Auditory cortex neurons that encode negative prediction errors respond to omissions of sounds in a predictable...

Area of Science:

• Neuroscience
• Computational Neuroscience

Background:

• Predictive coding theory suggests the brain anticipates sensory input, generating prediction errors when reality deviates from expectations.
• Negative prediction errors signal when actual input is less than predicted, potentially linked to responses to omitted stimuli.
• The precise information encoded by these omission responses and their neuronal basis remain incompletely understood.

Purpose of the Study:

• To identify and characterize neurons encoding negative prediction errors in the auditory cortex.
• To investigate the role of omission responses in confirming negative prediction error signaling.
• To explore the computational mechanisms and circuit properties underlying prediction error asymmetry.

Main Methods:

• Recorded single-unit activity in the rat auditory cortex during an auditory omission paradigm.
• Manipulated tone probabilities to systematically alter prediction content and evidence accumulation.
• Developed a circuit model of laterally interconnected prediction-error neurons to simulate observed neural activity.

Main Results:

• Identified neurons exhibiting robust responses to omitted tones, correlating with predictability and evidence accumulation.
• Observed an asymmetry in neuronal responses: selective omission responses but broad tone responses.
• Demonstrated that a circuit model with lateral connections qualitatively reproduced this asymmetry and enhanced prediction encoding efficiency.

Conclusions:

• Neurons responding to omissions in the auditory cortex exhibit characteristics of negative prediction-error units.
• An asymmetric error signaling mechanism, potentially mediated by laterally interconnected neurons, is proposed.
• Lateral connections within prediction-error circuits may improve the precision and efficiency of sensory prediction encoding, consistent with the free energy principle.