Abstract
We investigated whether perceptual states, and voluntary control over perceptual states during ambiguous structure-from-motion (SFM) can be predicted by decoding activation from retinotopic visual cortex and parietal motion-sensitive areas, using functional magnetic resonance imaging (fMRI) and multivariate statistics (support vector machines, SVM).
Our results indicate that prediction of perceptual states can be very accurate, when using activation from dorsal visual areas V3A, V4D, V7, MT+ and parietal areas responsive to SFM to decode and predict. Detailed analysis of eye movements revealed that strategic eye movement patterns are not the cause of the prediction accuracy based on cortical activation.
To examine the role of voluntary control, subjects were instructed to either switch their perceptual states as often as possible, or to hold the perceptual state as long as possible. Within a single run (of a total of 10), we interleaved the switch epochs (switch as often as possible) with hold epochs (hold the current perceptual state as long as possible). Prediction of the intention of subjects (switching/holding) was highly accurate when data from the frontal eye fields was used, decoding intention using the activation of dorsal visual areas was poor, but still possible with an accuracy higher than chance.
We conclude that during perceptual rivalry, dorsal visual areas, area MT+ and parietal cortex actively represent the content of visual awareness. In addition, frontal eye field activation is predictive for the intention of subjects: the voluntary control task (switching or holding a perceptual state).