Brain Representation in Conscious and Unconscious Vision

Area of Science:

• Neuroscience of Consciousness
• Cognitive Neuroscience
• Computational Neuroscience

Background:

• Understanding neural representations of conscious and unconscious perception is crucial for advancing consciousness research.
• Previous functional MRI (fMRI) studies demonstrated decoding of unconscious content from brain activity, but this doesn't fully explain neural representations.
• A detailed, information-based approach is needed to analyze how the brain represents perceptual content across awareness states.

Purpose of the Study:

• To investigate neural representations of both conscious and unconscious perceptual content using advanced computational models.
• To compare brain activity patterns with representations derived from convolutional neural network (CNN) models.
• To explore the functional role of the fronto-parietal cortex in conscious perception.

Main Methods:

• Re-analysis of high-precision fMRI data.
• Application of representational similarity analysis (RSA).
• Utilizing convolutional neural network (CNN) models from computer vision to predict brain responses.

Main Results:

• CNN model representations significantly predicted brain responses in the ventral visual cortex and fronto-parietal regions for both conscious and unconscious content.
• These predictive patterns were consistent across different participants, even when models were trained and tested independently.
• The findings held true even for participants with no detectable perceptual sensitivity, indicating robust unconscious processing.

Conclusions:

• Neural representations of unconscious information are widely distributed across the brain.
• The fronto-parietal cortex's role in conscious perception may involve generating meta-representations rather than broadcasting information, challenging the global neuronal workspace theory.
• Higher-order theories of consciousness provide a more fitting framework for the observed fronto-parietal cortex functions.