Saccade learning with concurrent cortical and subcortical basal ganglia loops

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• The Basal Ganglia (BG) plays a crucial role in motor control and decision-making.
• Specific BG circuits, including cortical and subcortical loops, are implicated in saccade target selection.
• Understanding how these loops interact is key to explaining learning in spatial and feature-based tasks.

Purpose of the Study:

• To investigate how the structural relationships within saccadic loops influence learning of spatial and feature-based tasks.
• To propose and evaluate a computational model of saccade generation incorporating reinforcement learning.
• To explore the interplay between different cortico-basal and tecto-basal loops in target selection.

Main Methods:

• Development of a reinforcement learning model for saccade generation based on prior models of the BG and superior colliculus.
• Structuring the model around interactions between two parallel cortico-basal loops (spatial and non-spatial) and one tecto-basal loop.
• Testing the model's learning capabilities and interactions across various saccade tasks.

Main Results:

• The model successfully learned basic target selection based on both spatial and non-spatial criteria.
• The model reproduced and provided an explanation for training-dependent express saccades towards spatially cued targets.
• Simulations indicated that the spatial loop tends to dominate target selection in the absence of prefrontal cortex control.

Conclusions:

• The proposed model effectively captures the dynamics of saccade generation and target selection learning within BG circuitry.
• The findings highlight the distinct roles of spatial and non-spatial information processing loops in decision-making.
• The model offers insights into the neural mechanisms underlying saccadic behavior and potential clinical applications.