Asymmetric stimulus representations bias visual perceptual learning.

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• The primate visual cortex shows regional specialization for stimulus properties like motion, shape, and color.
• Further specialization within regions leads to over-representation of specific features (e.g., orientations), causing perceptual biases.

Purpose of the Study:

• Investigate how cortical asymmetries influence visual learning.
• Determine if learning is optimal or affected by existing perceptual biases.
• Distinguish between learning theories predicting optimal vs. biased learning outcomes.

Main Methods:

• Trained human observers to discriminate between expanding and contracting motion patterns.
• Analyzed perceptual biases and their changes with training.
• Simulated neural network models with different learning rules.

Main Results:

• Observers demonstrated biased percepts of motion stimuli.
• Training affected these biases in often suboptimal ways.
• A neural network model adjusting only decision criteria, not connection weights, explained the observed data.

Conclusions:

• Cortical asymmetries significantly influence visual perception and learning.
• Human visual learning strategies can be suboptimal.
• Learning rules involving decision criteria adjustments, not just connection weights, may underlie these effects.