Formation of cognitive maps in large-scale environments by sensorimotor integration

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• The hippocampus is crucial for navigation and spatial memory in mammals.
• Understanding cognitive map formation in large-scale environments remains a challenge.
• Existing research has limited investigation into large arenas.

Purpose of the Study:

• To elucidate computational mechanisms of cognitive map formation in large environments.
• To propose a neural network model of the entorhinal-hippocampal circuit.
• To investigate the integration of spatial and non-spatial information.

Main Methods:

• Developed a neural network model of the entorhinal-hippocampal circuit.
• Integrated spatial information from medial entorhinal cortex (MEC) grid units.
• Incorporated non-spatial (object) information from lateral entorhinal cortex (LEC) visual units via deep neural networks.
• Used a competitive learning rule for synaptic weight adjustment.
• Simulated the model in a large box maze.

Main Results:

• Place units in the hippocampal model formed multiple, irregularly spaced fields.
• Dominant visual input (non-spatial information) led to conjunctive and egocentric place unit responses.
• The model demonstrated the hippocampus's role in balancing information from MEC and LEC.

Conclusions:

• The proposed model successfully simulates cognitive map formation in large environments.
• Hippocampal function involves integrating spatial cues (MEC) and object information (LEC).
• The balance between spatial and non-spatial inputs influences hippocampal representations.