Combining computational modeling and neuroimaging to examine multiple category learning systems in the brain

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Neuroimaging

Background:

• Human category learning is supported by distinct neural systems: one for conscious rule inference and another for implicit information integration.
• Interactions between these systems during learning remain understudied.
• The Parallel Interactive Neural Networks Active in Category Learning (PINNACLE) model simulates competing categorization systems.

Purpose of the Study:

• To investigate the neural correlates of internal cognitive states during category learning.
• To examine system interactions and competition within a multiple-system framework.
• To explore covert learning activity in non-dominant systems.

Main Methods:

• Re-analysis of two prior functional magnetic resonance imaging (fMRI) studies using the PINNACLE computational model.
• Identification of neural correlates associated with hypothesized internal cognitive states on a trial-by-trial basis.
• Examination of brain regions active during periods of maximal system competition.

Main Results:

• Identified additional brain regions supporting distinct category learning systems.
• Found heightened activity in regions associated with maximal system competition.
• Observed evidence of covert learning in the "off system"—the system not actively driving behavior.

Conclusions:

• The PINNACLE model offers a plausible framework for understanding the neural organization of competing category learning systems.
• Synergistic use of computational modeling and fMRI provides access to complex cognitive processes in decision-making.
• Findings highlight the dynamic interplay between conscious and implicit learning mechanisms in the brain.