As you get better at recognizing familiar faces in a crowd, spotting typos at a glance, and predicting your next move, the game-sensing neurons in your brain become more attuned to sharing information rather than acting independently. That’s the conclusion of a new study by researchers at the University of Rochester and its Del Monte Neuroscience Institute. scienceThis challenges the long-standing assumption in neuroscience that learning increases efficiency by minimizing repetitions between neural signals.
The study, led by Shijiao Liu, a graduate student in the labs of faculty members Ralph Hefner and Adam Snyder in the Department of Brain and Cognitive Sciences, shows that learning actually increases shared activity between neurons. The discovery could provide insights into learning disabilities and lead to more flexible, human-like artificial intelligence tools.
“The dominant view in neuroscience is that learning allows neurons to act more independently, making the brain more efficient and able to read information more cleanly,” Liu says. “Our findings support the alternative idea that the brain’s sensory cortex is not just passively encoding the world, but actively performing inferences by combining what comes in and the brain’s learned expectations.”
How learning reshapes neural teamwork
For decades, researchers believed that learning reduces shared activity between neurons, streamlining the way the brain processes information and allowing it to be read more efficiently. This idea has shaped how researchers think about everything from perception to decision-making.
But the work of Liu, Haefner, Snyder, and their team suggests a different mechanism. As learning progresses, neurons become less independent and more coordinated, increasing the amount of information they share, especially when the brain is actively working on a task and making decisions.
This adjustment reflects the brain’s increasing reliance on internal expectations. As learning progresses, feedback from higher brain regions appears to shape how sensory neurons respond, allowing them to incorporate both incoming information and what the brain has learned from past experiences into perception.
Track neurons as learning unfolds
The researchers tracked the activity of the same small network of neurons in the visual cortex over several weeks as the subjects learned to distinguish between different visual patterns. The researchers measured whether neurons became more independent or shared more information as learning progressed.
The researchers found that before learning, neurons work mostly independently. But as the subjects honed their visual skills, the neurons began to behave more like a well-trained sports team, communicating and cooperating in a coordinated way.
It’s like a group of people solving a problem. Instead of everyone working in isolation and being as efficient as possible, they learn and become more communicative. When information is shared, each individual becomes more informed and the group can become more flexible and adaptable. ”
Adam Snyder, faculty member, Department of Brain and Cognitive Sciences, University of Rochester
Importantly, this coordination effect appeared only when subjects were actively performing the task and making decisions based on what they saw. If they viewed the same image passively without having to respond, the effect disappeared.
The neurons most important to the task showed the greatest boost in coordination, especially at the moment a decision was made.
However, these are flexible changes, not permanent changes. The researchers believe that these changes are guided by feedback signals from higher brain regions, allowing neurons to adjust their behavior on the fly depending on the task.
The results support the growing idea in neuroscience that the brain is more than a simple conveyor belt transmitting information forward. Instead, it constantly blends what we see with what we expect to see, creating a richer, more informed picture of the world. And that mixing requires groups of neurons to work together rather than separately.
Insights on health and AI
Understanding how the brain modulates neurons during learning may provide new insights into learning disorders and conditions that affect perception. It could also help scientists design more generalizable artificial intelligence systems, taking inspiration from the way the brain flexibly blends prior expectations with new sensory information.
“Most current artificial intelligence systems are built on discriminative architectures that directly map sensory inputs to outputs,” Hefner says. “Our new research suggests that incorporating generative feedback loops in which internal models shape sensory representations may lead to systems that learn faster from limited data, are more robust to uncertainty, and adapt more flexibly to changing tasks.”
sauce:
Reference magazines:
Liu, S. Others. (2026). Task learning increases the information redundancy of neural responses in monkey visual cortex. Science. DOI: 10.1126/science.adw7707. https://www.science.org/doi/10.1126/science.adw7707
