Explaining Integration of Evidence Separated by Temporal Gaps with Frontoparietal Circuit Models

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Decision Neuroscience

Background:

• Perceptual decisions depend on accumulating sensory evidence over time.
• Integrating evidence from temporally separated cues enhances performance, but lacks a neural model explaining accuracy and confidence.
• The random dot motion task with separated evidence presents a challenge for existing decision models.

Purpose of the Study:

• To develop and test a neural model that accounts for accuracy and confidence in perceptual decisions with temporally separated evidence.
• To investigate the neural mechanisms underlying evidence accumulation from discrete cues.
• To identify the brain regions and network dynamics crucial for integrating information over time.

Main Methods:

• Utilized behavioral data and electroencephalography (EEG) from a random dot motion visual choice task.
• Investigated three candidate distributed neural network models.
• Focused on models incorporating recurrent cortical dynamics and an uncertainty-monitoring module.

Main Results:

• Decisions based on temporally separated evidence are best explained by models integrating recurrent dynamics between centro-parietal and frontal brain areas.
• The proposed neural network model successfully accounts for both accuracy and confidence in the task.
• EEG data provided insights into the temporal dynamics of evidence integration.

Conclusions:

• The interplay of recurrent cortical dynamics in specific brain areas, coupled with uncertainty monitoring, is crucial for decisions based on separated evidence.
• This study provides a novel neural model for understanding complex evidence accumulation in decision-making.
• Findings advance our understanding of how the brain handles time-varying sensory information for perception and choice.