Two strangers sit in front of one screen. They have never met before today. Between them they have exactly one game of Tetris, and the controls have been cut in half: one of them can slide the falling block left and right, the other can rotate it. Neither can do both. And they are not permitted to speak.

That is the apparatus at the center of a paper published in NeuroImage this month — the divided-controls detail comes from PsyPost's report on the full study rather than the abstract — and it produces a result that runs hard against the thing almost everyone assumes about solving a puzzle alongside another person.

The rig

The study is Inter-brain synchrony during collaborative gaming: an investigation using theta-burst stimulation at the right temporal-parietal junction, by S. H. Jessica Tan, S. P. Jessie Leuk, and Wei-Peng Teo of Nanyang Technological University in Singapore (NeuroImage, volume 332, 2026; PMID 41932592). Thirty-three same-gender stranger dyads came in for three separate sessions and played seven-minute Tetris games both together and individually.

While they played, both members of each pair were wired for functional near-infrared spectroscopy — optical sensors that track blood oxygenation in the cortex — in a setup called hyperscanning, where two brains are recorded at once so you can ask whether their activity rises and falls together. The measure that falls out of this is inter-brain synchrony: how tightly two people's cortical signals track one another while they are doing something jointly.

The researchers were after a causal question. Interpersonal synchrony tends to show up strongly at the right temporoparietal junction, a region heavily implicated in taking another person's perspective, so they used theta-burst magnetic stimulation to turn that region's activity up in one session and down in another, and left it alone in a third. If the rTPJ is where joint performance is manufactured, poking it should move the needle.

It did not. Stimulation changed neither the synchrony nor the play. And synchrony turned out to be stronger at the prefrontal cortex than at the rTPJ across all conditions — the planning-and-monitoring real estate, not the perspective-taking region.

The result nobody was hoping for

Synchrony did rise reliably when the pairs played together rather than separately. That part is the expected shape, and it replicates a decade of hyperscanning work.

Then came the correlation. The abstract states it flatly: "a negative relationship between collaborative Tetris performance and IBS was found." The pairs whose brains matched most closely played worse. PsyPost's account of the full paper locates the effect specifically in the combination moves — the plays where several rows come out at once, which are the ones that require setting up a structure several blocks in advance and holding it while the pieces fall. The more aligned two brains looked, the worse the pair was at building anything complicated.

The authors are appropriately measured about it. Their conclusion is that the functional significance of inter-brain synchrony, particularly at the rTPJ, is "more nuanced than previously assumed." Which is the polite scientific way of saying that a great many of us have been reading the signal backwards.

Synchrony as a bill, not a receipt

The reading I find hardest to argue with is one the authors themselves do not make, so let me flag it plainly as my inference.

Synchrony may be what two minds spend on each other rather than what they earn together.

When you cannot speak, and you control half of an object that someone else controls the other half of, you have no choice but to run a live model of your partner: what are they about to do, what do they think I am about to do, is this rotation a request or an accident. That modeling has to run somewhere, and the "somewhere" it lit up here was prefrontal — the same territory you need for holding a four-move plan open while the board fills. Two processes, one desk. The more of that desk you spend inferring your partner, the less of it is available for the combo.

There is a corroborating detail buried in an older study. Liu and colleagues, in Frontiers in Human Neuroscience in 2016, ran fNIRS hyperscanning on pairs playing Jenga face to face. They found strong inter-brain synchronization in prefrontal regions during cooperative play — and also during obstructive play, where one partner was actively working against the other. Parallel play, where the two people simply did their own thing side by side, produced no such synchrony.

Synchrony, in other words, shows up whenever you must model another mind. It does not care whether that mind is helping you or sabotaging you. What it appears to measure is attention paid.

The part that should make us uneasy

The strangest finding in the Tetris study is the softest one, and it too comes via the PsyPost writeup rather than the abstract: afterward, participants rated their partners as more likable — consistently, and regardless of how badly the pair had actually played.

The warm feeling of having been in it together was fully decoupled from having been any good at it. Which means the internal signal we all use to judge a collaboration ("that felt great, we were completely on the same wavelength") is not tracking the thing we think it is tracking. It may be tracking how hard we worked to read the other person.

What this does to a puzzle team

I keep thinking about this in the context of team solving, because team solving is exactly this problem with more people and higher stakes. The MIT Mystery Hunt draws over 200 teams and something on the order of 5,000 solvers across roughly 48 hours, with the largest teams running to 150 or 200 people. Nobody at that scale believes coordination is free. But the folk model of why it costs is usually logistical: rooms, spreadsheets, who has the grid open.

This study suggests the cost is also neural, and it is charged per partner, continuously, against the same faculty you need for the deep move. The pair that feels most beautifully in step may be the pair that has quietly reallocated its planning capacity to mutual surveillance.

If that holds, the design implication is uncomfortable and rather concrete. A cooperative puzzle might do better to hand each solver a clean, unambiguous channel for declaring intent — a shared notation, a called move, a physical token placed on the board — so that coordination stops being inferred and starts being stated. Free the desk. Let the partners spend their prefrontal cortex on the puzzle instead of on each other.

Escape room designers have been converging on something like this by instinct for years, with the walkie-talkie rooms and the two-sided puzzles where each half of the team holds information the other must be told. What I had assumed was a narrative device might be a load-bearing cognitive one.

Which leaves me with a question I do not know how to settle. If synchrony is the cost of guessing, and stating your intent removes the guessing, then the most efficient team is also the least synchronized one — and it will feel, from the inside, like the worst collaboration of the day. How many good teams have broken up over that feeling?