samples t-tests based on random field theory to evaluate left–right differences across the normalized stance period (MEI et al., 2021; Pataky, 2012). All analyses were conducted in MATLAB R2019a, significance set at 0.05 (two-tailed).
3. Results
Figure 2 presents the SPM results for left–right comparisons across the running stance phase, demonstrating pronounced inter-limb asymmetry in multiple biomechanical variables. For joint angles, significant differences were observed at the ankle (0–71%, 85–100%, p < 0.03), knee (0–24%, p < 0.02), and hip (0–100%, p < 0.01) over the stance period. Between-limb differences in angular velocity were detected at the ankle (5–75%, 79–100%, p < 0.01), knee (0–61%, 68–100%, p < 0.01), and hip (36–61%, p < 0.01). Analysis of joint moments revealed significant regions at the ankle (18–61%, p < 0.01), knee (0–16%, 77–89%, p < 0.01), and hip (65–91%, p < 0.01). Power profiles also showed asymmetry between limbs at the ankle (3–13%, 45–76%, p < 0.01), knee (0–17%, 66–75%, 94–100%, p < 0.02), and hip (0–7%, 7–17%, 18–55%, 64–78%, 83–97%, p < 0.01). Similarly, significant differences in joint forces were identified at the ankle (20–37%, p < 0.01), knee (17–61%, p < 0.01), and hip (0–72%, 75–100%, p < 0.01).
Figure 2
SPM results for left–right comparisons of ankle, knee, and hip kinematics and kinetics during the running stance phase. Shaded areas denote phases with significant between-limb differences, together with the corresponding t-statistics across participants.
Figure 3 presents the SPM results for left–right comparisons during the running-with-turning stance phase, indicating pronounced inter-limb asymmetry across several biomechanical variables. For joint angles, significant differences were identified at the ankle (18–100%, p < 0.01) and knee (42–100%, p < 0.01). Between-limb differences in angular velocity were observed at the ankle (69–100%, p < 0.01), knee (0–48%, 68–100%, p < 0.01), and hip (9–33%, p < 0.01). Analysis of joint moments showed significant regions at the ankle (3–84%, p < 0.01), knee (0–75%, 82–100%, p < 0.01), and hip (0–11%, 18–30%, 61–100%, p < 0.01). Power profiles also demonstrated asymmetry at the ankle (0–42%, 64–86%, p < 0.01), knee (0–30%, 33–73%,79–100%, p < 0.01), and hip (17–58%, p < 0.01). For joint forces, significant between-limb differences were found at the knee (21–31%, p = 0.03; 86–100%, p = 0.02).
Figure 3
SPM-based left–right comparisons of ankle, knee, and hip kinematics and kinetics during the running-with-turning stance phase. Shaded areas represent significant between-limb differences across participants, with dashed lines indicating the significance level (p = 0.05).
Figure 4 presents the SPM results for left–right comparisons during the walking stance phase, indicating inter-limb asymmetry across multiple biomechanical measures. For joint angles, significant differences were observed at the ankle (0–74%, p < 0.01), knee (0–57%, 64–100%, p < 0.01), and hip (0–100%, p < 0.01). Be-tween-limb differences in angular velocity were identified at the ankle (0–8%, 9–27%, 33–78%, 80–91%, 92–100%, p < 0.01), knee (0–27%, 29–90%, 92–100%, p < 0.01), and hip (3–16%, 20–89%, 92–100%, p < 0.01). Analysis of joint moments revealed significant regions at the ankle (18–50%, p < 0.01), knee (0–30%, 45–96%, p < 0.01), and hip (0–53%, 60–92%, p < 0.01). The figure illustrates statistically significant differences in the ankle (0–5%, 7–31%, 38–81%, 85–100%, p < 0.01), knee (11–28%, 37–53%, 82–91%, p < 0.01), and hip (0–10%, 13–93%, p < 0.01) joint power during the walking stance phase. The figure illustrates statistically significant differences in the ankle (0–18%, 50–65%, p < 0.01), knee (0–16%, 48–69%, p < 0.01), and hip (2–11%, 17–47%, 69–83%, p < 0.03) joint force during the walking stance phase.
Figure 4
SPM results for left–right comparisons of ankle, knee, and hip kinematic and kinetic variables during the walking stance phase. Shaded regions indicate time intervals with significant differences, and dashed lines represent the critical threshold (p = 0.05).
Figure 5 presents the SPM results for left–right comparisons during the walk-ing-with-turning stance phase, demonstrating notable inter-limb asymmetry. For joint angles, significant differences were observed at the ankle (10–100%, p < 0.01), knee (11–76%, p < 0.01), and hip (0–53%, p < 0.01). The figure illustrates statistically significant differences in the ankle (0–19%, 21–86%, 89–100%, p < 0.01), knee (0–14%, 20–34%, 41–63%, 72–100%, p < 0.01), and hip (11–27%, 33–41%, 46–60%, 62–89%, p < 0.02) joint velocity during the turning while walking stance phase. The figure illustrates statistically significant differences in the ankle (8–36%, 82–90%, p < 0.03), knee (0–98%, p < 0.01), and hip (0–99%, p < 0.01) joint moment during the turning while walking stance phase. The figure illustrates statistically significant differences in the ankle (19–64%, 74–92%, p < 0.01), knee (2–9%, 12–60%, 62–100%, p < 0.01), and hip (0–81%, 82–100%, p < 0.01) joint power during the turning while walking stance phase. The figure illustrates statistically significant differences in the ankle (0–10%, 21–37%, p < 0.05), knee (18–42%, p < 0.01;), and hip (1–25%, p = 0.01) joint force during the turning while walking stance phase.
Figure 5
SPM results for left–right comparisons of ankle, knee, and hip kinematic and kinetic variables during the walking with turning stance phase. Shaded regions indicate time intervals with significant differences, and dashed lines represent the critical threshold (p = 0.05).
4. Discussion
This study used SPM to systematically compare, under running and walking conditions (straight-line and turning), left–right differences in lower limbs kinematics and kinetics. The results showed asymmetries across all four conditions, with clear differences in the temporal intervals and joint distributions of these asymmetries. Overall, the results obtained in this study are generally consistent with our hypotheses, running exhibited broader left–right differences than walking, and turning tasks showed greater asymmetry than straight-line tasks.
During straight-line running, left–right asymmetry appeared in the ankle, knee, and hip. Hip angle showed near-continuous asymmetry; other variables showed windowed asymmetries. This suggests a functional division between dominant and non-dominant limbs. Prior studies report that the dominant limb tends to play a greater role in propulsive function, whereas the non-dominant limb more often pro-vides braking and stabilization (Delgado-García et al., 2025; D’Hondt et al., 2024). We did not classify limb dominance in the present study. Ankle asymmetry occurred mainly in early to midstance (Monte et al., 2020; Vial et al., 2023). This phase involves cushioning, energy absorption, and elastic storage. The Achilles ten-don and plantar-flexor unit absorb energy after foot strike. They release it during mid-stance and push-off. Reliance on elastic return has been shown to increase with speed (Kharazi et al., 2021; Monte et al., 2020). We did not manipulate speed. Thus, ankle asymmetry may reflect side-specific tendon energy return.
During the turning run, asymmetry inter-limb differences become more pronounced, particularly at the ankle and knee show significant differences during most of the stance, while the hip shows differences mainly in the early and late phases, with asymmetry also appearing in the mid-phase power. This pattern reflects the greater centripetal (inward-directed) and lateral-control demands of cornering, which amplify inter-limb differences in lower-extremity motion patterns and joint mechanics (Chang & Kram, 2007; Diaz et al., 2024). When turning, the outside leg bears more braking and mediolateral (centripetal) loading, while the inside leg redirects the path and drives propulsion; a smaller turn radius further accentuates these roles and the resulting asymmetry (Pietraszewski et al., 2021). It should be emphasized that the above differentiation reflects task-specific neurome-chanical adaptation and does not necessarily equate to a permanent distinction be-tween the dominant and non-dominant limbs (individual, speed, and geometric conditions can all influence expression) (Ohnuma et al., 2018; Taboga et al., 2016).
When walking straight ahead, although asymmetry can be observed between limbs, the overall amplitude is smaller than when running; healthy adults typically exhibit strong dynamic symmetry at low to moderate speeds (Baček et al., 2025). In straight-line walking, significant windows were prominent at mid-stance, with additional early and late intervals, which aligns with the step-to-step redirection of the body’s COM, where the leading limb performs negative work and the trailing limb provides push-off (Adamczyk & Kuo, 2009; Soo & Donelan, 2010). These differences are also related to frontal plane stability control: strategies such as step width and trunk lateral tilt can regulate hip adduction torque and stability requirements, suggesting that mid-term joint differences may reflect an individual’s fine balance of stabilization-propulsion (Subasinghe Arachchige et al., 2024). Using SPM for whole-segment timing statistics helps reveal the aforementioned asymmetric distribution, which is dominated by the middle period and interspersed with early/late, rather than relying solely on discrete peaks (Mestanza Mattos et al., 2023).
Under turning-walking conditions, left–right differences increase markedly: the ankle, knee, and hip show prolonged asymmetries linked to the in/outside leg and the radius of the turn (Rasmussen et al., 2022). Our pattern resembles turning runs but with smaller magni-tudes at lower speed. Turning gait imposes greater demands for centripetal/lateral control and path redirection, altering muscle output and coordination and thereby disrupting gait symmetry (Lautzenheiser & Kramer, 2025). In addition, curved walking is associated with a higher slip risk, which is influenced by turn radius and the phase at which perturbations occur (Rasmussen et al., 2022, 2024).
This study is limited by a modest sample, a stance-only analysis, and a con-trolled laboratory setup with a single turn geometry and fixed pace; thus, ecological generalizability and scaling with speed/radius cannot be assessed. Future work should include swing-phase analyses, vary speed and turn radius, and test more eco-logical tasks.
5. Conclusions
SPM revealed task-dependent inter-limb asymmetry, which is greater during running than walking and further amplified during turning. In running, hip asymmetry was near-continuous across stance, whereas in turning, ankle and knee asymmetries spanned large portions of stance. These patterns refine our understanding of motor control and may inform injury prevention, rehabilitation, and sport-specific conditioning.
Data Accessibility Statement
Data are available for research purposes upon reasonable request to the corresponding author.
Ethics and Consent
The Ethics Committee at Ningbo University approved the protocol of this study (protocol code: TY2024022), which was conducted in accordance with the Declaration of Helsinki.
Informed consent was obtained from the parents or legal guardian of the children participated in this study.
Author Contributions
Conceptualization: WL, GF. Methodology: WL. Software: WL. Validation: WL. Formal analysis: WL, GF. Investigation: WL, GF. Data curation and validation: WL, GF. Visualization: WL, GF. Project administration and supervision: WL, GF. Writing—original draft preparation: WL, GF. Writing—review and editing: WL, GF. All authors contributed to manuscript revision and approved the final version of the manuscript.