Source-linked AI summary
Gaze Under Strain: How Oculomotor Behavior Tracks Cybersickness Progression in VR
Nevzat Umut Demirseren, Mehedi Hasan Raju, Oleg Komogortsev, Isayas Berhe Adhanom
TL;DR
How gaze behavior evolves alongside discomfort during prolonged VR exposure remains poorly understood, despite its relevance to visual processing and oculomotor control. This study models repeated discomfort ratings and eye-tracking measures during a 20-minute realistic VR navigation session. Fixation, saccadic behavior, and gaze allocation show distinct relationships with discomfort and exposure time.
Problem
How eye-movement behavior changes alongside discomfort during prolonged VR exposure remains underexplored, limiting behavioral understanding of cybersickness progression.
Source: §1 · p. 1, §2.4 · p. 3Method
Results
Fixation duration decreases with increasing discomfort and more strongly over time; saccade amplitude and velocity show time-dependent discomfort relationships, while gaze eccentricity varies with exposure duration rather than discomfort.
Source: §7 · p. 8Takeaways & Limitations
Different oculomotor measures show distinct temporal responses, so characterizing gaze alongside repeated discomfort ratings captures patterns that aggregate or post-exposure approaches cannot resolve.
Source: §1 · p. 2, §7 · p. 8Because discomfort accumulates with elapsed time, the observational design cannot fully separate cybersickness-specific changes from time-on-task effects such as fatigue or adaptation.
Source: §6.4 · p. 8
Abstract
from arXiv · showhide
Cybersickness remains a major barrier to comfortable, accessible virtual reality (VR). Arising from mismatches between visual and vestibular signals, it places sustained demands on gaze stabilization and visual processing. Because gaze is the mechanism through which users actively sample the virtual environment, cybersickness-induced disruptions may alter both what users perceive and how discomfort itself emerges. Yet, how eye-movement behavior evolves alongside discomfort during prolonged VR exposure remains poorly understood. We present a behavioral analysis of eye-tracking data from 19 participants navigating a realistic virtual environment over a 20-minute session, with discomfort rated repeatedly throughout the experiment. Linear mixed-effects models examined gaze constructs spanning fixation stability, visual exploration, and gaze allocation as functions of discomfort, exposure duration, and their interaction. Our results show that fixation stability degraded progressively with increasing discomfort, with the effect strengthening over time. Visual exploration showed no marginal discomfort effect but exhibited significant discomfort-by-time interactions, indicating that saccadic behavior adapts dynamically rather than responding uniformly to discomfort. Gaze allocation was not associated with discomfort but shifted systematically with time. These findings show that cybersickness does not uniformly disrupt gaze. Instead, different aspects of eye-movement behavior respond differently and change in ways that only become visible when discomfort is tracked repeatedly over time. Understanding how gaze evolves during cybersickness progression provides the behavioral grounding needed to better interpret how discomfort shapes the user's visual experience of immersive environments and to design more sensitive detection approaches.
1 Introduction
The introduction identifies a gap in understanding how gaze behavior changes alongside discomfort during prolonged VR exposure. It presents a time-resolved analysis to characterize those changes beyond prediction or aggregated measures.
1 Introduction
- Gaze matters because it is how users sample a virtual scene, linking oculomotor disruption to both perception and discomfort.Cybersickness is closely tied to visual motion processing and oculomotor control.Source: §1 · p. 1
- Prior work leaves open how gaze behavior changes in step with discomfort when ratings are collected repeatedly throughout VR exposure.Existing approaches often emphasize prediction, session-aggregated features, condition contrasts, or post-exposure scores.Source: §1 · pp. 1–2
- The study analyzes eye tracking during a realistic 20-minute VR navigation session with repeated discomfort ratings, modeling gaze metrics over time.It examines both marginal discomfort relationships and how they change with continued exposure.Source: §1 · p. 2
- Fixation duration shows a strengthening negative association with discomfort, while saccadic exploration has discomfort-by-time interactions and gaze allocation shifts with exposure duration.These time-varying patterns motivate behaviorally grounded interpretation beyond prediction and aggregate-feature approaches.Source: §1 · p. 2
2 Background
The background describes proposed mechanisms of cybersickness, common assessment approaches, and prior evidence connecting gaze to discomfort. It highlights the limited understanding of how event-derived gaze measures evolve alongside repeated discomfort ratings over time.
2.1 Causes of Cybersickness
- No single theory fully explains cybersickness, though established frameworks offer perspectives on its underlying mechanisms.The section introduces sensory conflict, postural instability, and eye-movement accounts.Source: §2.1 · p. 2
2.1.1 Sensory Conflict Theory.
- Sensory Conflict Theory proposes that discomfort can arise when visual motion cues conflict with vestibular and proprioceptive signals.In VR, visually perceived self-motion without corresponding physical motion can produce intersensory conflict.Source: §2.1.1 · p. 2
2.1.2 Postural Instability Theory.
- Postural Instability Theory attributes cybersickness to prolonged difficulty maintaining postural control, with symptoms persisting until stable control returns.The account holds that balance instability precedes sickness onset.Source: §2.1.2 · p. 2
2.1.3 Eye Movement Theory.
- Eye Movement Theory proposes that excessive oculomotor strain during visually induced eye movements may contribute to motion-sickness symptoms.Prolonged visual motion may increase demands on gaze stabilization and eye-movement control.Source: §2.1.3 · p. 2
2.2.1 Subjective Measures of Cybersickness.
- Subjective instruments assess cybersickness, but self-report may interrupt immersion and offers limited insight into real-time symptom development.The SSQ is a post-exposure measure, while FMS and discomfort questionnaires can support repeated assessments during exposure.Source: §2.2.1 · p. 2
2.2.2 Objective Measures of Cybersickness.
- Physiological and behavioral signals provide objective correlates of discomfort, but many studies use multimodal measures to predict sickness rather than characterize evolving processes.Examples include EDA, postural sway, body temperature, blink rate, and heart rate.Source: §2.2.2 · p. 2
2.3 Eye Movements and Cybersickness
- Eye-tracking measures have been associated with discomfort and user state, yet prior VR work has often focused on prediction, coarser temporal analyses, or specific metric associations.Fixations and saccades offer interpretable units for examining visual sampling, but their evolution with repeated discomfort ratings has received comparatively less attention.Source: §2.3 · pp. 2–3
2.4 Temporal Dynamics of Gaze Behavior and Cybersickness
- Fine-grained temporal relationships between gaze and evolving discomfort remain understudied because much prior work uses session summaries or prediction-focused gaze signals.Repeated within-session ratings can support examination of how event-derived gaze metrics change alongside discomfort during prolonged exposure.Source: §2.4 · p. 3
3 Methodology
The methodology derives fixation- and saccade-based gaze measures from eye tracking, aligns them with repeated discomfort ratings, and models their relationships over exposure. Measures cover fixation stability, visual exploration, and gaze allocation.
3 Methodology
- The analysis derives gaze metrics from eye-movement events, aggregates them within discomfort-rating intervals, and relates them to repeated ratings over VR exposure.The pipeline detects fixations and saccades before deriving the metrics used in the analysis.Source: §3 · p. 3
3.1 Eye-Tracking Data Recording
- Binocular eye tracking was recorded at 120 Hz, and head-centered gaze vectors capture eye-in-head orientation independently of HMD pose.The resulting event metrics are not confounded by rigid-body head rotations or HMD translation.Source: §3.1 · p. 3
3.2 Eye Gaze Data Processing Pipeline
- The processing pipeline removes invalid samples, interpolates gaps under 75 ms, smooths gaze, and classifies events using a velocity-threshold algorithm.Saccades exceed 35°/s; fixation segments use duration limits of 60–2000 ms, while saccades require 10–150 ms duration and at least 1° amplitude.Source: §3.2 · pp. 3–4
3.3 Eye Gaze Metrics
- Five event-derived metrics were selected across fixation stability, visual exploration, and gaze allocation; event counts and sequence measures were not analyzed.The selected measures target interpretable changes in fixation- and saccade-level magnitudes.Source: §3.3 · p. 4
3.3.1 Fixation Stability.
- Fixation duration measures how long gaze stays settled, while fixation dispersion measures the angular spread of gaze samples within a fixation.Together they distinguish temporal persistence from spatial precision of fixation.Source: §3.3.1 · p. 4
3.3.2 Visual Exploration.
- Saccade amplitude measures the spatial extent of gaze shifts, while saccade velocity measures the speed of transitions.These measures separate exploration breadth from exploration dynamics.Source: §3.3.2 · p. 4
3.3.3 Gaze Allocation.
- Gaze eccentricity is angular distance from a forward-facing reference; lower values indicate centerward gaze and higher values indicate peripheral gaze.Metrics were averaged in non-overlapping 60-second windows aligned with discomfort-rating intervals.Source: §3.3.3 · p. 4
4 Dataset
The study reuses a published VR locomotion dataset selected for continuous eye tracking and repeated discomfort ratings during prolonged exposure. The present analysis restricts the data to one realistic environment and its first session.
4 Dataset
- The analysis uses a previously published VR locomotion dataset, with broader experimental procedures reported in the original study.The dataset was selected for prolonged exposure, one-minute discomfort ratings, and continuous eye tracking during naturalistic locomotion.Source: §4 · p. 4
4.1 Experiment Design
- The source experiment involved waypoint navigation using controller-based locomotion and headsteered navigation, with sessions lasting up to 20 minutes.The present analysis selected the realistic environment from the first session, limiting analysis to one continuous navigation task and avoiding cross-session adaptation and habituation effects.Source: §4.1 · pp. 4–5
4.2 Procedure
Participants navigated the virtual environment for approximately 20 minutes while providing repeated discomfort ratings. The procedure allowed free gaze behavior during immersion.
4.2 Procedure
- Participants navigated for approximately 20 minutes, rating discomfort once per minute on a 0–10 slider that disappeared after each response.They could rotate their heads but did not physically walk; movement used a controller and steering used head rotation.Source: §4.2 · p. 5
- The environment imposed no stimulus-locked tracking task, allowing free gaze behavior throughout navigation.Source: §4.2 · p. 5
- Figure 1 presents representative screenshots of the virtual environment used in the experiment.Source: Figure 1 · p. 5
4.3 Participants
Of 22 participants who took part, data from 19 who completed the study were included in the analysis.
4.3 Participants
- Data from 19 participants were included: 11 male and 8 female, with mean age 22.8 years (SD = 3.6).Source: §4.3 · p. 5
- The study initially involved 22 participants: 11 male and 11 female, with mean age 23.45 years (SD = 3.9).Source: §4.3 · p. 5
- The study received Institutional Review Board approval.Source: §4.3 · p. 5
4.4 Measures
Cybersickness was measured using repeated in-session discomfort ratings and pre- and post-exposure SSQ scores.
4.4 Measures
- Participants rated discomfort on a 0–10 scale once per minute, preserving time-resolved measurements rather than aggregating ratings by session or trial.Source: §4.4.1 · p. 5
- Pre- and post-exposure SSQ scores assessed overall cybersickness severity across the session.Source: §4.4.1 · p. 5
- The SSQ oculomotor subscale was extracted for a convergent-validity analysis.Source: §4.4.1 · p. 5
4.5 Hypotheses
The hypotheses address fixation stability, visual exploration, and gaze allocation as discomfort rises, while treating changes across exposure duration as exploratory.
4.5 Hypotheses
- Prior associations relied largely on aggregate, session-level, or predictive analyses, leaving their short-window observability and evolution during prolonged exposure open.The temporal evolution of discomfort–gaze relationships was examined through interactions with exposure duration.Source: §4.5 · p. 5
4.5.1 Fixation Stability.
- As discomfort rises, fixations were hypothesized to become shorter and more dispersed, consistent with reduced temporal persistence and spatial precision [15].Source: §4.5.1 · p. 5
4.5.2 Visual Exploration.
- If discomfort reduces visual exploration, saccade amplitude and velocity were hypothesized to decrease.Source: §4.5.2 · p. 5
4.5.3 Gaze Allocation.
- If discomfort increases central visual attention, gaze eccentricity was hypothesized to decrease.Source: §4.5.3 · p. 5
5 Results
Across prolonged VR exposure, fixation duration became more negatively associated with discomfort, while saccadic relationships changed over time and gaze allocation shifted with exposure rather than discomfort. Post-exposure SSQ scores were not associated with any gaze metric.
5 Results
- Linear mixed-effects models tested each gaze metric against discomfort, exposure duration, and their interaction, with participant random intercepts and Holm–Bonferroni-adjusted p-values.Because participants followed the same waypoint route, exposure duration absorbed shared scene-related structure; discomfort effects were identified by comparing ratings at matched exposure.Source: §5 · pp. 5–6
5.2 Visual Exploration (H2)
- Saccade amplitude and velocity had no reliable marginal discomfort effects, but both showed positive discomfort × exposure interactions as early negative associations weakened and tended to reverse.Both interaction effects had βstd = 0.12.Source: Figure 2 · p. 6, §5.2 · p. 6
5.4 Convergence with SSQ Oculomotor Symptoms
- No gaze metric correlated with post-exposure SSQ oculomotor scores, which index overall session severity rather than within-session dynamics.The authors interpret the null correlations as distinct from the event-window analyses that track within-session progression.Source: §5.4 · p. 6
5.1 Fixation Stability (H1)
- Fixation duration shortened with discomfort, increasingly so over exposure, while fixation dispersion showed neither a discomfort effect nor an exposure-related change.The discomfort × exposure interaction for fixation duration was significant and negative (β = −0.0012, βstd = −0.12); dispersion showed no measurable spatial broadening.Source: §5.1 · p. 6, Figure 3 · p. 7
5.3 Gaze Allocation (H3)
- Gaze eccentricity increased with exposure duration but had no association with discomfort or discomfort-by-exposure interaction.The exposure-duration effect was β = 0.056, p = .001.Source: §5.3 · p. 6, Figure 3 · p. 7
6 Discussion
During prolonged VR exposure, fixation duration’s negative association with discomfort strengthened over time, while saccadic relationships changed with exposure and gaze eccentricity shifted with time rather than discomfort. These distinct patterns suggest cybersickness does not uniformly affect gaze, but the single-task observational design limits generalization and cannot fully separate discomfort from other time-on-task processes.
6.1 Fixation Stability
- Higher discomfort was associated with shorter fixation duration, and this negative association strengthened as exposure progressed; fixation dispersion showed no discomfort association or change over time.The contrast suggests a specific reduction in fixation persistence rather than generalized spatial instability.Source: §6.1 · p. 7
6.2 Visual Exploration
- Saccade amplitude and velocity showed no reliable marginal discomfort association, but both had significant positive discomfort-by-exposure interactions (βstd = 0.12), indicating changing relationships over time.The interaction does not establish the relationship’s sign or significance at any particular exposure stage.Source: §6.2 · p. 7
6.3 Gaze Distribution
- Gaze eccentricity increased with exposure duration, not discomfort; Figure 4 shows gaze density becoming more dispersed from the early to late session.The authors offer declining organization of visual scanning as a plausible, unconfirmed interpretation.Source: §6.3 · pp. 7–8, Figure 4 · p. 8
6 Discussion
- Across measures, gaze showed distinct temporal responses rather than one global disruption, and time-resolved analysis revealed dynamics that aggregate or session-level analyses could obscure.Source: §6.3 · p. 8
6.4 Limitations and Future Work
- The findings come from 19 participants in one VR environment, navigation task, and locomotion paradigm, limiting generalization to other VR experiences.Larger, more diverse samples and additional environments and locomotion methods are proposed to test robustness.Source: §6.4 · p. 8
- Because discomfort accumulates alongside elapsed time, this observational design cannot fully distinguish cybersickness-specific changes from fatigue or adaptation.The authors suggest varying cybersickness-related stimulation independently of exposure duration.Source: §6.4 · p. 8
- Without intermediate recalibration, measurement error, calibration drift, or headset slippage cannot be excluded as contributors to the observed patterns.The authors note that spatial metrics would be most vulnerable, while the central discomfort effect was confined to fixation duration.Source: §6.4 · p. 8
- Future work could test nonlinear models and investigate these temporal gaze patterns for real-time cybersickness prediction and adaptive VR systems.Source: §6.4 · p. 8
7 Conclusion
The study found that gaze metrics relate to discomfort differently over prolonged VR exposure: fixation duration decreased with discomfort more strongly over time, while saccadic measures showed time-dependent relationships and gaze eccentricity shifted with exposure duration. These results support examining interpretable gaze metrics alongside repeated discomfort measures to characterize cybersickness progression.
7 Conclusion
- Fixation duration decreased with increasing discomfort, with the relationship strengthening over time; saccade amplitude and velocity had time-dependent discomfort relationships, while fixation dispersion stayed stable and gaze eccentricity tracked exposure duration rather than discomfort.Source: §7 · p. 8