New Study Reveals How Brain Learns and Applies Rules

A groundbreaking study from the University of Toyama in Japan has shed light on how the brain learns and applies rules. The research, led by Assistant Professor Shuntaro Ohno, demonstrates that our ability to follow procedural rules is embedded in the dynamic patterns of neuronal activity within the medial prefrontal cortex (mPFC). The findings were published on July 1, 2025, in the journal Molecular Brain.

The research team conducted experiments with mice learning a Y-maze task, which involved navigating a maze for a reward. Initially, the mice could explore the maze freely. They were then required to wait for a light cue before moving to a water container, where they would receive a water reward by licking within a set time. As the training progressed, the mice became increasingly adept at obtaining rewards, despite following the same physical paths each time.

To capture the complexities of neuronal activity, the team employed a novel computational tool called iSeq. This tool utilizes convolutional non-negative matrix factorization to detect sequences of neuronal activation from imaging data without predefined behavioral labels. The analysis revealed that in the early training phases, the neural sequences were less predictive of outcomes. However, by the sixth day of training, significant differences in neural dynamics emerged between instances of success and failure in reward acquisition.

Dynamic Neural Patterns Reveal Learning Process

Dr. Ohno explained the importance of the iSeq tool, stating, “The development of iSeq allowed us to observe the brain’s internal organization of behavior in unprecedented detail.” The study revealed that as the mice learned, their mPFC underwent a dynamic reorganization. Neural activity patterns began to emphasize actions that reliably led to rewards, highlighting how learning reshapes brain activity over time.

An interesting aspect of the study is the changing composition of neurons participating in each sequence throughout the training period. The set of neurons involved on the first day differed significantly from those on the sixth day. This indicates that the mPFC does not rely on fixed neural circuits, but rather constantly reorganizes itself to refine behavior.

Dr. Ohno noted, “These results suggest that the brain does not store a rule as a static template. Instead, it continuously updates sequential activity patterns to link meaningful sensory cues, actions, and outcomes—essentially learning how to learn.”

Broader Implications for Neuroscience and AI

The implications of this research extend beyond basic neuroscience. The study bridges the gap between neural activity and the execution of behavioral rules, suggesting that procedural rules in the brain function like a sequence of events: stimulus leads to action, which results in reward. This chain of neural events evolves as the animal gains competence, with the brain reorganizing sequences to align with successful behaviors.

Understanding these mechanisms may inform rehabilitation strategies following brain injuries and even guide artificial intelligence in mimicking this adaptability. Furthermore, the iSeq method could be applied to examine sequence-based neural dynamics in other brain areas and species.

While the study focused on mice, it lays a foundational framework for understanding how the human brain learns to execute rules and adapt behaviors. The findings emphasize the role of temporal patterns in brain activity, suggesting they may be crucial for developing flexible, learned behaviors rather than relying solely on static neural connections.

The research team’s work promises to enhance our understanding of cognitive control, learning, and adaptation, providing valuable insights into the neural circuits that underpin these essential functions.