Structured synaptic inhibition has a critical role in multiple-choice motion-discrimination tasks

Area of Science:

• Computational neuroscience
• Systems neuroscience

Background:

• Previous neural network models for decision-making tasks struggled to interpret key experimental findings.
• Homogeneous connectivity assumptions in prior models limited their explanatory power for complex behaviors.

Purpose of the Study:

• To develop a structured recurrent neural network model that explains neural mechanisms in multiple-choice motion-discrimination tasks.
• To investigate the distinct roles of opposite-feature and similar-feature inhibition in decision-making processes.

Main Methods:

• Constructed a continuous recurrent neural network model with structured inhibitory connectivity between pyramidal cells and interneurons.
• Incorporated two specific inhibitory patterns: opposite-feature and similar-feature inhibition.
• Utilized a simplified mean-field model for analytical characterization of inhibition effects.

Main Results:

• The model successfully reproduced a wide range of physiological and behavioral data from monkey experiments, surpassing previous models.
• Opposite-feature inhibition was found to implement an elimination strategy, accelerating decisions at the expense of accuracy.
• Similar-feature inhibition enhanced choice accuracy among multiple options but slowed the decision process.

Conclusions:

• Structured synaptic inhibition is crucial for multiple-choice decision-making.
• The interplay between different inhibition types shapes decision speed and accuracy.
• A combination of cross-feature and similar-feature inhibition is predicted to be essential for categorical choices among numerous alternatives.