Stubborn Predictions in Primary Visual Cortex.

Area of Science:

• Cognitive Neuroscience
• Perception Research
• Machine Learning Theory

Background:

• Perception relies on integrating sensory data with prior experiences.
• Optimal learning models suggest updating predictions based on uncertainty (precision).
• Alternative theories propose biases in uncertainty estimation, leading to rigid predictions.

Purpose of the Study:

• To investigate whether perceptual systems form "stubborn" predictions.
• To compare prediction models by examining brain activity during environmental change.
• To understand how top-down expectations influence perceptual inference.

Main Methods:

• Participants learned novel perceptual predictions.
• Functional magnetic resonance imaging (fMRI) recorded brain activity.
• Multivariate pattern analysis (MVPA) decoded neural patterns.

Main Results:

• Previously expected events were decoded from primary visual cortex despite environmental changes.
• Neural representations persisted even when expectations became statistically invalid.
• This indicates a resistance to updating predictions.

Conclusions:

• Perceptual systems form robust, potentially "stubborn" predictions.
• Top-down expectations can impede accurate perception in changing environments.
• Understanding prediction biases is crucial for perceptual inference.