Cortical circuit-based lossless neural integrator for perceptual decision-making: A computational modeling study.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• Perceptual decision-making relies on integrating sensory evidence, often modeled by recurrent neural networks.
• Existing models struggle to explain how evidence is retained during sensory gaps.

Purpose of the Study:

• To investigate the neural mechanisms underlying evidence accumulation and retention in perceptual decision-making.
• To model how cortical circuits flexibly adapt to changing sensory evidence demands.

Main Methods:

• Utilized computational modeling to simulate neural circuit dynamics.
• Explored different readout mechanisms for sensory evidence within the models.

Main Results:

• Demonstrated that cortical circuits can switch between evidence "retention" and "integration" modes.
• Simulated both "stepping" and "ramping" neural activity patterns by varying evidence readout.
• Showed these patterns, observed in primate parietal cortex, can arise from a unified mechanism.

Conclusions:

• Cortical circuits possess flexible mechanisms for managing sensory evidence during decision-making.
• A single underlying mechanism can explain diverse neural activity patterns observed in perceptual decision tasks.
• Findings reconcile previous empirical studies on decision-making neural correlates.