The Faces of Predictive Coding.

Area of Science:

• Neuroscience
• Cognitive Neuroscience
• Computational Neuroscience

Background:

• Predictive coding theories propose that prediction errors (PEs) are signaled by neural activity.
• A key hypothesis suggests increased gamma-band activity (GBA) reflects PEs in cortical areas.
• Alternative views propose reduced neural responses for violated predictions.

Purpose of the Study:

• To test whether increased GBA signals PEs.
• To investigate neural responses to violated predictions of face orientation and illumination.
• To examine the temporal and spatial dynamics of PE signaling.

Main Methods:

• Magnetoencephalography (MEG) recorded neural responses in 48 human subjects.
• A Mooney face-detection task was used to induce PEs by violating predictions.
• Beamformer source analysis applied to high-frequency GBA (68-140 Hz).

Main Results:

• Behavioral data confirmed successful PE induction.
• Early PE signals for unexpected orientation and later signals for unexpected illumination were observed.
• Both PE signals correlated with increased high-frequency GBA.
• A late interaction effect in visual areas and attentional effects in other regions were also noted.

Conclusions:

• Results strongly support the hypothesis that increased GBA signals PEs.
• Findings align with predictive coding accounts of perception.
• GBA may also reflect attentional mechanisms in neural processing.