Information-Theoretic Evidence for Predictive Coding in the Face-Processing System.

Area of Science:

• Neuroscience
• Cognitive Science

Background:

• Predictive coding theory posits the brain predicts sensations using prior knowledge.
• Neurophysiological evidence for activated prior knowledge in predictive coding is limited.
• Maintaining prior knowledge until needed is crucial for predictions.

Purpose of the Study:

• Investigate the neurophysiological basis of activated prior knowledge in predictive coding.
• Measure maintained information in neural signals using active information storage (AIS).
• Examine the role of low-frequency brain activity in predictive coding.

Main Methods:

• Calculated AIS on magnetoencephalography (MEG) data from 52 subjects.
• Analyzed whole-brain source time courses during a Mooney face/house detection task baseline.
• Utilized beamformer-reconstructed source activity for analysis.

Main Results:

• Increased α- and β-band AIS in content-specific areas indicated preactivated prior knowledge for faces.
• AIS increases in the fusiform face area were behaviorally relevant.
• AIS enabled trial-by-trial decoding of cued categories.
• Activated prior knowledge and predictions are signaled by low-frequency activity (<30 Hz).

Conclusions:

• Activated prior knowledge and predictions are reflected in low-frequency neural activity.
• AIS provides a novel, assumption-free method to study predictive coding.
• Findings support predictive coding models of perception and cognition.