Conventional wisdom holds that the human brain cannot truly multitask. What feels like doing two things at once is, in reality, rapid task-switching, the prefrontal cortex can consciously attend to only one demanding task at a time. This limit, known as the “frontal bottleneck,” is a fundamental constraint on human cognition.
But a new study from Georgetown University, published in the Journal of Cognitive Neuroscience, reveals that the brain can learn to escape this bottleneck. With enough practice, tens of thousands of trials over weeks, the neural circuitry of a task can be relocated from the prefrontal cortex to specialized regions in the temporal lobe, freeing the frontal cortex to handle other tasks simultaneously.
“Our work shows that experience remodels the brain to bypass that frontal bottleneck,” said senior author Maximilian Riesenhuber, a professor of neuroscience at Georgetown University School of Medicine. “The prefrontal cortex then stays free for whatever else you want to do, increasing your capacity.”
Thirty thousand trials of training
The study, led by Patrick H. Cox (now at Lehigh University), used an intensive longitudinal design. Thirty-one participants were trained for 5 to 10 weeks on a smartphone app to categorize morphed grayscale car images into two arbitrary categories, “SOVOR” or “ZUPUD”, a subtle visual discrimination task that required them to identify which side of a 50% morph boundary each image fell on. Training progressed through levels of increasing difficulty; participants had to achieve at least 90% accuracy to advance.
The complete training regimen involved more than 30,000 trials. Fourteen participants completed all phases, and 11 (8 female, mean age 23.4 years) provided usable neuroimaging data. Each participant underwent fMRI and EEG scanning at two time points: once after initial learning (approximately 6,000 trials over 1 to 2 weeks) and again after extensive training (the full 30,000-plus trials over 5 to 10 weeks).
From frontal control to automatic perception
At the first scanning session, after initial learning, the categorization task heavily activated the prefrontal cortex, the classic signature of controlled, effortful processing. The ventral occipito-temporal cortex (vOTC), a region specialized for visual object recognition, responded to the physical shape of the images but was not selective for category membership.
After extensive training, the picture had changed dramatically. Category-selective responses had emerged in the vOTC that were not present before: the region now signaled whether an image was “SOVOR” or “ZUPUD,” not just what it looked like. Functional connectivity had shifted: the vOTC showed decreased coupling with the prefrontal cortex and increased coupling with motor output areas.
The neural locus of categorization had moved from a controlled, prefrontal-dependent circuit to a streamlined perception-action loop running from the visual system directly to motor output, completely bypassing the frontal bottleneck.
“This is not just speed-up,” Riesenhuber explained. “It is a genuine change in neural architecture.”
True multitasking, not rapid switching
To test whether this neural shift truly enabled parallel processing, the researchers gave participants a dual-task test: they performed the car categorization while simultaneously carrying out a second, unrelated task. The critical finding was a correlation: the more the car task had been offloaded from the prefrontal cortex, measured as the decrease in vOTC–prefrontal connectivity, the better participants performed on the second task.
This correlation is the signature of true parallel processing. If participants were merely switching rapidly between tasks, no such correlation would exist. Instead, the two tasks were running on separate neural circuits at the same time.
The authors are careful to note the boundaries of the effect. Tasks that share the same sensory modality, for example, texting while driving, both of which consume visual resources, cannot be parallelized because they compete for the same input channels. Only tasks that can be routed through fully separate neural circuits can run in parallel.
Implications for expertise, habits, and safety
The study helps explain how experts, radiologists who spot tumors in seconds, birdwatchers who identify species at a glance, chess masters who assess positions almost instantly, can make rapid, accurate categorizations with minimal conscious effort. The brain has offloaded the skill to specialized temporal cortex circuits that operate automatically, leaving the prefrontal cortex available for other demands.
It also sheds light on why deeply learned habits, including compulsive behaviors, are so resistant to conscious control. Once a behavior is encoded in temporal cortex circuits, reasoning and willpower (prefrontal functions) have limited access to it. “Thinking of something else” strategies fail precisely because the habit is being executed by brain regions that do not answer to the prefrontal cortex.
The study’s limitations include a small final sample of 11 participants and a high attrition rate typical of intensive longitudinal studies. The task was artificial, morphed car images with arbitrary category labels, and the extent to which the findings generalize to real-world expertise remains to be tested. The authors identify the cellular and molecular signals that trigger the relocation from prefrontal to temporal cortex as the next major research question.
Source: Cox, P.H., Scholl, C.A., Laws, M.L., Jaimes, N.E., Jiang, X. & Riesenhuber, M. “Extensive Experience Remodels Neural Task Circuitry to Escape the Frontal Bottleneck and Increase Automaticity of Categorization.” Journal of Cognitive Neuroscience (2026). DOI: 10.1162/JOCN.a.2618