Study finds sensory cortex joins decision-making earlier than expected

Decision-making in the brain may start earlier than many neuroscience models have assumed, according to researchers at the University of Illinois Urbana-Champaign. The finding matters because it challenges a feedforward view of brain processing that has influenced artificial intelligence systems.

The study, led by electrical and computer engineering professor Yurii Vlasov at the university’s Grainger College of Engineering, was published in the Proceedings of the National Academy of Sciences. The paper, by Alex G. Armstrong and Vlasov, examined neural signals tied to perceptual decisions in the primary somatosensory cortex, known as S1.

A challenge to a one-way model

According to the University of Illinois Grainger College of Engineering, many long-running models describe the brain as a hierarchy in which sensory information moves upward through increasingly complex areas before decisions arise in higher regions such as the frontal cortex. That idea has also helped shape artificial intelligence systems, including convolutional neural networks.

Vlasov’s team reported evidence for a more interactive process. In the study, early sensory regions did more than relay incoming information; they showed activity linked to decisions and appeared to receive fast feedback from higher brain areas, according to the university.

The researchers recorded brain activity in mice while the animals moved through a virtual reality corridor and made perceptual choices. They found decision-related signals in S1, one of the first cortical regions involved in processing touch and other body-related sensory information.

That result suggests decision-making depends on communication running in both directions across brain areas, rather than a straight handoff from sensation to judgment, according to the university’s summary of the work. The researchers said higher brain regions may rapidly shape activity in early sensory cortex through feedback loops.

Why engineers are watching

The work connects neuroscience to a long-running engineering goal. The National Academy of Engineering in 2008 listed reverse engineering the brain among its 14 grand challenges for the 21st century, according to the University of Illinois.

Vlasov said the research is motivated by the efficiency of biological intelligence, which can perform complex tasks while using far less energy than current AI systems. “We want to learn from a billion years of evolution,” he said, according to the university.

He also said current AI falls short at the level of decision-making and that brain architecture could offer useful lessons. The researchers, however, did not claim the study provides a direct design for new AI systems.

Instead, the university said the findings may help engineers think about artificial neural networks that use feedback and distributed processing more like biological brains. Vlasov said a systems-level understanding of brain organization could influence how more efficient AI is built.

Next steps

The team plans to study the timing of the signals in more detail, according to the University of Illinois. Researchers also plan to develop new tools for measuring neural activity so they can better see how feedback loops form and coordinate activity across processing stages.

Vlasov said examining fast neural dynamics may help reveal how those loops take part in decisions. The study’s central claim is narrow but significant: early sensory cortex appears to participate in decision-making, rather than waiting for higher brain regions to finish the job.

This story draws on original reporting from ScienceDaily.