Home · Blog · Research

Cortex Learns General Structure, Subcortex Handles Rewards When Memory Is Limited

Cortex Learns General Structure, Subcortex Handles Rewards When Memory Is Limited

Your brain has two learning systems that work together: a flexible but memory-hungry cortex and a simpler but efficient subcortex. New research from Matthew Farrell and Taro Toyoizumi shows that when cortical memory is limited, these systems take on distinct roles — the cortex learns the general structure of the world, while subcortical circuits focus on rewards.

The Research

Farrell and Toyoizumi from RIKEN Center for Brain Science extended a computational model that pairs a model-based (cortical) module with a model-free (subcortical) module. They added a key constraint: the model-based module had limited memory resources. In a simple decision-making task, they tested different memory allocation strategies. They found that when rewarded states change frequently, the best strategy for the model-based module is not to focus on the current reward but to capture the general structure of the environment (e.g., transition patterns between states). This work provides a theoretical foundation for the idea that the cortex supports general structure learning, while subcortical circuits specialize in reward-based learning. The preprint, posted May 30, 2026, is available on arXiv (2606.00667).

Why It Matters

This study offers a clear computational explanation for why your brain doesn't just memorize rewards but builds mental models. When you're learning a new game or navigating a new city, your cortex is building a map of how things work (general structure), while subcortical regions automatically track what's rewarding (e.g., which street has the coffee shop). Understanding this split can help you appreciate why some learning feels effortless and why breaking complex tasks into structured patterns can improve memory efficiency.

What You Can Do

  • Focus on understanding underlying rules rather than just memorizing outcomes — this leverages your cortex's strength for structure learning.
  • Practice in stable environments first to build strong mental models before tackling changing reward contexts.
  • Use spaced repetition and varied practice to reinforce the structural knowledge your cortex is built to capture.

Source: arXiv q-bio.NC

Curious about your own brain? Take our free adaptive IQ test or try 306 brain training levels.

Curious about your own IQ?

Take our free, scientifically designed adaptive test across 7 cognitive domains. No signup required.

Take the free test