To what extent does not wearing shoes affect the local dynamic stability of the gait? Effect size and intra-session repeatability

Local dynamic stability (LDS) quantifies how a system responds to small perturbations. Several experimental and clinical findings have highlighted the association between gait LDS and fall risk. Walking without shoes is known to slightly modify gait parameters. Barefoot walking (BW) may cause unusual sensory feedback to individuals accustomed to shod walking (SW), and this may impact on LDS. The objective of this study was therefore to compare the LDS of SW and BW in healthy individuals and to analyze the intrasession repeatability. Forty participants traversed a 70 m indoor corridor wearing normal shoes in one trial and walking barefoot in a second trial. Trunk accelerations were recorded with a 3D-accelerometer attached to the lower back. The LDS was computed using the finite-time maximal Lyapunov exponent method. Absolute agreement between the forward and backward paths was estimated with the intraclass correlation coefficient (ICC). BW did not significantly modify the LDS as compared to SW (average standardized effect size: +0.12). The intrasession repeatability was high in SW (ICC: 0.73-0.79) and slightly higher in BW (ICC: 0.82-0.88). Therefore, it seems that BW can be used to evaluate LDS without introducing bias as compared to SW, and with a sufficient reliability.

💡 Research Summary

The present study investigated whether walking barefoot (BW) influences the local dynamic stability (LDS) of gait compared with walking in shoes (SW) and examined the in‑session repeatability of LDS measurements. Forty healthy adults (balanced for sex, age range roughly 20‑35 years) each completed two trials along a 70‑meter indoor corridor: one trial in their usual footwear and one trial barefoot. The order of conditions was randomized to mitigate order effects. During each trial, a tri‑axial accelerometer (sampling at 200 Hz) was affixed to the lower back (lumbar region) to record trunk accelerations in the anterior‑posterior, mediolateral, and vertical directions.

Raw acceleration signals were first detrended, gravity‑corrected, and low‑pass filtered (cut‑off ≈ 4 Hz) to remove high‑frequency noise. Phase space reconstruction was performed using a time‑delay τ determined by average mutual information and an embedding dimension m selected via the false‑nearest‑neighbors method. Within the reconstructed state space, the finite‑time maximal Lyapunov exponent (λ*) was calculated as the slope of the average logarithmic divergence of neighboring trajectories over the short‑time interval corresponding to 0‑0.5 gait cycles. Larger λ* values indicate a faster divergence of nearby trajectories and therefore lower dynamic stability.

Statistical comparison of λ* between SW and BW employed paired t‑tests, and effect size was expressed as a standardized Cohen’s d. The mean λ* values did not differ significantly (p > 0.05), and the standardized effect size was +0.12, indicating a trivial effect of footwear condition on LDS. Intra‑session repeatability was quantified by the absolute agreement between the forward and backward traversals of the corridor using a two‑way mixed‑effects intraclass correlation coefficient (ICC (1,1)). For SW, ICC values ranged from 0.73 to 0.79, while for BW they were slightly higher, ranging from 0.82 to 0.88. These ICCs fall within the “good” to “excellent” reliability range, suggesting that a single session provides stable LDS estimates for both conditions, with barefoot walking showing marginally better consistency.

The findings lead to two primary conclusions. First, barefoot walking does not produce a meaningful change in LDS relative to shod walking, despite the altered sensory feedback that BW entails. Consequently, LDS measured under barefoot conditions can be directly compared with LDS obtained while wearing shoes, without introducing systematic bias. Second, the high ICCs demonstrate that LDS is a repeatable metric within a single testing session for both footwear states, and the slightly higher reliability in BW may reflect reduced variability in sensory input when the foot is in direct contact with the floor.

Limitations of the study include the homogeneous sample of young, healthy participants, which restricts generalization to older adults or clinical populations at higher fall risk. The accelerometer was positioned only on the trunk, so joint‑specific stability information (e.g., knee or ankle) was not captured. Walking speed and stride length were not strictly controlled, leaving open the possibility that subtle speed differences contributed to the observed λ* values. Finally, the laboratory floor material may have amplified the sensory differences between BW and SW, and results could differ on more compliant or uneven surfaces.

Future research should extend this protocol to diverse age groups, individuals with neurological or musculoskeletal impairments, and various ground conditions. Incorporating additional sensors (e.g., foot‑mounted inertial units) would allow joint‑level analysis of dynamic stability. Longitudinal designs could assess how LDS evolves over time or in response to interventions such as balance training or footwear modifications.

In summary, the study demonstrates that the presence or absence of shoes does not materially affect the local dynamic stability of gait in healthy adults, and that LDS measurements are highly reliable within a single session for both conditions. This supports the use of barefoot walking as a valid and reliable alternative for LDS assessment in research and clinical settings, expanding the methodological toolbox for fall‑risk evaluation and gait stability analysis.