Task learning increases information redundancy of neural responses in macaque visual cortex.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• The brain must optimize sensory information for efficient decision-making, especially in novel tasks.
• Two competing hypotheses exist: learning reduces neural redundancy for efficiency, or increases it based on Bayesian inference.
• Understanding this process is key to deciphering neural computation.

Purpose of the Study:

• To investigate how the brain optimizes sensory information during learning for decision-making.
• To test whether learning increases or decreases neural redundancy in visual processing.
• To differentiate between efficiency-based and Bayesian inference models of learning.

Main Methods:

• Tracking population neural responses in macaque cortical area V4.
• Analyzing neural data during visual discrimination task learning over weeks.
• Quantifying changes in neural redundancy and information content within trials and across training.

Main Results:

• Task learning significantly increased neural redundancy in area V4.
• This increase in redundancy occurred both over weeks of training and within single trials.
• Redundancy enhancement correlated with increased information carried by individual neurons, not decreased information.

Conclusions:

• Findings strongly support Bayesian inference predictions, indicating learning increases neural redundancy.
• Neural processing appears to utilize a generative inference process rather than a purely discriminative one.
• Increased redundancy may be a mechanism for robust and efficient sensory information processing in the brain.