Combined fMRI- and eye movement-based decoding of bistable plaid motion perception.

Combining eye-tracking and functional magnetic resonance imaging (fMRI) improves the ability to decode which of two alternating perceptions a person is experiencing while viewing a bistable visual motion stimulus. Twenty participants viewed a plaid pattern that induces spontaneous perceptual switches and reported their perceptions. Using a support vector machine, multivoxel pattern analysis of fMRI data alone achieved 86% accuracy, eye-tracking alone 88% accuracy, and the combined approach reached 91% accuracy. This suggests that integrating brain and eye signals can reduce reliance on active verbal reports, which can confound studies of conscious perception, and may enable more effective no-report paradigms for investigating the neural correlates of consciousness.

Study at a glance

Abstract

The phenomenon of bistable perception, in which perception alternates spontaneously despite constant sensory stimulation, has been particularly useful in probing the neural bases of conscious perception. The study of such bistability requires access to the observer's perceptual dynamics, which is usually achieved via active report. This report, however, constitutes a confounding factor in the study of conscious perception and can also be biased in the context of certain experimental manipulations. One approach to circumvent these problems is to track perceptual alternations using signals from the eyes or the brain instead of observers' reports. Here we aimed to optimize such decoding of perceptual alternations by combining eye and brain signals. Eye-tracking and functional magnetic resonance imaging (fMRI) was performed in twenty participants while they viewed a bistable visual plaid motion stimulus and reported perceptual alternations. Multivoxel pattern analysis (MVPA) for fMRI was combined with eye-tracking in a Support vector machine to decode participants' perceptual time courses from fMRI and eye-movement signals. While both measures individually already yielded high decoding accuracies (on average 86% and 88% correct, respectively) classification based on the two measures together further improved the accuracy (91% correct). These findings show that leveraging on both fMRI and eye movement data may pave the way for optimized no-report paradigms through improved decodability of bistable motion perception and hence for a better understanding of the neural correlates of consciousness.