Scientists at Georgetown University Medical Center have achieved a breakthrough in understanding how the brain physically rewires itself as a skill transitions from conscious control to complete automaticity.
In a massive study involving volunteers sorting car images, researchers demonstrated that the brain does something far more sophisticated than simply speeding up task execution. It physically relocates the task to an entirely different brain region, thereby freeing up space for true parallel processing. This debunked the long-standing myth that humans are merely capable of rapid task-switching.
Over five to ten weeks, volunteers completed more than 30,000 car sorting trials via a mobile app, learning to distinguish the subtlest nuances between similar images. Researchers scanned the participants' brains using functional MRI and electroencephalography (EEG) twice: once at the very start and again after the training was finished.
It was this longitudinal approach that allowed them to observe how intensive practice literally reshapes the brain's neural architecture, forging new neural circuits where none existed before.
In the early stages of learning, the task required heavy lifting from the prefrontal cortex—the brain region responsible for conscious decision-making, planning, and volitional control.
This area, it turns out, acts like a bottleneck: it can only focus on one complex task at a time. This is precisely why your full attention is consumed when you are first learning to drive a car. However, after weeks of intensive practice, a dramatic shift occurred: activity moved entirely to the temporal cortex, a region specializing in object recognition and long-term memory storage. Now, information could bypass the prefrontal cortex bottleneck and flow directly into areas responsible for rapid, automatic responses.
"Experience remodels the brain to bypass this frontal lobe bottleneck and increase automaticity," explained the study's senior author Maximilian Riesenhuber, a professor of neuroscience at Georgetown Medical Center and co-director of the Center for Neuroengineering.
The effect was strikingly clear: the more the task shifted to the temporal cortex, the better participants performed a second task simultaneously—providing direct and undeniable proof of true multitasking, rather than just rapid attention switching. The reality of multitasking, which had long been debated, was finally scientifically confirmed.
The study explains why habits are so remarkably difficult to change. Well-learned behaviors are embedded into neural circuits that operate almost independently of conscious control. That is why, once a bad habit becomes fully automated, simply "wanting to change" is not enough—the habitual action triggers without involving the prefrontal cortex, which typically provides willpower. This scientific finding has practical implications: it shows that changing deep-seated habits requires different approaches than mere promises or willpower.
The discovery also sheds light on a fundamental difference between the human brain and modern artificial intelligence. While neural networks can recognize patterns and process data, they cannot transfer learned skills into new contexts—they do not learn to restructure themselves in response to experience.
The human brain, however, uses old knowledge uploaded to autopilot as building blocks for new skills. This allows humans to quickly master new abilities by building on familiar ones. This fundamental distinction points to a vital path for developing AI that can truly learn from experience rather than just accumulating parameters.
The study, titled "Extensive Experience Remodels Neural Task Circuitry to Escape the Frontal Bottleneck and Increase Automaticity of Categorization," was published in the Journal of Cognitive Neuroscience on June 4, 2026. The study authors are Patrick Cox (first author), Clara Scholl, Marisa Looze, Nelson Heimes, Xiong Jiang, and Maximilian Riesenhuber, all from Georgetown. Funding was provided by the National Science Foundation, the Army Research Laboratory, and the ARCS Foundation.
Researchers are already planning the next phase: identifying exactly which neural signals trigger the skill transfer from one brain region to another and determining which types of tasks are actually capable of reaching true parallelism.
