Mismatch of Visual-Vestibular Information in Virtual Reality: Is Motion Sickness Part of the Brains Attempt to Reduce the Prediction Error?
PubDate: October 2021
Teams: Jena University Hospital;
Writers: Matthias Nürnberger1,2*, Carsten Klingner1,2, Otto W. Witte1 and Stefan Brodoehl1,2
Abstract
Visually induced motion sickness (VIMS) is a relevant limiting factor in the use of virtual reality (VR) devices. Understanding the origin of this problem might help to develop strategies to circumvent this limitation. Previous studies have attributed VIMS to a mismatch between visual, and vestibular information, causing ambiguity of the position of the body in relation to its surrounding. Studies using EEG have shown a shift of the power spectrum to lower frequencies while VIMS is experienced. However, little is known about the relationship between the intensity of the VIMS and the changes in these power spectra. Moreover, the effect of different varieties of VIMS on the causal relationship between brain areas is largely unknown. Here, we used EEG to study 14 healthy subjects in a VR environment who were exposed to increasing levels of mismatch between vestibular and visual information. The frequency power and the bivariate transfer entropy as a measure for the information transfer were calculated. We found a direct association between increasing mismatch levels and subjective VIMS. With increasing VIMS, the proportion of slow EEG waves (especially 1–10 Hz) increases, especially in temporo-occipital regions. Furthermore, we found a general decrease in the information flow in most brain areas but especially in brain areas involved in the processing of vestibular signals and the detection of self-motion. We hypothesize that the general shift of frequency power and the decrease in information flow while experiencing high intensity VIMS represent a brain state of a reduced ability to receive, transmit and process information. We further hypothesize that the mechanism of reduced information flow is a general reaction of the brain to an unresolvable mismatch of information. This reaction aims on transforming a currently unstable model with a high prediction error into a stable model in an environment of minimal contradictory information.
