Defying Expectations: How Neurons Compute Prediction Errors in Visual Cortex

Area of Science:

• Neuroscience
• Computational Neuroscience
• Visual Processing

Background:

• Predictive coding models suggest the brain anticipates sensory input.
• Understanding the neural computations underlying these predictions is crucial for explaining perception.
• Subthreshold neuronal activity plays a key role in integrating and processing information.

Purpose of the Study:

• To investigate the subthreshold computations involved in predictive coding within the mouse visual cortex.
• To determine which neuronal layers are involved in computing prediction errors.
• To elucidate the mechanisms by which predicted and actual sensory inputs are compared.

Main Methods:

• Whole-cell recordings were performed in vivo in the mouse visual cortex.
• Neuronal activity was recorded during locomotion to generate visual flow stimuli.
• Analysis focused on subthreshold membrane potential fluctuations to assess neuronal computations.

Main Results:

• Layer 2/3 neurons in the visual cortex exhibit computations consistent with prediction error signaling.
• These neurons appear to subtract predicted visual flow from actual visual flow.
• Layer 5/6 neurons do not show evidence of computing prediction errors in this manner.
• Subthreshold activity in layer 2/3 neurons reflects the comparison of expected and unexpected visual motion.

Conclusions:

• Layer 2/3 neurons are critical for computing prediction errors related to locomotion-generated visual flow.
• Subthreshold computations in these neurons implement a key aspect of predictive coding.
• These findings provide a cellular-level mechanism for how the brain updates its predictions based on sensory evidence.