Recordings from thousands of neurons show how the human brain represents reasoning acts abstractly. By analyzing neurons in patients with epilepsy, researchers have determined how the brain creates cognitive maps during deductive reasoning. The hippocampus, traditionally associated with mapping physical space, also structures cognitive processes. The study illuminates how experiential and verbal learning affect neural representations and offers ideas for potential neurological treatments.

Uncovering the neural mechanisms of inferential reasoning

Understanding how any two things in the world are related—whether bad weather affects transit delays or environmental conditions affect the evolution of species—requires a brain. A new study based on recordings from human brains has produced a groundbreaking trove of data that researchers are now using to reveal the neural regulation of inferential reasoning more clearly than ever before.

“We are just beginning to understand how the brain learns and how we gain knowledge from what we experience,” said Ueli Rutishauser, Ph.D., professor of neurology, neurosurgery and biomedical sciences at Cedars-Sinai Medical Center and co-author.

The study, conducted as part of an inter-institutional consortium funded by the U.S. National Institutes of Health under its Brain Research through Advanced Innovative Neurotechnologies Initiative, or BRAIN Initiative, was published today (August 14) in the journal Nature.

Neural recording during inferential reasoning tasks

Using electrical recordings from more than 3,000 neurons in 17 patients with epilepsy who underwent invasive monitoring in the hospital to identify the sources of seizures, the researchers collected “a unique dataset that allows us to watch for the first time how brain cells represent the process of learning, which is critical for inductive thinking,” said Stefano Fuzzi, Ph.D., principal investigator at Columbia University’s Zuckerman Institute for Brain-Behavior and another co-author of the paper.

As the neuron recordings show, the scientists presented participants with a simple logical thinking task. In this task, the participants used trial and error to find the correct, monetary reward associations between images, such as a picture of a car or a piece of fruit, and pressing a left or right button. After the participants memorized these associations for a series of images, the researchers varied which button corresponded to the correct association for each image and which button did not.

Cognitive changes during learning and inference

The volunteers initially made the wrong choice because they didn’t realize that previously learned associations had changed. But these errors allowed the volunteers to quickly infer that the new picture button rule had come into play, and they were also able to infer that all new picture button rules had changed, including ones they hadn’t yet encountered. The scientists are comparing this experimental task to real-life situations that travelers often have to face when traveling abroad.

Neural representation of mental state transitions

“If you live in both New York and London and you’re flying into the United Kingdom, you know to look right to cross the road. You’re in a different state of mind that reflects the traffic rules you learned when you lived in London,” says Dr. Fusi, professor of neurology at Columbia Vagelos College of Physicians and Surgeons and a member of Columbia’s Center for Theoretical Neurology.

“Even if you go to places in the UK you’ve never been to, like rural Wales, you’ll still find there are new rules,” he added. “You still have to look right, not left, when you cross the road.”

“This work illuminates the neural basis of conceptual knowledge that is essential for reasoning, inference, planning, and even emotion regulation,” said Daniel Salzman, co-author of the Nature paper and principal investigator of the Zuckerman Institute and professor of science, psychiatry, and neurology at the Columbia College of Physicians and Vagelos surgeon.

High-dimensional geometric shapes in brain activity

So how are these kinds of thoughts physically expressed in the activity of neurons? Using mathematical tools developed by Dr. Fusi to integrate recordings from thousands of neurons, the researchers transformed the volunteers’ brain activity into geometric images, albeit shapes that span thousands of dimensions rather than the three dimensions we are accustomed to visualizing.

“These are high-dimensional geometric shapes that we can’t imagine or visualize on a computer monitor,” Dr. Fusi said. “But we can use mathematical techniques to visualize their very simplified images in 3D.”

When the researchers compared the brain activity patterns between when the subjects made successful inferences and when their inferences were unsuccessful, striking differences emerged.

“We observed transitions in certain populations of neurons during learning from non-ordered representations to these beautiful geometric structures that are associated with the ability to think deductively,” Dr. Fusi said.

The relationship between hippocampal activity and cognitive mapping

What’s more, the researchers only observed these structures in recordings from the hippocampus, not in other areas of the brain where the scientists looked, such as the amygdala and the frontal lobes of the cerebral cortex. This is a surprising finding, the researchers say, because the hippocampus has long been thought of as the brain’s home to neural maps of physical space. The new evidence suggests that the device can also create cognitive maps related to brain functions such as inference and learning.

Effect of verbal instructions on neural representations.

Another striking result of the study, according to Dr. Rutishauser, is that volunteers who learned the associative connections between the pictures and the buttons not through trial-and-error experience but through verbal instructions alone still formed the same “beautifully structured neural representations in the hippocampus.” This is an important observation, he said, because while people often learn from each other through verbal exchanges, little is known about how verbal information changes neural representations.

“Verbal instructions are how we construct knowledge about things we have never encountered before,” Dr. Rutishauser added. “Our study shows that verbal instructions lead to very similar structured neural representations compared to those resulting from experiential learning.”

The importance of epileptic patients in brain research

The researchers emphasize that none of these discoveries would have been possible without the cooperation and voluntary participation of patients with drug-resistant epilepsy who were hospitalized after surgery. Electrodes to collect neural data were temporarily implanted in patients by doctors for the sole purpose of determining the source of each individual’s seizures, with the ultimate goal of using this information for further surgery or neuromodulation treatment.