A Sparse Probabilistic Code Underlies the Limits of Behavioral Discrimination

Area of Science:

• Neuroscience
• Computational Neuroscience
• Visual Perception

Background:

• Perception relies on cortical processing at behaviorally relevant timescales.
• Mice exhibit rapid visual stimulus integration (<100 ms) for orientation discrimination.

Purpose of the Study:

• Characterize cortical activity underlying rapid visual perception.
• Investigate how the brain achieves reliable perception from unreliable neural signals.

Main Methods:

• Measured single-unit responses in mouse visual cortex.
• Utilized a linear decoding model to assess information integration.
• Analyzed neural activity at timescales relevant to behavioral performance.

Main Results:

• Stimuli near perception threshold evoke sparse, unreliable neuronal responses.
• Individual neurons do not reliably encode stimulus orientation.
• Integrating information across 50-100 neurons sufficiently explains behavioral performance.

Conclusions:

• The visual system integrates probabilistic information from unreliable single neurons.
• Neural integration is crucial for generating reliable perception at the limits of detection.
• Cortical computation balances sparse coding with efficient information pooling.