Accelerated Split‐Belt Gait Adaptation in Patients with Functional Tremor: Clues to a Generalized Precipitating Trait?

Abstract

Background

Functional neurological disorders (FNDs) are common, disabling conditions that are genuinely experienced and arise from functional rather than structural abnormalities. Although theoretical frameworks implicate aberrant integration of prior expectations and sensory input, empirical evidence is limited, and it remains unclear whether such alterations reflect symptom‐specific deficits or broader trait characteristics.

Objective

To determine whether individuals with functional tremor exhibit altered integration of prior expectations and sensory input, by comparing their rate of adaptation to a split‐belt gait perturbation with that of healthy controls.

Methods

We examined split‐belt treadmill adaptation—a paradigm inducing mismatch between predicted and actual sensory feedback—in patients with functional upper‐limb tremor and age‐matched healthy control subjects. Fifteen patients and 15 control subjects were recruited; three patients were excluded postenrolment because of the presence of functional gait features resulting in an inability to walk unsupported. Participants completed fast and slow tied‐belt baseline walking, 7 minutes of split‐belt adaptation, and 4 minutes of washout (ie, slow tied‐belt walking). Adaptation and washout of step length asymmetry were quantified using single‐exponential fits.

Results

Baseline spatiotemporal gait parameters did not differ between groups. Patients with functional tremor adapted step length asymmetry significantly faster than control subjects (26.7 ± 24.9 seconds vs. 100.6 ± 82.2 seconds; P = 0.006) and showed faster adaptation when returning to tied‐belt modus (4.1 ± 5.9 seconds vs. 12.2 ± 11.3 seconds; P = 0.033).

Conclusions

These findings indicate that individuals with functional tremor exhibit enhanced adaptation to novel sensorimotor contexts, contrasting with slowed adaptation reported in other neurological populations such as Parkinson's disease. Because this study included people with upper‐limb functional tremor, accelerated adaptation may reflect a predisposing trait of FND rather than a symptom‐specific or task‐dependent phenomenon. © 2026 The Author(s). Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.

Functional neurological disorders (FNDs) are common, disabling conditions characterized by genuinely experienced symptoms and recognizable clinical signs that demonstrate internal inconsistency over time. ¹ Manifestations may involve motor, sensory, or cognitive domains. ² Functional tremor, for example, involves the motor domain and is typically defined by variability in frequency and amplitude, as well as entrainment or cessation in response to an externally cued rhythm. ³ Although the pathophysiological mechanisms underlying FND remain incompletely understood, there is consensus that these disorders arise not from structural lesions but from aberrant functioning within brain networks. ⁴ Etiological frameworks emphasize the contribution of predisposing, precipitating, and perpetuating factors to the occurrence of FND. ¹ Predisposing factors may include prior neurological illness, genetic susceptibility, anxiety, depression, emotional dysregulation, and early‐life trauma. Precipitating factors frequently involve acute stressors such as physical injury or significant life events. ⁵

Among several proposed mechanistic models, the Bayesian predictive coding framework has gained prominence. ⁶ In this account, perception of proprioceptive input and voluntary movement is shaped by top‐down predictions (priors) that are combined with incoming sensory signals to generate posterior beliefs, which in turn update future predictions. In the context of movement, this process results in adaptive behavior as the executed movement is adapted to fit the predicted proprioceptive feedforward signals. ⁷ FND symptoms are believed to result from overweighting faulty priors relative to sensory evidence, such as under the influence of attentional mechanisms. ⁶ For instance, functional tremor can be explained by incorrect proprioceptive priors involving scaling and dynamics of movement. However, empirical support for this model remains limited, ⁸ , ⁹ and it is unclear whether alterations in the integration of priors and sensory input are symptom specific or reflect a more generalized trait that predisposes to symptom development across tasks.

To explore this, we examined split‐belt gait adaptation in patients with upper‐limb functional tremor. This well‐established paradigm involves walking on a treadmill with two belts driven at independently controlled speeds. ¹⁰ , ¹¹ Initially, both belts move at the same speed; subsequently, the speed for each belt is altered separately, requiring the participant to adapt their gait. In Bayesian terms, the split‐belt perturbation creates a prediction error, ¹² that is, a mismatch between priors and sensory evidence. This prediction error results in an adaptation of the prior and a gradual adaptation of the gait pattern. Indeed, when the belts begin moving at different speeds, neurotypical adults initially exhibit an asymmetric gait pattern, but over minutes gradually adapt toward symmetry. ¹⁰ , ¹¹ , ¹³ In this study, we compared the time to achieve a symmetrical gait pattern between patients with functional tremor and healthy control subjects. We hypothesized that if alterations in the integration of priors and sensory input are generalized traits, patients with functional tremor would show longer adaptations times, because we expected them to overweigh prior expectations relative to incoming sensory signals.

Subjects and Methods

Subjects

We included 15 individuals with functional tremor and 15 age‐ and sex‐matched healthy control subjects. Because this was an explorative study, no formal sample size calculation was performed. Inclusion criteria were (1) age between 18 and 80 years; (2) a diagnosis of upper‐limb functional tremor based on positive clinical features consistent with the disorder, established by a movement disorders specialist; and (3) the ability to walk on a treadmill without using the handrails for support. Exclusion criteria included comorbidities hampering ambulation or cognition, and the presence of functional gait features. The study protocol was approved by the Central Ethics Review Board of the University Medical Center Groningen (no. 17044) and conducted in accordance with the Declaration of Helsinki. ¹⁴ All participants provided written informed consent before participation. Of the 15 individuals with functional upper‐limb tremor, 3 were excluded postenrolment because they were unable to walk on the treadmill without support because of functional gait impairments.

Procedures

The following participant characteristics were collected: Mini Balance Evaluation Systems Test (Mini‐BESTest), ¹⁵ Hospital Anxiety and Depression Scale (HADS), ¹⁶ and the Activities‐specific Balance Confidence Scale (ABC Scale). ¹⁷ In addition, the Simplified Functional Movement Disorders Rating Scale (S‐FMDRS) was independently scored for the functional tremor group by two movement disorders experts (A.M.M (Madelein) van der Stouwe, University Medical Center Groningen, Groningen), after which the mean score was calculated. ¹⁸

Participants walked on an instrumented split‐belt treadmill (M‐Gait, Motek Medical, Amsterdam, the Netherlands) under four conditions: 2 minutes of fast tied‐belt baseline walking (1.0 m/s), 2 minutes of slow tied‐belt baseline walking (0.5 m/s), 7 minutes of split‐belt adaptation (1.0: 0.5 m/s), and 4 minutes of slow tied‐belt washout (0.5 m/s). For safety, participants wore a fall‐prevention harness attached to an overhead support; the harness did not restrict movement or provide body weight support. Participants were instructed to look straight ahead and were informed that belt speed changes could occur and might be asymmetrical. However, they were not informed about the timing of belt speed changes or the duration of the different phases of the protocol.

Materials

The instrumented treadmill's embedded force plates recorded three‐dimensional ground reaction forces (GRFs; N), moments of force (N m), and two‐dimensional center of pressure (CoP) position (m) at a sample frequency of 1000 Hz through D‐Flow software (Motek Medical).

Data and Statistical Analysis

All data and statistical analyses were performed in MATLAB (version r2024b, The MathWorks Inc., Natick, MA, USA). GRFs and CoP were low‐pass filtered with a 15‐Hz second‐order zero‐phase Butterworth filter. Heel‐strike and toe‐off gait events were determined by detecting the time point at which the vertical GRF crossed the 50‐N threshold. Step length (m) was defined as the difference in anteroposterior CoP position between the left and right foot at heel strike. Step length asymmetry was calculated for every stride i with Eq. 1. This definition of step length asymmetry results in a negative step length asymmetry when the left step length is shorter than the right step length, a positive step length asymmetry when the left step length is longer than the right step length, and a step length asymmetry of 0 when the left and right step length are equally long. Step width (m) was defined as the difference in average mediolateral CoP position during an ipsilateral single support phase and the average mediolateral CoP position during the consecutive contralateral single support phase.

$$\begin{gathered} \text{Step length asymmetry}\left(i\right) \\ =\frac{{\text{step length}}_{\text{left}}\left(i\right)-{\text{step length}}_{\text{right}}\left(i\right)}{{\text{step length}}_{\text{left}}\left(i\right)+{\text{step length}}_{\text{right}}\left(i\right)} \end{gathered}$$ (1)

To assess whether the functional tremor and healthy control groups differed in baseline walking behavior, we calculated baseline step length and step width as the mean of the final 30 strides of slow baseline walking for each participant. Baseline step length variability and step width variability were calculated as the standard deviation of these same 30 strides. Between‐group differences in baseline walking performance were evaluated using independent t tests with an alpha level of 5%. To account for multiple comparisons, we applied a Holm–Bonferroni correction and reported the corrected P values.

To quantify the time to adapt to split‐belt treadmill walking, we fitted a single exponential to the step length asymmetry adaptation curve for each participant. Similarly, to quantify the time to washout of the adapted gait pattern after returning to tied‐belt walking, we fitted a single exponential to step length asymmetry during the washout phase for each participant. The resulting time constants (seconds) for adaptation and washout were tested for between‐group differences with independent t tests against an alpha of 5%.

Results

Participant characteristics are presented in Table 1. Three patients exhibited unilateral upper‐limb functional tremor, and nine patients had bilateral upper‐limb symptoms. Patients had higher scores on the depression and anxiety domains of the HADS. At baseline (Fig. 1), spatiotemporal gait parameters did not differ between patients with functional tremor and healthy control subjects for step length (P = 0.835), step length variability (P = 0.432), step width (P = 0.609), or step width variability (P = 0.876).

TABLE 1.

Patient characteristics

	Individuals with a functional tremor (n = 12)	Healthy control subjects (n = 15)	P
Age (yr)	48.9 ± 15.6	51.3 ± 13.8	0.674
Sex, M/F (% F)	6/6 (50%)	9/6 (40%)
Weight (kg)	89.5 ± 20.7	82.6 ± 11.3	0.280
Length (m)	1.81 ± 0.08	1.84 ± 0.08	0.388
Mini‐BESTest	23.5 ± 3.8	26.9 ± 1.2	0.003
HADS anxiety	5.1 ± 2.4	1.9 ± 1.8	<0.001
HADS depression	3.4 ± 3.6	1.0 ± 1.1	0.020
ABC Scale (%)	74.6 ± 19.8	95.4 ± 3.4	<0.001
S‐FMDRS a	7.7 ± 5.5

P values from independent t tests against an alpha level of 5%. Bold values indicate significant differences between groups.

^a The S‐FMDRS was not scored for two individuals.

M, male; F, female; Mini‐BESTest, Mini Balance Evaluation Systems Test (0–28); HADS anxiety, Hospital Anxiety and Depression Scale, anxiety subscale (0–21); HADS depression, Hospital Anxiety and Depression Scale, depression subscale (0–21); ABC Scale, Activities‐specific Balance Confidence Scale (0–100); S‐FMDRS, Simplified Functional Movement Disorders Rating Scale (0–54).

FIG 1.

Baseline walking characteristics for patients with functional tremor and healthy control subjects. Dots indicate data of individual participants. The vertical raincloud plots show average step length (top left panel), step length variability (top right panel), average step width (bottom left panel), and step width variability (bottom right panel) from the last 30 strides of slow baseline walking. None of these parameters were significantly different between the two groups (P > 0.05). [Color figure can be viewed at wileyonlinelibrary.com]

All participants were able to adapt to the split‐belt perturbation. Patients with functional tremor adapted significantly faster (26.7 ± 24.9 seconds) compared with healthy control subjects (100.6 ± 82.2 seconds; Figs. 2 and 3; t = −2.9956, P = 0.006). This faster adaptation was also seen in the washout phase (ie, when both belts returned to identical speeds). Specifically, adaptation times were 4.1 ± 5.9 seconds for patients versus 12.2 ± 11.3 seconds for control subjects (t = −2.2592, P = 0.033). The shorter adaptation in the functional tremor group was driven primarily by a more rapid increase in step length on the left (faster) belt, whereas the faster washout was predominantly driven by a more rapid increase in step length on the right belt (Fig. 4). Speed of adaptation in patients did not correlate with HADS anxiety scores (r = 0.086, P = 0.79).

FIG. 2.

Group averaged step length asymmetry time series for the functional tremor and healthy control groups. Dotted vertical lines indicate transitions between consecutive phases of the experiment: fast baseline (0–2 minutes), slow baseline (2–4 minutes), split‐belt adaptation (4–11 minutes), and washout (11–15 minutes). Shaded areas indicate standard error around the mean. [Color figure can be viewed at wileyonlinelibrary.com]

FIG. 3.

Vertical raincloud plots of adaptation (top panel) and washout (bottom panel) time constants (seconds) for the functional tremor and healthy control groups. Asterisks indicate significant between‐group differences (*P < 0.05). [Color figure can be viewed at wileyonlinelibrary.com]

FIG. 4.

Group averaged step length time series for the left leg (top panel) and right leg (bottom panel) for the functional tremor and healthy control groups. Dotted vertical lines indicate transitions between consecutive phases of the experiment: fast baseline (0–2 minutes), slow baseline (2–4 minutes), split‐belt adaptation (4–11 minutes), and washout (11–15 minutes). Shaded areas indicate standard error around the mean. [Color figure can be viewed at wileyonlinelibrary.com]

Discussion

We investigated split‐belt gait adaptation in patients with functional tremor and age‐matched healthy control subjects. In contrast with our hypothesis, patients achieved a symmetrical gait pattern significantly faster after split‐belt perturbation than control subjects. To our knowledge, this is the first report of accelerated gait adaptation in a neurological population. By contrast, previous studies have reported slower adaptation in patients with Parkison's disease and freezing of gait, ¹⁹ people with blepharospasm and writer's cramp, ²⁰ and persons after stroke. ²¹

Our findings suggest that patients with an FND may demonstrate an increased propensity to adapt to novel contexts involving strong or unusual sensory input. ⁶ Because this study focused on individuals with functional tremor, the results support the view that accelerated adaptation may represent a predisposing characteristic of FND, rather than a symptom‐specific or task‐dependent phenomenon. This interpretation is consistent with prior experimental evidence demonstrating fundamental alterations in sensorimotor processing in FND compared with healthy control subjects, including differences in sensory information processing ²² , ²³ , ²⁴ , ²⁵ , ²⁶ and perception–action integration. ²⁷ , ²⁸ Our findings are particularly reminiscent of a previous study demonstrating altered probabilistic reasoning in individuals with functional movement disorders using the bead task, a paradigm in which participants infer from which of the two jars beads are being drawn based on sequentially presented evidence. ²⁹ In that study, patients tended to reach decisions with less evidence compared with control subjects and updated their beliefs more readily when new information was presented. An open question remains whether the differences between people with FND and healthy control subjects found in our study, as well as those reported previously, should be understood as intrinsic traits of individuals with FND or as secondary consequences of the disorder.

In contrast with the previous studies investigating sensorimotor processing in FND, the enhanced adaptability observed in patients with FND in this study appeared advantageous. However, this observation raises the question of why healthy individuals adapt more slowly, typically requiring several minutes to reach full adaptation. ¹⁰ , ¹¹ , ¹³ A plausible explanation is a trade‐off between adaptation speed and motor stability. Several theoretical and computational models have been proposed to explain adaptive behavior in humans. In this article, we discuss three potential mechanisms, each grounded in Bayesian principles involving the weighting of sensory or perceptual information and prior beliefs. First, as outlined in the introduction, the Bayesian predictive coding framework has been applied to explain symptoms and signs in FNDs. ⁶ In this model, the precision assigned to sensory input and prior beliefs determines the rate of adaptation. ³⁰ High sensory precision promotes rapid associative learning, whereas high precision of prior beliefs slows adaptation. Individuals with FND may have a predisposition to assign greater precision to sensory input—potentially under the influence of attentional processes—thereby facilitating faster adaptation. This interpretation is supported by previous work using a force matching paradigm in patients with functional movement disorders. ²⁴ In that task, patients showed reduced sensory attenuation, failing to normally dampen the sensory consequences of their own movements, consistent with the notion of heightened weighting or gain of sensory feedback. Second, adaptive processes likely depend on evaluating the relevance of perceived errors under conditions of uncertainty. ³¹ Because sensory and motor signals are inherently noisy, the brain must infer whether an observed error is meaningful. Individuals with FND may attribute greater precision to perceptual errors than healthy control subjects, enhancing adaptability within the present paradigm but conferring a vulnerability to maladaptive learning and, consequently, motor instability in other contexts. Third, the contextual inference model, a more recent and complementary Bayesian framework, posits that the brain maintains multiple distinct motor memories corresponding to different contexts and uses Bayesian inference to identify the most relevant one. ³² In this framework, the split‐belt perturbation constitutes a novel context, requiring the creation of a new memory. Individuals with FND may switch more readily between contextual states, facilitating faster adaptation. Again, this flexibility may come at the cost of motor stability, although this should be explored in future studies, because this was not seen in the current split‐belt experiment.

Bayesian models for motor adaptation have primarily been tested in discrete, episodic tasks such as arm reaching, ³³ where errors do not propagate across episodes. In contrast, locomotion is a continuous task in which errors can accumulate and directly threaten balance. To capture this, recent work has modeled locomotor adaptation as the interaction between a fast stabilizing controller and a slower reinforcement learner that improves performance through local exploration via natural step‐to‐step variability and memory from previously encountered situations. ³⁴ This framework, using energy minimization with a minor penalty for asymmetry as its performance criterion, has successfully explained learning and generalization across multiple prior and model‐guided experiments. Within this model, the accelerated adaptation in patients with functional tremor observed in this experiment may be explained by several nonexclusive mechanisms. First, patients may permit greater motor variability, thereby enhancing exploration and accelerating gradient estimation. However, patients did not exhibit differences in gait variability during baseline walking. A second explanation is that their cost function may assign greater weight to symmetry, effectively steepening the learning gradient under split‐belt perturbations. Third, the model predicts faster adaptation under conditions of reduced sensory noise. Although described as noise introduced by the treadmill, one may hypothesize that a reduction in sensory noise also reflects a higher precision assigned to incoming sensory inputs or perceived errors. Such an interpretation provides a conceptual bridge between reinforcement learning and Bayesian accounts of motor adaptation.

Irrespective of the mechanisms underlying motor adaptation, we did not find evidence for maladaptive adaptation in the present task. Unlike a previous study employing the “broken escalator” paradigm, ³⁵ which showed stronger aftereffects in patients with functional gait disorders than in healthy people, we found that patients were faster when returning to tied‐belt modus. Notably, when examining the adaptation phase in that earlier study, patients exhibited faster motor adaptation, reflected by a more rapid decrease in trunk acceleration during the initial trials, findings that are consistent with the present results.

It remains challenging to disentangle alterations in motor learning from the effects of heightened state anxiety. Patients scored higher on the HADS than healthy control subjects, although this did not correlate with speed of adaptation. Including a psychiatric control group, such as individuals with generalized anxiety disorder, could help to clarify the specificity of the observed effects.

Future studies should aim not only to replicate our findings, including retention of split‐belt adaptation, but also to implement computational paradigms capable of probing Bayesian inference in this population. ³³ Although such studies are methodologically challenging and require extensive empirical data, they represent a crucial next step in elucidating the mechanisms underlying FND. Moreover, because these paradigms typically involve consciously controlled voluntary movements, they may help determine whether accelerated adaptation is observed in both automatic movements (eg, gait) and more consciously controlled actions. In addition, future work may try to clarify the state versus trait nature of the findings by examining patients longitudinally over time.

Author Roles

(1) Research Project: A. Conception, B. Organization, C. Execution; (2) Statistical Analysis: A. Design, B. Execution, C. Review and Critique; (3) Manuscript Preparation: A. Writing of the First Draft, B. Review and Critique.

J.N.: 1A, 1B, 2C, 3A.

D.V.: 1C.

H.C.: 3B.

M.A.J.T.: 1A, 3B.

T.J.W.B.: 1A, 1B, 1C, 2A, 2B, 3A.

Financial Disclosures and Conflicts of Interest

The authors have no (financial) conflicts of interest to declare. Author disclosures are available in the Supporting Information.

Data Availability Statement

The data reported in this study are available from the corresponding author on reasonable request.