(Photo by Pheelings media on Shutterstock)
Your brain is constantly guessing what others think. Scientists just found the signal.
In A Nutshell
- Researchers identified a specific brain signal that activates each time a person revises their mental model of someone else during a strategic interaction.
- A network of social brain regions, centered on the temporoparietal junction, coordinates this process, and people with stronger network connectivity adapted to opponents more effectively.
- A machine learning algorithm predicted that brain signal from scans alone with an average correlation of 0.82, and the pattern held in an independent, more diverse group without retraining.
- The researchers say the neural marker could eventually help assess social reasoning difficulties in conditions like autism spectrum disorder, though no clinical testing has been done yet.
At some point, most people have sensed they were being outmaneuvered in a conversation, or felt a quiet confidence that they had finally figured out how someone else thinks. That instinct, that constant mental tracking of another person’s next move, has a specific home in the brain. Scientists say they have now identified a brain signal linked to this process with unusual detail.
A team from the University of Zurich scanned the brains of people playing a strategic game against opponents of varying skill levels and found a measurable brain signal that activates when a person updates their mental model of someone else. Individuals with stronger, more coordinated social brain networks were better at adapting to their opponents in real time.
Published in Nature Neuroscience, the research offers one of the clearest neural markers yet identified for this kind of moment-to-moment social reasoning for what scientists call adaptive mentalization, the brain’s ongoing effort to track what another person is thinking as new information arrives.
Most adults do this constantly, in every meeting, negotiation, argument, and casual exchange. Much prior research treated it more like a stable trait, something a brain either had or lacked to some degree. This study focused instead on its adaptive quality, and that shift in focus turned out to matter.
Why Rock-Paper-Scissors Reveals How the Social Brain Really Works
To study this under controlled conditions, the researchers needed a setting where social reasoning could be measured precisely, without the noise of language or facial expression. Rock-paper-scissors turned out to be nearly ideal.
In its standard form, the game should be unwinnable over many rounds because pure random play is theoretically optimal. But humans are notoriously bad at being random. People fall into patterns, and a sharp opponent can exploit those patterns by staying one strategic step ahead. Working from a numeric version of the game, the researchers designed computer opponents calibrated to play at one of three distinct levels, from a simple pattern-repeater to a multi-step strategist.
Making those artificial opponents believable required careful staging. Participants were told their opponents were located in an adjacent lab, connection checks appeared on screen before each session, and researchers simulated delays as though waiting for other players to get ready. In a formal Turing-test-like evaluation, participants could not reliably tell the computer opponents apart from real human players.
Across nine sub-studies involving 553 total participants, players faced these opponents in sequences where the strategic level shifted between rounds without warning, forcing continuous adaptation. A computational model called CHASE, short for Cognitive Hierarchy Assessment, calculated after each round how strongly the opponent’s move caused the player to revise their mental read on that opponent. That revision signal, called a belief update, became the core measurement of the study.
What Lights Up in the Social Brain When Reading Other People
For the brain-scanning portion, 50 participants, all male, played the game inside an fMRI scanner, the same type of machine used in hospitals to detect strokes and tumors, while researchers tracked their brain activity in real time.
Two regions proved central. One is the temporoparietal junction, an area near the back and sides of the brain long associated with perspective-taking and reading others’ intentions, though research has established that activity there is not exclusive to social reasoning. When an opponent’s move prompted a significant rethinking of strategy, the temporoparietal junction fired alongside it. A second region, the anterior insula, tied to social awareness and interpersonal processing, tracked the same signal. Together with parts of the prefrontal cortex, these areas formed a coordinated network whose activity rose and fell in step with how actively a participant was revising their view of the opponent.
Individual differences added a telling dimension. Participants whose right temporoparietal junction showed stronger connections with the rest of the social brain network tended to update their beliefs about opponents more strongly, responding to changes in opponent strategy with more pronounced mental revisions. Being a more adaptive social reasoner was not about one region firing harder; it was about a wider circuit operating in tighter sync.
How Accurately a Brain Scan Can Predict the Belief-Updating Signal
Using machine learning, the team extracted a whole-brain activity pattern tied to the belief-updating process. To be specific, the goal was to predict, from brain data alone, how strongly a participant’s mental model of the opponent shifted on each round, not to predict the opponent’s next move. Accuracy was strikingly high. The correlation between what the brain scans predicted and what the computational model calculated averaged 0.82 across participants, meaning the two tracked each other very closely.
When the researchers then applied that same brain pattern to an entirely separate group of 47 participants, a more demographically diverse sample that was 57 percent female and skewed older, accuracy remained strong without any retraining. That matters because a primary limitation of the original sample was that it consisted entirely of young men. A brain-based marker that holds up in a different population without recalibration carries considerably more weight than one confined to a narrow demographic slice.
What This Means When the Social Brain’s Reading Ability Breaks Down
For people with conditions that disrupt social cognition, the failure of this adaptive process carries real consequences. Autism spectrum disorder is among the most studied, and current diagnosis relies heavily on behavioral interviews and questionnaires, a process that is time-consuming and shaped by subjective judgment. A brain-based measure that directly reflects how well someone adjusts their social reasoning in real time could, in principle, offer a clearer path toward diagnosis.
No clinical testing has yet been done, and the researchers frame that application as a future direction rather than a near-term reality. That level of accuracy in the prediction work does suggest the approach could eventually have clinical relevance, though it has not yet been tested in patient populations.
The study is specific in its scope: it tracked a particular brain signal, during a particular task, in a controlled lab setting. Whether that signal maps cleanly onto the richer, messier social reasoning of everyday life remains an open question. What the research does establish is that when the brain revises its model of another person, something measurable happens, it varies across individuals, and it can be read from a brain scan. That is a more concrete foundation than the field has had before.
Paper Notes
Limitations
Several constraints shape how broadly these findings can currently be applied. The primary brain-imaging sample included only 50 participants, all male and relatively young, recruited from a university economics department in Zurich, Switzerland. While the replication sample was more demographically varied, neither group represents the general population. The study relied on a single game format, and whether the same brain signal appears in more naturalistic social settings remains untested. No testing has been conducted in clinical populations, meaning proposed applications to conditions like autism spectrum disorder remain speculative. The authors also note that mentalization encompasses dimensions, such as emotional empathy, that the current model does not measure, and that temporoparietal junction activation is not exclusive to mentalization processes.
Funding and Disclosures
Funding was provided by the European Research Council under the European Union’s Horizon 2020 research and innovation program (grant agreement 725355, ERC consolidator grant BRAINCODES, awarded to Christian C. Ruff). Additional support came from the Swiss National Science Foundation (grant number 10.006.863) and the University Research Priority Program “Adaptive Brain Circuits in Development and Learning” (URPP AdaBD) at the University of Zurich. All authors declared no competing interests.
Publication Details
Authors: Niklas Buergi and Gökhan Aydogan (co-first authors), Arkady Konovalov, and Christian C. Ruff, affiliated with the Zurich Center for Neuroeconomics, Department of Economics, University of Zurich, Zurich, Switzerland. Additional affiliations include the Max Planck Institute for Biological Cybernetics (Buergi) and the School of Psychology, Centre for Human Brain Health, University of Birmingham (Konovalov). Journal: Nature Neuroscience. Paper title: “A Neural Signature of Adaptive Mentalization.” DOI: https://doi.org/10.1038/s41593-026-02219-x. Published online: March 9, 2026.







