Reactive and predictive processes during unpredictable driving hazards in virtual reality: an exploratory brain and body study with multimodal neurophysiological monitoring
- 1The Brain and Cognition research center (CerCo), CNRS
Abstract
Anticipatory neurophysiological activity before unpredictable emotional events has been reported for decades, but the effects are small, difficult to replicate, and obtained almost exclusively in stripped-down laboratory paradigms. Whether they reflect genuine predictive processes or methodological artifacts remains debated; one rarely tested possibility is that such settings mute it. We examined anticipation in an immersive virtual reality (VR) driving simulation whose collision outcomes were assigned by a quantum random number generator, while EEG, photoplethysmography (PPG), EMG, EOG and inertial (IMU) signals were recorded from a single wireless dry-electrode headset built into the VR display. Sixteen participants passively observed 120 driving sequences ending in a collision or no collision. Length-matched pre- and post-stimulus windows were analysed identically with mass-univariate hierarchical general linear models under permutation cluster correction, plus machine-learning classification across held-out participants. Two post-stimulus clusters differentiated the conditions, the largest peaking at 451.6 ms (d = -1.35), with broadband power modulation robust to baseline modelling, decoded from held-out participants with up to 93.8% accuracy. No time-domain effect or decodable information preceded stimulus onset, but a broadband spectral difference did, surviving preregistered controls for expectation bias (run length), time-on-task build-up and ocular or facial-muscle artifact: directionally consistent in 13 of 15 participants, with pre- and post-stimulus power coupled in all 16. Heart rate differentiated the conditions neither alone nor when added to the classifier. The study was preregistered; the BIDS-format data, analysis code and an open-source EEGLAB plugin are publicly available.
Related
- Multimodal computer vision and machine learning for predictive monitoring of planar-flow casting of amorphous steel foils
- Unpredictable chronic stress redistributes novelty-evoked exploratory space use and yields method-dependent identification of resilient-like archetypes in zebrafish
- Individualized brain-body signatures of positive emotions revealed by large-scale multimodal neurophysiological recording
- Metabolite-centric Knowledge Graph Unveils Predictive and Mechanistic Insights through Self-supervised Multimodal Learning
- A multimodal deep learning for rice crop health monitoring using vegetation index feature selection
BibTeX
@misc{cannard2026reactive,
title = {{Reactive and predictive processes during unpredictable driving hazards in virtual reality: an exploratory brain and body study with multimodal neurophysiological monitoring}},
author = {C. Cannard and D. Demet Yesilbas},
howpublished = {bioRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.24.754154},
url = {https://www.biorxiv.org/content/10.64898/2026.09.24.754154v1},
note = {CC BY-NC},
}
RIS
TY - UNPB
TI - Reactive and predictive processes during unpredictable driving hazards in virtual reality: an exploratory brain and body study with multimodal neurophysiological monitoring
AU - C. Cannard
AU - D. Demet Yesilbas
PY - 2026
DA - 2026/09/24
PB - bioRxiv
DO - 10.64898/2026.09.24.754154
UR - https://www.biorxiv.org/content/10.64898/2026.09.24.754154v1
AB - Anticipatory neurophysiological activity before unpredictable emotional events has been reported for decades, but the effects are small, difficult to replicate, and obtained almost exclusively in stripped-down laboratory paradigms. Whether they reflect genuine predictive processes or methodological artifacts remains debated; one rarely tested possibility is that such settings mute it. We examined anticipation in an immersive virtual reality (VR) driving simulation whose collision outcomes were assigned by a quantum random number generator, while EEG, photoplethysmography (PPG), EMG, EOG and inertial (IMU) signals were recorded from a single wireless dry-electrode headset built into the VR display. Sixteen participants passively observed 120 driving sequences ending in a collision or no collision. Length-matched pre- and post-stimulus windows were analysed identically with mass-univariate hierarchical general linear models under permutation cluster correction, plus machine-learning classification across held-out participants. Two post-stimulus clusters differentiated the conditions, the largest peaking at 451.6 ms (d = -1.35), with broadband power modulation robust to baseline modelling, decoded from held-out participants with up to 93.8% accuracy. No time-domain effect or decodable information preceded stimulus onset, but a broadband spectral difference did, surviving preregistered controls for expectation bias (run length), time-on-task build-up and ocular or facial-muscle artifact: directionally consistent in 13 of 15 participants, with pre- and post-stimulus power coupled in all 16. Heart rate differentiated the conditions neither alone nor when added to the classifier. The study was preregistered; the BIDS-format data, analysis code and an open-source EEGLAB plugin are publicly available.
ER -