Effect of Tai Chi Practice on the Adaptation to Sensory and Motor Perturbations While Standing in Older Adults

Abstract

Tai Chi provides an age-appropriate exercise to decrease fall risks in older adults. However, the exact mechanism underlying the benefits of Tai Chi practice remains an open question. Thus, this study examined how aging and Tai Chi practice impact adaptation to sensory and motor perturbations while standing. We hypothesized that older Tai Chi practitioners would exhibit a decreased reliance on visual processes as sensory and motor perturbations increased, relative to naive healthy older adults. Using rambling and trembling decompositions of the center of pressure (COP) and frequency-domain features, we examined changes in low (0–0.3 Hz), medium (0.3–1 Hz), and high (1–3 Hz) frequency components, reflecting contributions from the visual, vestibular/somatosensory, and proprioceptive systems, respectively, in healthy young adults (HYA), healthy older adults (HOA), and Tai Chi practicing older adults (TCOA). Our results revealed statistically significant condition-by-group interactions in high-frequency COP-x and rambling-x and COP-y components, medium-frequency COP-y components, and all low-frequency components in COP and trembling (p < 0.05). Further, a significant trial-by-group interaction in high-frequency rambling-y was observed (p < 0.05). These results indicate age and Tai-chi-related differences in modulation of sensory contributions to balance as perturbations increase, and with repeated practice, which merit further investigation.

1. Introduction

With an increasing number of individuals over 65 years of age around the world, an increasing number of people face functional impairments associated with aging, including reduced balance function and increased fall risk [1]. Nearly 30% of older adults report falling [2], which leads to injuries, reductions in quality of life, and even death [2,3,4]. Considering the severe consequences of falls in older adults, it is essential to further our understanding of the mechanisms underlying falls and the evaluation of factors that may contribute to the effectiveness of training programs aimed at reducing falls in older adults.

Postural control [5] declines with advancing age and is strongly associated with an increased fall risk in older adults [6]. Postural control is defined as the act of maintaining, achieving, or restoring a state of balance that may involve either a fixed-support or a change in support response [7]. Postural control is commonly assessed using posturography, which quantitatively evaluates postural stability [5]. Compared to younger adults, older adults show significantly reduced performance under sway-referenced surface conditions when visual input is absent or unavailable, suggesting age-related declines in vestibular function and overall balance control [8]. This impairment is even more pronounced in older adults at higher risk of falling, who demonstrate a reduced ability to rely on vestibular input when both visual and somatosensory information are inaccurate [5,8,9]. Similarly, a study on older women found diminished stability under these conditions, along with an age-related decline in somatosensory function. These findings point to a general deterioration of both somatosensory and vestibular systems with age and emphasize the growing compensatory importance of visual input in maintaining balance [10]. Nonetheless, further research is needed to assess the effectiveness of targeted, evidence-based interventions, such as Tai Chi, for improving postural control and reducing fall risk in older adults [11].

Tai Chi is a Chinese martial art that combines physical practice, breathing techniques, and meditation, which has been widely used to improve postural control and prevent falls in older adults [11,12,13]. Prior work has demonstrated the effectiveness of Tai Chi at improving postural control [14]. Specifically, Tai Chi practitioners have exhibited greater postural stability by reducing their reliance on perturbed visual or proprioceptive inputs while relying more on unperturbed visual, proprioceptive, or vestibular inputs under various sensory perturbation conditions [15]. However, there have been ambiguous findings on the exact mechanism that may be underlying Tai Chi practice benefits in postural control [16].

Upright postural and balance control requires a complex interplay within and between the sensory and the motor systems. Furthermore, there is strong evidence for the crucial contribution of the cerebral cortex in the control of balance [17,18,19]. The variability of the center of pressure (COP) in postural control can be decomposed into two main components: rambling and trembling [20,21]. Rambling reflects the movement of a reference point, which estimates the body’s dynamic state and is formed through the integration of various sensory inputs [22]. Trembling consists of high-frequency oscillations around the reference point, primarily regulated by the intrinsic stiffness of muscles and spinal reflexes [20,21]. In line with the work performed by Tahayori et al. [23], rambling is defined as the trajectory of the COP at instantaneous equilibrium points (IEPs)—the moments when the horizontal ground reaction force is zero. At these points, the body’s position is considered to be in mechanical equilibrium, and the COP at IEPs effectively reconstructs the voluntary, low-frequency component of postural control. As a result, the rambling trajectory closely resembles the original COP time series, capturing the overall movement path of the body’s center of mass. The trembling component, by contrast, is assessed as the residual, the difference between the actual COP trajectory and the reconstructed rambling trajectory. This residual isolates the high-frequency, involuntary fluctuations around the equilibrium path, providing insight into the fine motor adjustments that occur during quiet standing. These two components are distinguished by their underlying control mechanisms; rambling is associated with supraspinal processes, whereas trembling is attributed to peripheral mechanisms, such as spinal reflexes and muscle properties [20,21]. This distinction provides insights into whether postural control is primarily mediated by sensory integration at the cortical level or by reflex-level regulation [20,21,22].

In this study, we investigated the impact of aging and Tai Chi practice on the adaptation to sensory and motor perturbations while standing, using a rambling and trembling decomposition. We hypothesized that older Tai Chi practitioners would exhibit a decreased reliance on visual processes as sensory and motor perturbations increased, relative to naive healthy older adults. Further, we hypothesized that older adults, relative to younger adults, would exhibit decreased adaptation to sensory and motor perturbations, indicative of a decreased flexibility in sensorimotor integration, as measured by a smaller change in power of visual processes across repeated trials.

2. Materials and Methods

2.1. Protocol

This cross-sectional study consisted of a single session. Community-dwelling adults with the following inclusion criteria were recruited: (1) right-handed; (2) young adults between 18 to 30 years of age and older adults over 65 years of age with and without Tai Chi practice experience; (3) free of chronic or acute neurological conditions, such as Parkinson’s disease, Huntington’s disease, stroke, epilepsy, and seizures; and (4) free of severe heart conditions, such as heart attack, heart failure, and coronary heart disease, including, symptoms such as angina pectoris. Exclusion criteria included: (1) cognitive impairment, as defined by a Modified Telephone Interview for Cognitive Status (TICS-M) questionnaire score lower than 18 [24]; (2) physical disability or inability to walk independently without an assistive device; and (3) severe chronic pain that limits physical function. The Tai Chi practice cohort was defined by having at least 50 practice hours. Once in the study, all participants read and signed a written informed consent form. The protocol and procedures were reviewed and approved by the Institutional Review Board of the University of Illinois Urbana-Champaign (IRB protocol number 15317).

2.2. Participants

We recruited 23 healthy young adults, 21 healthy older adults, and 15 older adult Tai Chi practitioners from the local community to participate in this study (See

Table 1. Sample characteristics by group.

	Young Adult (N = 23)		Older Adult (N = 21)		Tai Chi Practitioner (N = 15)
	Mean	SD	Mean	SD	Mean	SD	p-Value
Age (y)	21.4	2.0	70.8	5.5	76.7	5.6	<0.001
Height (cm)	170.9	11.0	171.0	10.4	168.1	9.0	0.649
Weight (kg)	71.5	11.5	72.0	11.9	64.2	7.0	0.073
Foot Size (cm)	25.9	1.7	25.9	1.6	25.5	1.4	0.654
SOT Score [0–100]	75.8	4.7	75.4	6.7	73.0	6.8	0.369
SCK Difficulty [0–4]	0.2	0.4	1.1	0.9	0.9	0.9	<0.001
Brisk Activity/Week	4.0	2.3	4.8	3.4	3.0	2.2	0.265
FES-I Score [16–64]	18.0	2.1	19.9	3.2	22.7	3.9	<0.001
	Number	Percent	Number	Percent	Number	Percent	p-Value
Females	12	52.2	11	52.4	9	60	0.874

One-way analysis of variance was used for group differences in continuous variables. The chi-square test was used for group differences in the distribution of females. Abbreviation: SOT score, sensory organization test equilibrium score; SCK, stooping, crouching, or kneeling; FES-I, Falls Efficacy Scale-International.

), as part of an ongoing study at the Mobility and Fall Prevention Research Lab at the University of Illinois Urbana-Champaign. The Tai Chi practitioner group had at least 50 practice hours of a modified Yang style (Range = 61–4680 h). All practitioners came from two local Tai Chi groups and practiced in a similar style (i.e., Yang style 24 form or modified Yang style).

2.3. Postural Control Tasks

Participants were asked to stand as still as possible, with their medial malleoli aligned to a center horizontal line and calcanei aligned to a height-dependent stance width, during the performance of the sensory organization test (SOT), while the COP data were recorded (Figure 1). The SOT is a clinically used standardized instrumented balance test performed using the SMART EquiTest-Clinical Research System (SECRS, Neurocom, a division of Natus). The SOT is designed to assess a patient’s use of sensory systems that contribute to balance and identify any abnormalities in the systems [25]. The six conditions of the SOT manipulate or eliminate information normally delivered to the patient’s eye, head, feet, and joints. Specifically, there are three trials per condition and 20 s per trial in the SOT. The SOT measures an individual’s ability to suppress the misleading information from the conflicting senses and use the remaining sensory input to maintain an upright stance [26]. Thus, in this study, the SOT introduces visual and somatosensory perturbations using sway-referenced mechanical ankle rotations, as part of the different sensory and minor mechanical perturbations presented to participants. To quantify postural control performance in the SOT, the SOT equilibrium score [0–100] was calculated using the following formula: SOT score = ((12.5 − (Maximum y-direction sway − Minimum y-direction sway)/12.5) × 100, which quantifies how well the participant’s sway remains within the expected angular limits of stability during each SOT condition, with higher scores indicating higher performance. Furthermore, self-reported functional capacity, physical activity, and psychological function were evaluated to help control for characterization of the three different groups. Self-reported functional capacity and physical activity were evaluated using a subset of the Established Populations for Epidemiologic Studies of the Elderly questionnaire [27]. In particular, stooping, crouching, or kneeling difficulty was assessed using a 0–4 scale ranging from 0 = no difficulty at all to 4 = just unable to do it [28], and brisk activity frequency per week. Lastly, the fall risk of the participants was assessed by the Falls Efficacy Scale-International (FES-I), with scores ranging from 16 to 64, with a higher score indicating a greater fear of falling [29].

2.4. Data Analysis

COP data were collected through the SERCS at a 100-Hz sampling rate. In both datasets, the primary outcome measure from the COP data was the analysis of rambling and trembling components, which are indicative of postural stability and control. According to the authors of [23], rambling and trembling components were extracted from the COP signals in both the x-, or mediolateral, and y-, or anteroposterior, axes. To process the COP data, each trial was analyzed using a custom Python script (version 3.12.7). Force signals were first smoothed using a 4th-order low-pass Butterworth filter, with a 6-Hz cutoff, which facilitated accurate identification of zero-crossings, or IEPs, without attenuating the high-frequency content necessary for subsequent trembling analysis. The COP signal was then mean-centered to remove any offset in COP position, focusing the analysis on postural excursions. To decompose the COP trajectory, IEPs were identified as the time points where the horizontal ground reaction force crossed zero, signifying moments of mechanical equilibrium. The COP values at these IEPs were interpolated using cubic splines to reconstruct the rambling trajectory. Trembling was subsequently defined as the residual signal obtained by subtracting the rambling trajectory from the mean-centered COP. The decomposition algorithm was designed to robustly handle missing data and to maintain reliable interpolation, even in cases of sparse IEPs.

For the spectral analysis, the power spectral density (PSD) of both the rambling and trembling components was computed using the Fast Fourier Transform (FFT), consistent with prior work [30]. The PSD was then integrated within predefined frequency bands (low frequency: 0–0.3 Hz, medium frequency: 0.3–1 Hz, high frequency: 1–3 Hz, reflecting contributions from the visual, vestibular/somatosensory, and proprioceptive systems, respectively) to quantify the power in each band [30]. An extraction of output files that carry out the frequency analysis on the COP data exclusively without decomposition into rambling and trembling was also performed in order to confirm the use of specific sensory modalities.

2.5. Statistical Analysis

All the statistical analyses were performed using R (R 4.2.2, RStudio 2023.06.0 + 421). There were four sets of statistical analyses that were performed to answer the research questions. A one-way analysis of variance and chi-square test were used to examine cohort demographic differences. For primary outcome measurements, linear mixed-effect models (LMMs) were used to identify the cohort differences for postural control performance. Specifically, LMMs were applied to the power of the low-, medium-, and high-frequency bands of the COP in the x- and y-directions, as well as the rambling and trembling components in the x- and y-directions. These models tested the effects of fixed effects of condition, trial, group, and age, as well as their interactions, including condition-by-trial, condition-by-group, trial-by-group, and condition-by-trial-by-group on the relative power during the SOT. Analyses included condition, trial, and group as fixed effects, age as a covariate, and rankit-transformed variables to address the non-normality of outcome measures.

A significance level of p  <  0.05 was used to determine statistical significance. Moreover, Spearman correlations were used to investigate the relationship between power and SOT score in each condition. Type III ANOVA tables with Satterthwaite’s method were used to evaluate the significance of main effects and interactions, providing F-values and p-values (

Table 2. Analysis of variance table for the linear mixed model examining main effects and interactions of condition, trial, and group on the low-, medium-, and high-frequency band of the center of pressure in x and y, with age as a covariate.

	Low Frequency		Medium Frequency		High Frequency
Source of Variance	F-Value	p-Value	F-Value	p-Value	F-Value	p-Value
Center of Pressure in x
Condition	6.47	<0.001	247.45	<0.001	373.98	<0.001
Trial	0.04	0.962	25.71	<0.001	49.83	<0.001
Group	0.39	0.680	2.22	0.118	1.66	0.200
Age	1.22	0.274	4.60	0.036	1.92	0.172
Condition × Trial	0.43	0.935	1.36	0.196	2.40	0.008
Condition × Group	2.11	0.021	1.02	0.423	2.13	0.020
Trial × Group	0.64	0.631	1.41	0.229	0.23	0.924
Condition × Trial × Group	0.16	1.000	1.36	0.131	1.33	0.148
Center of Pressure in y
Condition	69.96	<0.001	283.50	<0.001	541.60	<0.001
Trial	6.51	0.002	17.78	<0.001	75.75	<0.001
Group	2.61	0.083	1.95	0.152	2.47	0.093
Age	0.03	0.869	2.62	0.111	0.14	0.709
Condition × Trial	2.75	0.002	1.13	0.339	3.14	0.001
Condition × Group	2.14	0.020	2.06	0.025	1.99	0.032
Trial × Group	0.31	0.873	1.95	0.101	0.31	0.870
Condition × Trial × Group	0.82	0.697	1.04	0.408	1.19	0.258

,

Table 3. Analysis of variance table for the linear mixed model examining main effects and interactions of condition, trial, and group on the low-, medium-, and high-frequency band of the x and y rambling components, with age as a covariate.

	Low Frequency		Medium Frequency		High Frequency
Source of Variance	F-Value	p-Value	F-Value	p-Value	F-Value	p-Value
Rambling in x-direction
Condition	202.97	<0.001	148.08	<0.001	94.63	<0.001
Trial	22.74	<0.001	9.99	<0.001	14.50	<0.001
Group	1.35	0.268	1.93	0.154	0.62	0.539
Age	2.06	0.157	3.28	0.075	2.37	0.129
Condition × Trial	1.08	0.378	1.73	0.070	1.65	0.088
Condition × Group	0.69	0.734	1.72	0.072	2.71	0.003
Trial × Group	0.31	0.869	1.57	0.180	1.42	0.227
Condition × Trial × Group	1.36	0.131	1.32	0.156	1.19	0.250
Rambling in y-direction
Condition	332.66	<0.001	212.60	<0.001	161.44	<0.001
Trial	26.71	<0.001	17.62	<0.001	25.19	<0.001
Group	0.39	0.680	4.34	0.018	3.73	0.030
Age	0.29	0.592	7.15	0.010	7.36	0.009
Condition × Trial	1.65	0.088	1.47	0.146	1.54	0.119
Condition × Group	0.93	0.505	1.34	0.204	1.04	0.404
Trial × Group	2.21	0.066	2.37	0.051	3.44	0.008
Condition × Trial × Group	0.68	0.847	0.56	0.942	0.84	0.663

and

Table 4. Analysis of variance table for the linear mixed model examining main effects and interactions of condition, trial, and group on the low-, medium-, and high-frequency band of the x and y trembling components, with age as a covariate.

	Low Frequency		Medium Frequency		High Frequency
Source of Variance	F-Value	p-Value	F-Value	p-Value	F-Value	p-Value
Trembling in x-direction
Condition	1.89	0.094	121.49	<0.001	214.49	<0.001
Trial	0.20	0.817	15.27	<0.001	24.75	<0.001
Group	0.33	0.722	3.85	0.027	2.08	0.134
Age	1.14	0.290	6.14	0.016	3.82	0.056
Condition × Trial	0.37	0.961	1.24	0.261	1.65	0.087
Condition × Group	2.22	0.015	1.15	0.323	1.25	0.258
Trial × Group	0.57	0.686	1.03	0.393	1.08	0.365
Condition × Trial × Group	0.19	1.000	1.00	0.456	1.07	0.380
Trembling in y-direction
Condition	116.23	<0.001	389.96	<0.001	546.63	<0.001
Trial	6.37	0.002	24.40	<0.001	76.31	<0.001
Group	1.65	0.201	1.65	0.201	3.15	0.050
Age	0.03	0.873	0.14	0.711	0.18	0.670
Condition × Trial	2.40	0.008	0.90	0.536	3.56	<0.001
Condition × Group	2.09	0.023	1.11	0.350	1.78	0.060
Trial × Group	1.27	0.279	1.36	0.245	0.80	0.526
Condition × Trial × Group	0.58	0.928	0.63	0.896	1.16	0.285

).

3. Results

3.1. Overall Results

Participant characteristics are tabulated per group (Table 1). Age, self-reported stooping crouching kneeling difficulty, and Falls Efficacy Scale-International score were found to be significantly different within the three groups. No other significant differences were observed amongst the three groups. Mean height, weight, foot size, sensory organization test score, brisk activity frequency per week, and proportion of females within each group were comparable in all groups. Given the significant association between increased age and increased self-reported stooping, crouching, and kneeling difficulty and fear of falling [31,32], analyses were adjusted for age to account for this confounding factor in the Tai Chi practitioner group. To provide an integrative overview of spectral power variations across trajectory components and frequency bands, we plotted the mean and standard error of the mean (SEM) for each group and each condition (Figure 2).

3.2. Center of Pressure Trajectories

The overall LMMs demonstrated a statistically significant condition (F = 6.47, p < 0.001) and condition-by-group interaction (F = 2.11, p = 0.021) effect in low-frequency power of COP in the x-direction; and condition (F = 69.96, p < 0.001), trial (F = 6.51, p = 0.002), condition-by-trial interaction (F = 2.75, p = 0.002), and condition-by-group interaction (F = 2.14, p = 0.020) effect in low-frequency power of COP in the y-direction (See Table 2). In addition, there was a statistically significant condition (F = 247.45, p < 0.001), trial (F = 25.71, p < 0.001), and age (F = 4.60, p = 0.036) effect in medium-frequency power of COP in the x-direction; and condition (F = 283.50, p < 0.001), trial (F = 17.78, p < 0.001), and condition-by-group interaction (F = 2.06, p = 0.025) effect in medium-frequency power of COP in the y-direction. Lastly, there was also a statistically significant condition (F = 373.98, p < 0.001), trial (F = 49.83, p < 0.001), condition-by-trial interaction (F = 2.40, p = 0.008), and condition-by-group interaction (F = 2.13, p = 0.020) effect in high-frequency power of COP in the x-direction; and condition (F = 541.60, p < 0.001), trial (F = 75.75, p < 0.001), condition-by-trial interaction (F = 3.14, p = 0.001), and condition-by-group interaction (F = 1.99, p = 0.032) effect in high-frequency power of COP in the y-direction.

3.3. Rambling Trajectories

The overall LMMs demonstrated a statistically significant condition (F = 202.97, p < 0.001) and trial (F = 22.74, p < 0.001) effect in low-frequency power of rambling in the x-direction. In addition, there was a statistically significant condition (F = 332.66, p < 0.001) and trial (F = 26.71, p < 0.001) effect in low-frequency power of rambling in the y-direction (See Table 3). Moreover, there was a statistically significant condition (F = 148.08, p < 0.001) and trial (F = 9.99, p < 0.001) effect in medium-frequency power of rambling in the x-direction; and condition (F = 212.60, p < 0.001), trial (F = 17.62, p < 0.001), group (F = 4.34, p = 0.018), and age (F = 7.15, p = 0.010) effect in medium-frequency power of rambling in the y-direction. Lastly, there was a statistically significant condition (F = 94.63, p < 0.001), trial (F = 14.50, p < 0.001), and condition-by-group interaction (F = 2.71, p = 0.003) effect in high-frequency power of rambling in the x-direction; and condition (F = 161.44, p < 0.001), trial (F = 25.19, p < 0.001), group (F = 3.73, p = 0.030), age (F = 7.36, p = 0.009), and trial-by-group interaction (F = 3.44, p = 0.008) effect in high-frequency power of rambling in the y-direction.

3.4. Trembling Trajectories

The overall LMMs demonstrated a statistically significant condition-by-group interaction (F = 2.22, p = 0.015) effect in low-frequency power of trembling in the x-direction. In addition, there was a statistically significant condition (F = 116.23, p < 0.001), trial (F = 6.37, p = 0.002), condition-by-trial interaction (F = 2.40, p = 0.008), and condition-by-group interaction (F = 2.09, p = 0.023) effect in low-frequency power of trembling in the y-direction (See Table 4). There was also a statistically significant condition (F = 121.49, p < 0.001), trial (F = 15.27, p < 0.001), group (F = 3.85, p = 0.027), and age (F = 6.14, p = 0.016) effect in medium-frequency power of trembling in the x-direction; and condition (F = 389.96, p < 0.001) and trial (F = 24.20, p < 0.001) effect in medium-frequency power of trembling in the y-direction. Furthermore, there was a statistically significant condition (F = 214.49, p < 0.001) and trial (F = 24.75, p < 0.001) effect in high-frequency power of trembling in the x-direction; and condition (F = 546.63, p < 0.001), trial (F = 76.31, p < 0.001), and condition-by-trial interaction (F = 3.56, p < 0.001) effect in high-frequency power of trembling in the y-direction.

3.5. Correlation Analyses

The overall Spearman correlations between low-, mid-, and high-frequency power and SOT performance for each cohort are presented in Figure 3. Correlations ranged from −0.29 to 0.013 in HOA, −0.148 to 0 in HYA, and −0.202 to 0.249 in TCOA. Similarly, a breakdown of each condition is provided in Figure S1, where increased correlations in low-frequency power were observed in TCOA in trembling and overall COP in trials 1–3. Lastly, a summary by trial is provided in Figure S2, where modulation of low-frequency power was observed across all groups.

4. Discussion

In this study, the impact of aging and Tai Chi practice on the adaptation to sensory and motor perturbations while standing was observed, using frequency-domain features of the COP and rambling and trembling decompositions. We hypothesized that older Tai Chi practitioners would exhibit a decreased reliance on visual processes as sensory and motor perturbations increased, relative to naive healthy older adults. However, no significant condition-by-group interactions were observed in the low-frequency components between these groups. Additionally, we hypothesized that healthy older adults, relative to healthy younger adults, would exhibit decreased adaptation to sensory and motor perturbations across repeated trials. Partially supporting this hypothesis, a significant trial-by-group interaction was observed in the high-frequency rambling-y component, although this effect was not clearly linked to specific sensory or motor perturbation conditions.

Contrary to our first hypothesis, Tai Chi practitioners did not exhibit a measurable reduction in visual dependence across increasing perturbation levels. This suggests that Tai Chi practice may not lead to enhanced down-weighting of visual input in challenging sensory and motor contexts. Prior studies have reported mixed findings regarding visual dependence in Tai Chi practitioners. Some have shown reduced visual reliance, such as improved balance with eyes closed [15], while others reported increased dependence [33] or no difference at all [34]. Nonetheless, these studies consistently indicate that Tai Chi improves overall balance by enhancing sensory re-weighting [15,33,34]. In line with this, our results suggest that the effects of Tai Chi may reflect broader enhancements in multisensory integration rather than selective changes in reliance on a specific modality.

In partial support of our second hypothesis, younger adults exhibited greater reductions in high-frequency power over repeated trials in the rambling-y component, suggesting more effective proprioceptive adaptation. Notably, older adults showed lower overall high-frequency power in this component, as reflected by a significant main effect of group. This pattern aligns with prior findings that aging is associated with diminished proprioceptive adaptation in postural control [35]. However, the lack of a significant three-way interaction (condition-by-trial-by-group) suggests that this adaptation was not strongly modulated by specific perturbation types.

To evaluate whether these adaptation patterns were associated with postural performance, we examined correlations between high-frequency power and SOT scores. Among older adults, proprioceptive reliance showed a weak-negative correlation with stability in Trial 1 (r = −0.144), which diminished by Trial 3 (r = −0.0299). In contrast, younger adults showed a slightly increasing correlation across trials (Trial 1: r = −0.0746; Trial 3: r = −0.137), although still weak. This suggests that reduced proprioceptive reliance over time may be modestly associated with improved postural stability in younger adults, potentially reflecting more effective proprioceptive adaptation. For older adults, however, this relationship was minimal, possibly reflecting age-related declines in proprioceptive sensitivity and integration that limit the benefits of sensory adaptation [36].

These adaptation effects were observed primarily in the rambling-x component, suggesting that age-related differences may reflect disparities in cortical-level sensory integration. This interpretation is supported by evidence linking rambling to sensory re-weighting processes [37] and EEG studies implicating cortical activity during postural control under SOT conditions [38]. The observed effects in the mediolateral direction further suggest that such integration differences are particularly evident in mediolateral balance control, a known predictor of fall risk in older adults [39]. Thus, the rambling-x component may serve as a sensitive marker of cortical sensory integration efficiency, particularly in mediolateral stability.

In addition to high-frequency changes, older adults showed significantly lower mid-frequency power than younger adults in the rambling-y and trembling-x components, indicating reduced reliance on vestibular and somatosensory inputs. These findings are consistent with age-related declines in vestibular function due to hair cell loss [40] and diminished somatosensory processing, as reflected in reduced gating and cortical alpha oscillations in MEG studies [41]. These reductions likely contribute to degraded balance control in older adults.

We further examined whether mid-frequency power was associated with postural performance. In the rambling-y component, correlations with SOT scores were weak for both groups (older adults: r = −0.0385; younger adults: r = −0.0272), indicating negligible relationships. This is consistent with prior studies suggesting that, although vestibular and somatosensory ratios decline with age, their impact on postural performance may be limited under stable conditions [10].

In contrast, the trembling-x component showed a moderate negative correlation with postural stability in older adults (r = −0.29) and a weaker correlation in younger adults (r = −0.13). In older adults, lower mid-frequency power, indicative of reduced vestibular and somatosensory reliance [15], was associated with better postural performance. This may reflect a compensatory shift toward biomechanical strategies, such as increased ankle co-contraction, which enhances joint stiffness and stabilizes posture despite diminished sensory input [42]. Given that the trembling component has been linked to mechanical and reflexive properties of muscle control [43], these findings suggest that older adults may increasingly rely on reflex-level mechanisms to maintain ML stability when sensory inputs are less reliable.

While all groups demonstrated a similar frequency of brisk activity per week, group differences in functional performance and fear of falling may aid in explaining findings. Supplementary results examining the association between balance performance and fear of falling demonstrated consistently strong negative associations in older adults.

Limitations of the current work include the limited sample size and heterogeneity in the number of practice hours and the timeframe for completing practice hours of the Tai Chi older adult practitioners. In addition, age, functional performance, and fear of falling may play a role in adaptation to more challenging postural control tasks, which may benefit from further investigation. Future work may benefit from larger sample sizes, the use of neuroimaging to confirm sensorimotor integration changes, and the use of an intervention to evaluate changes in sensorimotor integration at different time points of an intervention.

5. Conclusions

This study demonstrated the effect of aging and Tai Chi practice on the adaptation to sensory and motor perturbations while standing and highlights the utility of using frequency-domain features of the COP and rambling and trembling decompositions to further our understanding of postural control. The analysis of rambling and trembling decompositions during balance-demanding tasks allowed for the identification of age and Tai Chi-related changes in the use of different sensory modalities as task demands change and adaptation to repeated exposures. These insights provide a foundation for further inquiry into the benefits of mindful movement practices in older adults.