Heart rate variability dynamics during virtual reality roller coaster experience: correlations with vestibular function and motion sickness susceptibility

Abstract

Virtual reality (VR) increasingly causes motion sickness, yet physiological mechanisms and individual susceptibility factors remain unclear. Understanding autonomic nervous system responses and their relationship to vestibular function is crucial for developing safer VR applications. This study aimed to characterize autonomic responses to visually induced motion sickness (VIMS) exposure through heart rate variability (HRV) analysis and determine correlations between vestibular function and motion sickness susceptibility. Thirty participants underwent finger PPG monitoring during baseline, VR exposure, and recovery phases, and additionally completed motion sickness questionnaires and vestibular evoked myogenic potential (VEMP) testing. HRV parameters were analyzed using a mixed-design ANOVA, with Pearson correlations examining vestibular-autonomic relationships. VR exposure significantly increased heart rate (3.67%, ) and HRV parameters (SDNN: + 41.12 ms, RMSSD: + 43.50 ms), followed by pronounced recovery decreases. High motion sickness susceptibility individuals showed the greatest autonomic reactivity. Associations emerged between VR symptoms and ocular VEMP amplitudes (, ), and between vestibular function and autonomic responses during VR (, ); these are interpreted as hypothesis-generating findings. VR exposure triggers measurable autonomic responses that correlate with individual motion sickness susceptibility and vestibular function. These findings support using HRV monitoring and vestibular assessment as candidate markers of VR-induced discomfort.

Introduction

Visually induced motion sickness (VIMS) is motion sickness primarily caused by stimulation of the visual system, known as a physiologic reaction that includes various symptoms and signs: nausea and/or gastrointestinal disturbance, thermoregulatory disruption, alteration in arousal, dizziness and/or vertigo, headache and/or ocular strain. When exposure ends, these disturbances resolve¹. Since symptoms were first documented in the 1950s in laboratory and military domains, technology has advanced and humans are now exposed to new simulators such as 4D theaters, virtual reality, and metaverse environments^2,3.

From the perspective of VIMS or motion sickness studies, personalisation and objectivity present the greatest challenges. Not everyone experiences motion sickness with the same intensity, symptoms, or signs. For prediction and prevention of motion sickness, various studies have suggested relationships between vestibular laboratory results and motion sickness; however, these relationships remain unclear, even in new environments such as virtual reality (VR)^4,5.

VR technology has rapidly evolved from experimental prototypes to mainstream applications across healthcare, entertainment, education, and training. This widespread adoption has raised important questions about the physiological effects of immersive VR experiences on users^6–8. Of particular concern is the phenomenon commonly known as cybersickness or VR sickness, which shares symptomatic similarities with motion sickness yet occurs without actual physical movement^9,10. Understanding the underlying physiological mechanisms of VR-induced discomfort is crucial for developing more comfortable and accessible VR applications¹¹.

The sensory conflict theory, one of the predominant explanations for motion sickness, suggests that symptoms arise from discrepancies between expected and actual sensory inputs across visual, vestibular, and proprioceptive systems^12,13. In VR environments, users experience visual motion cues without corresponding vestibular stimulation, potentially creating such sensory conflicts. The autonomic nervous system (ANS), responsible for regulating unconscious bodily functions, appears to play a significant role in the physiological response to these conflicts¹⁴.

The present study aimed to characterize the ANS responses to a virtual roller coaster experience by analyzing heart rate variability (HRV) parameters across baseline, VR exposure, and recovery phases. Additionally, we sought to determine if photoplethysmography (PPG) characteristics change among individuals with different levels of motion sickness susceptibility. Vestibular evoked myogenic potentials (VEMPs) are electrophysiological tests that assess otolith organ function through reflex pathways. Cervical VEMP (cVEMP) evaluates saccular function via the vestibulocollic reflex, while ocular VEMP (oVEMP) assesses utricular function through the vestibulo-ocular reflex. Previous studies have demonstrated that individuals with motion sickness susceptibility often exhibit altered VEMP responses, suggesting that otolith organ sensitivity may serve as a candidate marker of motion sickness vulnerability^15–17. By incorporating VEMP measurements and validated questionnaires assessing motion sickness susceptibility, we explored potential correlations between objective physiological markers and subjective symptom reporting. This comprehensive approach may help identify candidate biomarkers for individual susceptibility to VR-induced discomfort and inform the development of personalised adaptation strategies for VR applications.

We hypothesized that VR exposure would induce significant changes in HRV parameters, reflecting ANS modulation in response to visual-vestibular conflict, and that the magnitude of these changes would correlate with subjective reports of discomfort and individual motion sickness susceptibility. Furthermore, we anticipated that PPG characteristics would show measurable differences based on an individual’s susceptibility to motion sickness, potentially offering a candidate biomarker for VR discomfort.

Materials and methods

Participants

A total of 30 participants (23 female) with a mean age of 37 ± 10 years were recruited for this prospective study conducted at Chungnam National University Hospital (CNUH) between January and February 2025. Participants were consecutive volunteers presenting to the outpatient neurology clinic and were not recruited based on vestibular symptom status. The female-predominant composition of the sample (76.7%) reflects the demographic profile of the outpatient neurology clinic population, consistent with the higher prevalence of vestibular symptoms and motion sickness susceptibility reported in women in the literature. All participants voluntarily presented at the medical institution. The study protocol was approved by the CNUH Institutional Review Board (IRB No. 2024-07-043-002), and written informed consent was obtained from all participants prior to enrollment. Inclusion criteria were: age 18–65 years, normal or corrected-to-normal vision , and capacity to provide written informed consent. Exclusion criteria were: active cardiovascular disease, use of vestibular-suppressant or cardiovascular medications, prior diagnosis of a vestibular disorder, acute illness on the day of testing, significant sleep deprivation or alcohol consumption in the preceding 24 h, and inability to tolerate head-mounted display use. All participants confirmed absence of acute exclusionary conditions via clinical interview on the day of testing.

Intervention

The experimental intervention consisted of three sequential phases as shown in Fig. 1. Initially, participants completed a comprehensive pre-test assessment battery to establish baseline motion sickness susceptibility and symptoms prior to VR exposure. This assessment included the Motion Sickness Susceptibility Scale (MSS-F for frequency (0-100%) and MSS-S for severity (0-20)), the Motion Sickness Assessment Questionnaire with multiple subscales (MSAQ-1 (0-3), MSAQ-A-1 (0-33), MSAQ-A-2 (0-44), MSAQ-A-3 (0-44), MSAQ-B-1 (0-33), MSAQ-B-2 (0-44), MSAQ-B-3 (0-44), MSAQ-sum representing the sum of scores of MSAQ-A-2,3,B-2, and 3, and MSAQ-sum/freq calculated as MSAQ total points divided by frequency), and the Headache Impact Test (HIT-6, 0-30)¹⁸. The complete questionnaire items for all assessment scales are provided in the Supplementary Material.

Fig. 1.

Study design and data processing workflow.

Following the pre-test assessment, participants underwent physiological monitoring via finger PPG using a BIOPAC MP160 data acquisition system (BIOPAC Systems Inc., Goleta, CA, USA) at 2000 Hz sampling rate. The PPG recording protocol began with a 60-second baseline period during which participants sat quietly in a controlled environment. Subsequently, participants were immersed in a virtual roller coaster experience (Epic roller coaster, B4T Games, Brazil) presented through an Oculus Quest 2 VR headset (Meta Platforms, Inc., USA) for 162 s. The VR content consisted of experiencing roller coasters, which had the potential to induce varying degrees of visual-vestibular conflict. Immediately following VR exposure, a 60-second recovery period was recorded while participants remained seated in the same position.

Upon completion of the physiological monitoring phase, participants answered additional post-exposure questions to assess acute symptoms induced by the VR experience the Virtual Reality Sickness Susceptibility Scale (VR-SSS-1 for severity 0-20 and VR-SSS-2 for symptoms 0-48).

The final component of the intervention consisted of VEMP using Nicolet-Biomedical equipment (Madison, WI, USA) testing conducted within the hospital setting to evaluate vestibular function and potential correlations with susceptibility to VR-induced symptoms. Both cVEMP and oVEMP were recorded by a single examiner to ensure consistency across all measurements.

For cVEMP recording, participants lay supine with the head raised approximately 30 and rotated to one side to contract the sternocleidomastoid muscle (SCM). Short burst of alternating tones (110 dB nHL, 123.5 dB SPL, 500 Hz, rise time = 2 ms, plateau = 3 ms, and fall time = 2 ms) were applied at 2.1 Hz frequency monaurally via headphones. The signal was sampled at 48 kHz, amplified, and bandpass-filtered at 30–1,500 Hz. No rectification or smoothing was performed during cVEMP response recording. Up to 80 stimuli responses were averaged for each test, with responses obtained at least twice for each ear and mean values calculated. The p13-n23 complex was identified and peak-to-peak amplitudes were measured. Absolute cVEMP amplitudes were normalised by dividing by the mean tonic activation of the SCM during recording. Peak latencies of p13 and n23 components were measured from stimulus onset. To compare the normalised p13-n23 amplitudes between the right and left sides, the interaural difference (IAD) was calculated as [100 (ARight - ALeft)/(Aright + Aleft)], where A represents the p13-n23 amplitudes.

For oVEMP recording, responses were elicited by tapping the hairline at the AFz using an electric reflex hammer (VIASYS Healthcare, Conshohocken, PA, USA). Bilateral responses were recorded simultaneously after applying up to 60 tapping stimuli at 2 Hz frequency with approximately 0.45 g of force. The responses were averaged for each test, and the average latency of the initial negative peak (n1) and n1-p1 amplitude were determined. oVEMP responses were obtained at least twice, and the mean was calculated. The IAD (%) of the oVEMP amplitudes was calculated as [100 (ARight - ALeft)/(Aright + Aleft)], where A represents the n1-p1 amplitude.

Data processing and analysis

Raw PPG signals were processed using custom algorithms developed in MATLAB R2023a (MathWorks, Natick, MA, USA). Signal preprocessing included artifact removal, baseline correction, and R-peak detection algorithms to accurately identify inter-beat intervals. HRV parameters were extracted for each experimental phase to assess ANS. These parameters included heart rate (HR) as an indicator of overall cardiovascular response, standard deviation of normal-to-normal intervals (SDNN) reflecting general autonomic adaptability, and root mean square of successive differences (RMSSD) representing parasympathetic nervous system activity^19,20.

SDNN was calculated as the standard deviation of all normal RR intervals using Eq. (1), where represents individual RR intervals, is the mean of all RR intervals, and N is the total number of RR intervals.

1

RMSSD was calculated as the root mean square of successive differences between adjacent RR intervals using Eq. (2), where and are consecutive RR intervals, and N-1 is the number of successive RR interval differences.

2

HRV analysis has emerged as a valuable non-invasive method for assessing ANS function, particularly the balance between sympathetic and parasympathetic activity^9,21.

Survey data from motion sickness susceptibility assessments and post-VR symptom evaluations were compiled and analyzed. MSAQ total scores were calculated according to established protocols. Participants were stratified into three motion sickness susceptibility categories based on MSAQ-sum score terciles (33rd and 67th percentiles): high susceptibility (MSAQ-sum over 24), medium susceptibility (MSAQ-sum 9 to 24), and low susceptibility (MSAQ-sum 0 to 8) groups^22–24.

VEMP data were processed following established clinical protocols with responses contaminated by artifacts or demonstrating inadequate signal-to-noise ratios excluded from analysis. All VEMP measurements were performed by a single experienced examiner to ensure consistency and minimize inter-rater variability.

Statistical analysis

Descriptive statistics including means, standard deviations, medians, and interquartile ranges were calculated for all continuous variables. Data distributions were assessed for normality using Shapiro-Wilk tests combined with visual inspection of histograms and quantile-quantile plots.

Within-subject changes in HRV parameters across the three experimental phases were evaluated using a 3 (Group: High/Medium/Low) 3 (Time: Baseline/VR/Recovery) mixed-design ANOVA, with Group as the between-subjects factor and Time as the within-subjects factor. Main effects of Group and Time, and the Group Time interaction were tested. Partial eta-squared ( ) was computed as a measure of effect size. Post-hoc pairwise comparisons for the Time factor used paired t-tests with Bonferroni correction; between-group comparisons at each time point used independent t-tests with Bonferroni correction.

Relationships between vestibular function measures derived from VEMP testing, autonomic responses quantified through HRV parameters, and subjective symptom reports were investigated using Pearson correlation coefficients. Correlation analyses were performed systematically across all variable pairs to identify associations between objective physiological markers and subjective symptom experiences. This included correlations between VEMP features and VR sickness scale scores, between vestibular function measures and motion sickness susceptibility scores, and between VEMP measurements and HRV parameters across experimental phases.

To address the risk of inflated false-positive findings arising from the large number of cross-domain correlation pairs, Benjamini-Hochberg False Discovery Rate (FDR) correction was applied to all 1,239 Pearson correlation p-values computed across the VEMP, PPG, and survey domains. No correlations survived FDR correction at ; all reported correlations are therefore interpreted as hypothesis-generating findings.

To identify the most important features in our dataset, we conducted correlation analyses between vestibular measures (VEMP latencies, amplitudes, IAD), physiological responses (PPG data), and subjective reports (previous motion sickness susceptibility, headache, and after VR exposure reports). Features were ranked based on the frequency of significant correlations with other variables throughout the dataset.

The significance threshold was maintained at = 0.05 for all statistical tests. All statistical analyses were performed using MATLAB R2023a and Python 3.12 (Python Software Foundation, Delaware, USA).

Results

A total of 30 participants completed the study protocol, consisting of 23 females (76.7%) and 7 males (23.3%) with a mean age of 37.0 ± 9.6 years. Baseline motion sickness susceptibility and VR-induced symptom scores are shown in Table 1.

Table 1.

Participant demographics and motion sickness assessment scores.

Characteristic	Mean (SD)	Median (IQR)	Range
Demographics
Sample size		—	—
Female, n (%)	23 (76.7)	—	—
Male, n (%)	7 (23.3)	—	—
Age (years)	37.0 (9.6)	34.0	24–58
Clinical measures
MSS-F	0.21 (0.28)	0.12 (0.00–0.25)	0.00–1.00
MSS-S	4.17 (4.56)	5.00 (0.00–8.75)	0.00–15.00
MSAQ-sum	19.60 (18.93)	12.50 (6.25–33.25)	0.00–80.00
MSAQ-sum/freq	1.75 (1.96)	1.04 (0.49–2.24)	0.00–8.76
HIT-6	13.13 (5.95)	12.00 (9.25–18.75)	2.00–24.00
VR-SSS-1	7.67 (4.69)	10.00 (5.00–10.00)	0.00–20.00
VR-SSS-2	10.53 (9.33)	8.00 (4.00–13.75)	0.00–36.00

MSS-F = Motion Sickness Susceptibility Scale–Frequency; MSS-S = Motion Sickness Susceptibility Scale–Severity; MSAQ = Motion Sickness Assessment Questionnaire; HIT-6 = Headache Impact Test–6; VR-SSS-1 = Virtual Reality Sickness Scale–Severity; VR-SSS-2 = Virtual Reality Sickness Scale–Symptoms; SD = standard deviation; IQR = interquartile range.

HRV responses to VR exposure

Following baseline characterization, we examined physiological responses to VR exposure through HRV analysis. The percentage changes in heart rate relative to different phases were calculated to quantify autonomic responses. The mean heart rate change during VR exposure compared to baseline was 3.67% (±5.68%). During recovery, heart rate decreased by 6.30% (±6.98%) compared to the VR phase. When comparing recovery to baseline, a mean reduction of 3.10% (±5.90%) in heart rate was observed.

The HRV parameters across baseline, VR exposure, and recovery phases are presented in Table 2. The mixed-design ANOVA revealed significant autonomic modulation during VR exposure, with a significant main effect of Time across all three metrics shown on Table 3.

Table 2.

HRV parameters across experimental phases.

Phase comparison	Phase 1 Mean (SD)	Phase 2 Mean (SD)	Mean difference	t	p
HR (bpm)
Baseline vs. VR exposure	79.09 (13.18)	81.66 (12.20)	2.58	– 3.18	0.004**
VR exposure vs. Recovery	81.66 (12.20)	76.33 (11.37)	– 5.34	4.84	< 0.001***
Baseline vs. Recovery	79.09 (13.18)	76.33 (11.37)	– 2.76	3.04	0.005**
SDNN (ms)
Baseline vs. VR exposure	51.37 (34.63)	92.49 (71.04)	41.12	– 3.04	0.005**
VR exposure vs. Recovery	92.49 (71.04)	66.77 (51.83)	– 25.72	2.07	0.047*
Baseline vs. Recovery	51.37 (34.63)	66.77 (51.83)	15.40	– 1.41	0.171
RMSSD (ms)
Baseline vs. VR exposure	55.92 (49.65)	99.42 (101.42)	43.50	– 2.28	0.030*
VR exposure vs. Recovery	99.42 (101.42)	68.16 (68.90)	– 31.26	1.74	0.092
Baseline vs. Recovery	55.92 (49.65)	68.16 (68.90)	12.25	– 0.83	0.414

Paired t-test comparisons of HRV parameters between experimental phases. HR = heart rate; SDNN = standard deviation of normal-to-normal intervals; RMSSD = root mean square of successive differences; SD = standard deviation. *, **, ***.

Table 3.

Mixed-design ANOVA results for HRV parameters: 3 (Group) 3 (Time).

Variable	Effect	F	df	p	Sig.
HR (bpm)	Time	15.318	(2, 54)	< 0.001	***	0.362
	Group	0.231	(2, 27)	0.795		0.017
	Group Time	0.517	(4, 54)	0.723		0.037
SDNN (ms)	Time	5.764	(2, 54)	0.005	**	0.176
	Group	0.514	(2, 27)	0.604		0.037
	Group Time	1.257	(4, 54)	0.298		0.085
RMSSD (ms)	Time	3.494	(2, 54)	0.037	*	0.115
	Group	0.442	(2, 27)	0.647		0.032
	Group Time	1.697	(4, 54)	0.164		0.112

Group = MSAQ susceptibility group (High/Medium/Low); Time = experimental phase (Baseline/VR/Recovery). = partial eta-squared (effect size). HR = heart rate; SDNN = standard deviation of normal-to-normal intervals; RMSSD = root mean square of successive differences. *, **, ***.

Motion sickness susceptibility groups and physiological responses

Participants were stratified into three motion sickness susceptibility categories based on MSAQ-sum scores. The stratification used clinically meaningful cut-offs: high susceptibility (n = 10, MSAQ-sum over 24), medium susceptibility (n = 9, MSAQ-sum 9 to 24), and low susceptibility (n = 11, MSAQ-sum 0 to 8) groups as shown in Fig. 2. These cut-offs created three balanced groups for comparison.

Fig. 2.

HRV parameters across experimental phases by MSAQ susceptibility group. Mean ± SEM for heart rate (HR), SDNN, and RMSSD across baseline, VR exposure, and recovery phases, stratified by motion sickness susceptibility group (High n = 10, Medium n = 9, Low n = 11). Brackets indicate the significant Time main effect from the mixed-design ANOVA; asterisks denote significance level (*, **, ***). The Group Time interaction was not significant for any metric (all ).

A 3 (Group: High/Medium/Low) 3 (Time: Baseline/VR/Recovery) mixed-design ANOVA was conducted to examine HRV parameter changes across experimental phases , as shown in Table 3. A significant main effect of Time was observed for all three metrics: HR (F(2,54) = 15.318, , = 0.362), SDNN (F(2,54) = 5.764, p = 0.005, = 0.176), and RMSSD (F(2,54) = 3.494, p = 0.037, = 0.115). The main effect of Group was not significant for any metric (HR: F(2,27) = 0.231, p = 0.795; SDNN: F(2,27) = 0.514, p = 0.604; RMSSD: F(2,27) = 0.442, p = 0.647). The Group Time interaction did not reach significance (HR: F(4,54) = 0.517, p = 0.723; SDNN: F(4,54) = 1.257, p = 0.298; RMSSD: F(4,54) = 1.697, p = 0.164), although the RMSSD interaction showed a medium effect size ( = 0.112), suggesting the study may have been underpowered to detect differential autonomic response patterns across susceptibility groups.

Vestibular function and VR sickness correlations

Pearson correlation analysis was conducted between VEMP features and VR sickness scale scores. VR sickness symptoms (VR-SSS-2) showed correlations with right-side ocular VEMP hammer amplitudes (, ), left-side ocular VEMP hammer amplitudes (, ), and left-side cervical VEMP amplitudes (, ). VR sickness severity (VR-SSS-1) showed a significant correlation with right-side ocular VEMP hammer amplitudes (, ).

Motion sickness susceptibility frequency (MSS-F) showed a significant correlation with cervical VEMP interaural difference asymmetry (, ). The correlation between MSS-F and cervical VEMP amplitude average approached statistical significance (, ). No significant correlations were observed between MSAQ-sum scores and cervical VEMP latency parameters (P13 latency: , ; N23 latency: , ) as seen on Fig. 3.

Fig. 3.

Relationship between cVEMP amplitude, latency, and asymmetry measures with subjective dizziness scores. Scatter plots showing correlations between cVEMP parameters and motion sickness scores. (A) Correlation between average cVEMP amplitude and MSS-F scores (, ). (B) Correlations between MSAQ-sum scores and cVEMP latencies: P13 latency (circles, solid line, , ) and N23 latency (triangles, dashed line, , ). (C) Correlation between absolute cVEMP inter-aural difference and MSS-F scores (, ). Lines show linear regression with 95% confidence intervals. MSS-F = Motion Sickness Susceptibility - Frequency; MSAQ = Motion Sickness Assessment Questionnaire.

Vestibular function and PPG parameter correlations

To explore the relationship between vestibular function and autonomic responses during VR exposure, Pearson correlation analysis was conducted between VEMP parameters and HRV measures, as presented in Table 4.

Table 4.

Significant correlations between VEMP and HRV parameters

VEMP variable	HRV variable	r	p
oVEMP Hammer amplitude (R)	Heart Rate SD (VR)	0.511**	0.004
oVEMP Hammer amplitude (L)	RMSSD (VR)	0.392*	0.032
oVEMP Air P13 latency (L)	RMSSD (VR)	– 0.505**	0.004
	pNN50 (VR)	– 0.496**	0.005
	SDNN (VR)	– 0.479**	0.007
oVEMP Hammer amplitude (R)	RMSSD (VR)	0.473**	0.008
oVEMP Hammer amplitude (L)	Heart Rate SD (VR)	0.409*	0.025
oVEMP Hammer O latency (L)	Ratio VR/Recovery	0.408*	0.025
cVEMP P13 latency (R)	Ratio VR/Recovery	0.397*	0.030
	Change Recovery-VR	– 0.394*	0.031

oVEMP = ocular vestibular evoked myogenic potential; cVEMP = cervical vestibular evoked myogenic potential; RMSSD = root mean square of successive differences; SDNN = standard deviation of NN intervals; pNN50 = percentage of NN intervals > 50 ms; HR SD = heart rate standard deviation; R = right ear; L = left ear. *, **.

The strongest correlations were observed between ocular VEMP hammer right peak-to-peak amplitude and heart rate standard deviation during VR (, ), and between ocular VEMP air P13 latency (left) and RMSSD during VR (, ). Right-sided ocular VEMP hammer amplitude measures showed positive correlations with HRV parameters during VR exposure. Left-sided ocular VEMP air P13 latency demonstrated negative correlations with multiple parasympathetic indicators during VR: RMSSD (, ), pNN50 (, ), and SDNN (, ).

Cross-domain feature correlations

Correlation analysis was conducted across all vestibular, physiological, and survey measures to identify variables with the highest number of significant correlations. Figure 4 shows the top 10 features ranked by frequency of significant correlations.

Fig. 4.

Most frequently correlated variables across all domains. Features were ranked by their frequency of significant correlations with variables from other domains (uncorrected ). This ranking is an exploratory descriptive heuristic and should not be interpreted as a formal measure of predictive importance. No correlations survived Benjamini-Hochberg FDR correction; counts therefore reflect uncorrected exploratory associations only.

The right-sided ocular VEMP hammer amplitude had 14 significant correlations, followed by Ratio VR/Recovery (11 correlations) and MSAQ-B-3 (11 correlations). Change Recovery VR had 10 correlations, HIT-6 had 9 correlations, and left-sided ocular VEMP hammer amplitude had 8 correlations. Change RMSSD Recovery VR had 8 correlations, oVEMP air latency-R-P13 had 7 correlations, MSAQ-A-3 had 7 correlations, and Ratio Recovery Baseline had 6 correlations.

Discussion

Our PPG analysis revealed significant autonomic responses to VR exposure consistent with sympathetic activation during VR immersion. Heart rate increased significantly during VR compared to baseline (mean increase: 2.58 bpm, ), representing a 3.67% increase from baseline values. More pronounced was the substantial heart rate decrease during recovery (-5.34 bpm from VR phase, ), resulting in values significantly lower than baseline (-2.76 bpm, ). This pattern indicates parasympathetic rebound following VR withdrawal, suggesting robust autonomic regulation in response to visual-vestibular conflict. HRV parameters showed marked increases during VR exposure, with SDNN rising by 41.12 ms () and RMSSD increasing by 43.50 ms () compared to baseline.

The concurrent increase in heart rate and HRV parameters (SDNN and RMSSD) during VR exposure warrants specific discussion, as these patterns may appear contradictory given that SDNN and RMSSD are conventionally interpreted as indices of parasympathetic activity. Several non-exclusive mechanisms likely account for this observation. Respiratory sinus arrhythmia (RSA), the primary short-term driver of RMSSD, is substantially amplified by the deep and irregular breathing associated with emotional arousal during immersive VR, which would inflate RMSSD independently of the sympatho-vagal balance. Additionally, intense sensory stimuli can produce dual autonomic co-activation, with simultaneous sympathetic and parasympathetic engagement rather than simple reciprocal inhibition. Finally, the short recording windows used in this study mean that RMSSD and SDNN estimates are sensitive to transient respiratory and movement artifacts inherent to the VR condition. Future studies should incorporate respiratory monitoring to disentangle vagal and respiratory contributions to HRV during VR exposure.

Previous research has documented changes in cardiovascular parameters during VR exposure, but findings have been inconsistent across studies due to variations in VR content, exposure duration, and participant characteristics^25,26. Intense visual stimuli, such as virtual roller coasters, may be particularly effective at inducing strong autonomic responses due to their simulation of acceleration, drops, and rotational movements^27,28. These cardiovascular changes could serve as accessible candidate biomarkers of autonomic activation during VR experiences, particularly valuable given the widespread availability of PPG sensors in consumer wearable devices.

The differential physiological responses observed between motion sickness susceptibility groups provide important insights into individual differences in VR tolerance. The mixed-design ANOVA confirmed a significant main effect of Time across all HRV metrics, while the Group Time interaction did not reach significance. However, the RMSSD interaction showed a medium effect size ( = 0.112, p = 0.164), suggesting the study may have been underpowered to detect differential autonomic response patterns across susceptibility groups. Future studies with larger samples should reexamine this interaction. Descriptively, the high susceptibility group demonstrated the most pronounced autonomic responses . This heightened autonomic reactivity in motion-sensitive individuals aligns with previous research suggesting enhanced physiological sensitivity in those prone to motion sickness¹¹. These findings suggest a relationship between autonomic system sensitivity and motion sickness susceptibility, with autonomic response patterns differing qualitatively between susceptibility groups, as supported by previous studies¹¹.

Our correlation analyses revealed significant associations between vestibular function measures and VR-induced symptoms, providing evidence for the role of peripheral vestibular status in VR tolerance. The strongest correlations were observed between VR sickness symptoms (VR-SSS-2) and bilateral ocular VEMP hammer amplitudes (right: , ; left: , ), suggesting that otolith organ sensitivity is particularly relevant for VR-induced discomfort. The significant correlation between motion sickness susceptibility frequency (MSS-F) and cervical VEMP interaural asymmetry (, ) aligns with previous research by Fowler et al.²² demonstrating relationships between vestibular function and motion sickness susceptibility. This asymmetry measure may serve as an associated physiological marker of individual susceptibility to visually induced motion sickness, with greater vestibular imbalance associated with increased symptom severity.

It is important to note that VEMP responses reflect stable, trait-level otolith function rather than a direct electrophysiological response to VR stimulation. Within the framework of sensory conflict theory, individuals with higher baseline otolith sensitivity, reflected in larger VEMP amplitudes, may generate stronger visuo-vestibular mismatch signals when exposed to compelling visual motion cues without corresponding vestibular input. Head-mounted VR creates a high level of vection that generates a strong expectation of vestibular stimulation; when this expected input is absent, the resulting sensory conflict is more pronounced in those with heightened vestibular sensitivity, leading to more severe symptoms. This interpretation is consistent with our finding that oVEMP amplitude correlates with VR symptom severity even in the absence of direct vestibular stimulation during the VR task.

It is also important to clarify why VR exposure specifically may provoke motion sickness in the absence of vestibular stimulation, whereas conventional video viewing typically does not. Head-mounted VR creates a compelling and immersive sense of vection, generating a strong expectation of corresponding vestibular input. When this expected input is absent, the resulting visuo-vestibular conflict is sufficient to trigger autonomic and symptomatic responses consistent with motion sickness. The wide field of view and high perceptual immersion of head-mounted displays are the key factors that differentiate VR from conventional 2D video viewing, which does not reliably generate vection of sufficient magnitude to produce sensory conflict.

The predominance of ocular VEMP correlations over cervical VEMP associations with VR symptoms reflects the specific nature of visually induced motion sickness (VIMS). Unlike conventional motion sickness involving physical acceleration, VR exposure primarily activates the vestibulo-ocular reflex through visual-vestibular conflict. The oVEMP responses, which assess utricular and superior vestibular nerve function, are therefore more directly relevant to VIMS than saccular responses measured by cVEMP. Our study demonstrated meaningful associations between oVEMP findings and VIMS symptoms. The relatively weak relationship with cVEMP responses may be associated with conventional motion sickness components or potentially influenced by the physical weight effect of the head-mounted VR device.

The significant correlations between VEMP measurements and HRV parameters reveal important vestibular-autonomic interactions during VR exposure. The strongest relationship was observed between right-sided ocular VEMP hammer amplitudes and heart rate standard deviation during VR (, ), suggesting that individuals with more robust vestibular responses exhibit greater autonomic reactivity to visual-vestibular conflict. Left-sided ocular VEMP air P13 latency showed negative correlations with multiple parasympathetic indicators during VR, including RMSSD (, ), pNN50 (, ), and SDNN (, ). These vestibular-autonomic relationships provide physiological support for the concept of vestibulosympathetic reflexes, which have been documented in previous studies of vestibular disease. Though rare, vestibulopathy patients report vestibular syncope, which could represent transient autonomic failure due to vestibular imbalance²⁹. Our study suggests that vestibular response and autonomic reaction can determine VIMS susceptibility during VR exposure, extending this understanding from clinical vestibular conditions to visually induced motion sickness.

Correlation analyses did not survive Benjamini-Hochberg FDR correction for multiple comparisons (1,239 cross-domain pairs tested), consistent with the exploratory nature and small sample size of this study; all reported correlations are therefore interpreted as hypothesis-generating findings requiring confirmation in larger, adequately powered studies. Multivariate approaches such as multiple regression and network analysis, which better account for shared variance among physiological indices, are recommended for future work with larger cohorts.

Several methodological considerations specific to oVEMP measurement should be acknowledged. The mechanical tap stimulus used in this study has inherent trial-to-trial variability in stimulus intensity, which may introduce measurement noise into amplitude estimates. Furthermore, unlike cVEMP, no normalisation for muscle activation level was performed for oVEMP recordings, meaning that absolute amplitude values may partially reflect differences in muscle activation rather than true vestibular sensitivity. These factors should be considered when interpreting the oVEMP-symptom correlations, and future studies are encouraged to adopt controlled vibration stimuli and standardised gaze elevation protocols to improve measurement consistency.

Clinical implications and future directions

The feature correlation analysis identified several measures with high interconnectivity across physiological and subjective domains. Right-sided ocular VEMP hammer amplitude emerged as the most frequently correlated variable (14 uncorrected correlations), followed by Ratio VR/Recovery and MSAQ-B-3 (11 correlations each) . These counts are based on uncorrected p-values and serve as an exploratory descriptive heuristic; as noted above, no correlations survived FDR correction. The practical implications of these findings are significant for VR application development and user experience optimization. PPG monitoring offers a non-invasive, accessible method for real-time assessment of user physiological states during VR experiences, with widespread availability in consumer wearable devices making this approach particularly practical for implementation in consumer VR systems. Vestibular function assessment, particularly ocular VEMP testing, may provide valuable information for identifying individuals at risk for VR-induced discomfort, potentially informing personalised VR adaptation strategies as a candidate biomarker worthy of further investigation in larger samples.

Several limitations should be considered when interpreting these findings. Our sample size of 30 participants, while sufficient for detecting the reported effects, may have limited statistical power for identifying more subtle correlations or subgroup differences. The recovery period was relatively short (60 s), potentially limiting characterisation of complete physiological recovery patterns following VR exposure. The focus on a single type of VR content (roller coaster simulation) presents another constraint, as different VR applications with varying visual intensity, motion patterns, and cognitive demands may elicit different physiological responses. The female-predominant sample (76.7%) limits generalisability to male populations; future studies should aim for balanced sex representation. Participants were recruited as consecutive volunteers at a single academic medical centre, which may have introduced selection bias toward individuals with greater health-related concerns or vestibular awareness; the distribution of motion sickness susceptibility in our sample may not reflect population-level distributions. The absence of concurrent respiratory monitoring limits the interpretation of HRV parameters, as respiratory sinus arrhythmia may have contributed to the observed increases in RMSSD and SDNN during VR exposure independently of autonomic state. Additionally, oVEMP findings reflect a spectrum of physiological variation in the general population and do not necessarily indicate vestibular pathology.

Future research should explore extended monitoring periods to better characterise recovery dynamics and investigate diverse VR content types to establish generalisability across different virtual environments. Larger sample sizes would enable more robust subgroup analyses and potentially reveal additional predictive relationships. The integration of real-time physiological monitoring with adaptive VR systems represents a promising avenue for reducing VR-induced discomfort, with systems that adjust content intensity or provide comfort interventions based on physiological indicators potentially improving user experience and expanding VR accessibility.

In conclusion, VIMS, especially VR exposure-induced motion sickness, can be associated with individual vestibular traits and previous conventional motion sickness experience. Real-time monitoring of autonomic reactions with non-invasive PPG monitoring can predict symptoms and may prevent uncomfortable feelings in future VR applications. The identification of candidate biomarkers and the feasibility of real-time physiological monitoring provide foundations for developing personalised approaches to VR comfort and safety.

Author contributions

C.A.C.P., E.K., and Y.K. conceived the experiment(s). C.A.C.P., E.K., and Y.K. conducted the experiment(s). C.A.C.P., E.K., and Y.K. analysed the results. C.A.C.P. prepared the original draft. C.A.C.P., E.K., Y.K., and S.H.J. reviewed and edited the manuscript. E.K. and S.H.J. provided resources. E.K. acquired funding and administered the project. E.K. and Y.K. supervised the study. All authors approved the final manuscript.

Funding

This work(research) was supported by Chungnam National University Hospital Research Fund, 2023-CF-023.

Data availability

The datasets used and analyzed during the current study are available from the corresponding authors on reasonable request.

Competing interests

The authors declare no competing interests.

Ethical approval

This study followed the tenets of the Declaration of Helsinki and was performed according to the guidelines of the Institutional Review Board (IRB 2024-07-043-002) of the Chungnam National University Hospital, Daejeon, South Korea. Informed consent was obtained from all participants prior to their involvement in the study.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-50827-1.