WorthDay Icon
EvidenceHealthFootwear / Barefoot

Barefoot walking is more stable in the gait of balance recovery in older adults

Study type:
Within-subject laboratory experiment with fixed (non-randomised) condition order, n = 14 healthy older adults (68.29 ± 3.41 years), comparing treadmill walking shod vs. barefoot during normal gait and trip-like perturbed recovery gait, measuring four spatiotemporal gait-variability outcomes
Added:
2026-09-11
Status:
published

Strength of evidence

Very Low

Source

Ren X, Kebbach M, Bruhn S, Yang Q, Lin H, Bader R, Tischer T, Lutter C. Barefoot walking is more stable in the gait of balance recovery in older adults. BMC Geriatr. 2022;22:904. doi:10.1186/s12877-022-03628-w

Open original study
Science LogbookAll studies

In a within-subject laboratory experiment, 14 healthy older adults walked on a treadmill shod and barefoot, during both normal gait and trip-like perturbation recovery gait (belt deceleration of 3 m/s² held for 300 ms). Four gait-variability outcomes were calculated from marker trajectories. Only one of the four - swing time variability - showed a footwear effect and a gait-pattern-by-footwear interaction (p = .031 and p = .006), with lower swing time variability in barefoot walking during the first recovery step after a trip. The other three outcomes (stride length, stride time and step width variability) showed no footwear effect at all. No falls occurred, all endpoints were computationally derived variability coefficients rather than observed events, the sample was 14 participants, and the shod condition always ran before the barefoot condition without counterbalancing. Under WorthDay Evidence Criteria this earns very low-quality evidence: the hypothesis is tested experimentally, but exclusively with indirect surrogate endpoints, without randomisation of condition order, without blinding and without a null control.

Key findings

  • Design: within-subject laboratory experiment, two-way repeated-measures ANOVA with factors gait pattern (normal vs. trip-Rec1) and footwear condition (shod vs. barefoot), Tukey post hoc tests.
  • Participants: 14 healthy older adults (age 68.29 ± 3.41 years; body height 1.76 ± 0.10 m; body mass 81.14 ± 14.52 kg). Sex distribution is not reported in the published abstract or indexed record.
  • Protocol: four trials - (1) 3 min shod normal walking, (2) six trip perturbations shod, (3) and (4) the identical procedure repeated barefoot. Each session began with a 6-min familiarisation at self-selected speed.
  • Perturbation: treadmill belt deceleration at 3 m/s² triggered at right heel strike, reducing speed by about 1.2 m/s, held for 300 ms, then returning to normal; intervals of 15-20 s, pseudo-random timing.
  • Data handling: the first recovery step of perturbations 2-6 was averaged (first perturbation excluded to remove the anticipatory response); 25 consecutive strides averaged for normal gait.
  • Outcomes: coefficient of variation of stride length, stride time, step width and swing time, computed from marker trajectories processed by the Human Body Model in the Gait Offline Analysis Tool.
  • Gait pattern main effects - significant in all four outcomes: stride length variability F(1,13) = 51.24, p < .0001; stride time variability F(1,13) = 32.47, p < .0001; step width variability F(1,13) = 5.180, p = .0404; swing time variability F(1,13) = 14.94, p = .0019.
  • Footwear condition effects - NOT significant for three of four outcomes: stride length variability p = .0675, stride time variability p = .7293, step width variability p = .1430.
  • Swing time variability - the only outcome with footwear effects: condition effect F(1,13) = 5.852, p = .031, partial eta squared = 0.0632; interaction F(1,13) = 11.06, p = .006, partial eta squared = 0.0829. Tukey post hoc: significantly lower swing time variability barefoot vs. shod during trip-Rec1 gait.
  • Gait speed: barefoot walking was significantly slower than shod in both normal gait (p = .0103) and trip-Rec1 gait (p < .0001).
  • Safety: no actual falls occurred in any participant during the experiment.
  • Authors' conclusion: swing time variability can differentiate footwear conditions; lower swing time variability in perturbed recovery gait suggests barefoot walking may be superior to shod walking for perturbation-based balance training in older adults.

Background

  • Perturbation-based balance training on a treadmill - deliberately tripping people and letting them recover - is an emerging training method reported to improve balance-recovery strategies and reduce falls. An open question is which footwear condition should be used during such training: shod or barefoot. This study tested whether barefoot or shod walking produces more stable gait during normal walking and, more importantly, during the critical first recovery step after a trip-like perturbation in older adults.

Methods

  • Design: within-subject laboratory experiment; two-way repeated-measures ANOVA (gait pattern x footwear condition) with Tukey post hoc comparisons.
  • Participants: 14 healthy older adults (age 68.29 ± 3.41 years; body height 1.76 ± 0.10 m; body mass 81.14 ± 14.52 kg). Sex distribution not stated in published abstract or indexed record.
  • Apparatus: Gait Real-time Analysis Interactive Lab treadmill; marker trajectories processed with the Human Body Model in the Gait Offline Analysis Tool.
  • Protocol (fixed order): (1) shod normal walking at self-selected speed, 3 min; (2) shod walking with six trip-like perturbations; (3) barefoot normal walking, 3 min; (4) barefoot walking with six trip-like perturbations. Each session began with a 6-min familiarisation period.
  • Perturbation: triggered at right heel strike; belt decelerated at 3 m/s² until speed dropped by about 1.2 m/s, held for 300 ms, then returned to normal; intervals of 15-20 s with pseudo-random timing.
  • Data selection: first recovery step (Rec1) of perturbations 2-6 averaged, with perturbation 1 excluded to eliminate the anticipatory response; 25 consecutive strides averaged for normal gait.
  • Outcomes: coefficient of variation for stride length, stride time, step width and swing time.

Results

  • Significant gait pattern main effects (normal vs. trip-Rec1) in all four outcomes, with largest effects for stride length variability (partial eta squared = 0.4253) and stride time variability (partial eta squared = 0.3752).
  • No significant footwear condition effect and no significant interaction in three of four outcomes: stride length variability (p = .0675 / .0627), stride time variability (p = .7293 / .9551), step width variability (p = .1430 / .3587).
  • Swing time variability showed both a footwear condition effect (F(1,13) = 5.852, p = .031) and a gait-pattern-by-footwear interaction (F(1,13) = 11.06, p = .006). Tukey post hoc: barefoot showed lower (better) swing time variability during trip-Rec1 gait.
  • Barefoot walking speed was significantly lower than shod speed in normal gait (p = .0103) and trip-Rec1 gait (p < .0001).
  • No actual falls occurred in any participant.

Interpretation

  • From a conventional biomechanics perspective, the study suggests that during the destabilising recovery step after a trip, barefoot walking produces less variable swing timing than shod walking, and therefore that barefoot conditions may be preferable for perturbation-based balance training in older adults.
  • From the WorthDay perspective, the honest reading is considerably narrower:
  • This is an experiment with a manipulated independent variable, which is a genuine strength over uncontrolled before-after reports.
  • But all four endpoints are indirect, computationally derived surrogate measures - coefficients of variation calculated from marker trajectories by modelling software. They are not observed events.
  • No falls occurred. The entire fall-relevance of the study rests on the assumption that gait variability indexes fall risk. That is an inference from surrogate variables, which is exactly the situation the WorthDay criteria treat as a downgrade.
  • Only one of four outcomes showed any footwear effect. Three were null. Across 4 outcomes x 3 effects = 12 statistical tests, two reached significance, and no correction for multiple comparisons is reported. A single significant surrogate among four is weak ground for the headline claim.
  • The experiment was not randomised in condition order: both shod trials always preceded both barefoot trials. Familiarisation, learning and fatigue are therefore indistinguishable from the footwear effect itself. The 6-min familiarisation before each session reduces but does not remove this confound.
  • Barefoot walking was significantly slower than shod walking in both gait patterns. Since gait speed influences variability measures, this is a plausible alternative explanation for the swing-time finding. The authors counter that swing time variability is speed-independent - but that claim is imported from other studies, not demonstrated in these data. Under WorthDay rules, an ad-hoc assumption used to neutralise a confound does not resolve it.
  • No assessor blinding is documented, participant blinding is impossible by design, and there is no untreated or null control condition.
  • n = 14 is a very small sample for a repeated-measures design with interaction terms.
  • Under WorthDay Evidence Criteria this yields very low-quality evidence: the hypothesis is tested experimentally, but exclusively through indirect surrogate endpoints, without randomised condition order, without blinding, without a null control, and with the clinical relevance resting entirely on inference.

Internal inconsistencies in the published report

  • The abstract reports the gait-pattern-by-footwear interaction for swing time variability as p = 0.0055, while the results section reports the same effect as F(1,13) = 11.06, p = 0.006. Minor, but the two figures do not match.
  • For stride time variability the paper reports F(1,13) = 0.1250, p = 0.7293 alongside partial eta squared = 0.1790. An F below 1 cannot correspond to a partial eta squared of 0.179; at least one of these three numbers is wrong.
  • The step-width variability interaction is printed as "F(1,13) = 0.2.431", which is a typographical error rather than a valid F value.
  • The title and conclusion state that barefoot walking "is more stable", but the effect appears in one of four variability outcomes and only in the perturbed recovery condition - the headline is broader than the significant result.
  • These do not change the direction of the finding, but they are reporting-quality problems and are listed here so readers can judge the reliability of the numbers themselves.

Limitations

  • All four outcomes are indirect, software-derived gait-variability coefficients; no falls or any other hard clinical event was measured or observed.
  • Condition order was fixed, not randomised or counterbalanced - shod always ran first, barefoot always second - making order, familiarisation and fatigue indistinguishable from the footwear effect.
  • Very small sample (n = 14) for a design testing interaction effects.
  • Only one of four outcomes showed a footwear effect; three were null, and no correction for multiple comparisons is reported.
  • Barefoot gait speed was significantly lower than shod speed in both gait patterns, a plausible alternative explanation for the swing-time result that the study does not exclude with its own data.
  • The claim that swing time variability is independent of gait speed is imported from external studies - an assumption, not a finding of this experiment.
  • No untreated or null control condition; the study compares shod vs. barefoot only.
  • No documented blinding of the researchers performing data processing or analysis.
  • Acute single-session laboratory exposure on an instrumented treadmill; nothing here measures a training effect over time.
  • Healthy older adults only; results may not apply to people with gait impairment, neuropathy, foot pathology or high fall risk.
  • Sex distribution of the 14 participants is not reported in the published abstract or indexed record.

WorthDay perspective

  • S0063 is the direct counterpoint to S0062. In S0062, minimalist shoes produced better stability indices than bare feet during normal walking. Here, barefoot walking produced the less variable swing time during trip-recovery gait. Two small laboratory studies, different tasks, opposite directions.
  • That opposition is the honest state of the evidence, and WorthDay reports it rather than picking a side. Neither study measured falls, injuries or any hard clinical outcome; both measured modelled variability indices in a laboratory.
  • The most defensible reading of S0063: during the first recovery step after an induced trip, barefoot walking showed less variable swing timing than shod walking in 14 healthy older adults on a treadmill. That is a narrow, single-outcome, acute laboratory finding - not evidence that barefoot walking prevents falls.
  • Users should also note the direction of the speed result: participants walked slower when barefoot. Slower, more cautious gait in an unfamiliar condition is itself a plausible reason for reduced swing-time variability.
  • For WorthDay users the practical takeaway is unchanged from S0062: barefoot or minimalist walking is a legitimate self-experiment, introduced gradually, with attention to how the body responds. The transition question is not resolved by high-quality evidence, and the two best available barefoot gait experiments point in different directions.
  • Barefoot walking is a plausible element within the Health and Necessities & Routines space - alongside movement, recovery and routine - not a proven fall-prevention intervention.

Permalink: https://www.worth-day.com/science-compass/s0063-barefoot-walking-is-more-stable-in-the-gait-of-balance

Related studies

Science meets practice

Turn proven habits into your daily 100 points.

Start your free WorthDay now