It was never all happening in sentences

There’s a quiet pressure running through most wellbeing advice: put it into words. Name it. Write it down. Say what you’re feeling.

Underneath sits an assumption almost nobody examines — that a thought isn’t quite real until it’s a sentence. That the inside of a head is basically a voice talking.

A study published this week in PNAS puts a dent in that assumption. Not in the place you’d expect, and not in the way a headline would like. But the dent is worth looking at.

What they did

Hope Kean and Evelina Fedorenko at MIT, with colleagues in the US and UK, took a two-part run at an argument that has been going for a very long time: is language the medium in which complex thought happens, or just the thing we use to report it afterwards?

First, they scanned 29 healthy adults with fMRI. Each person’s language network was mapped individually — by having them read real sentences versus lists of pronounceable non-words — so the researchers knew exactly which regions to watch in that particular brain.

Then those same people worked through three kinds of reasoning problem. Inductive: work out the rule that turns one list of numbers into another. Deductive: read a classic syllogism and judge whether the conclusion actually follows. Matrix: find the odd one out in a grid of geometric patterns.

The second part was smaller and, in its way, more striking. Two people who had sustained extensive damage to the left hemisphere of the brain, leaving them with profound aphasia and severe grammatical and linguistic impairments, sat the non-verbal versions of the inductive and matrix tasks alongside 40 adults without brain injury. The syllogism task was dropped, because it requires reading complex sentences.

What they found

In the scanner, the language regions behaved exactly as you’d expect while people read sentences — strong, reliable activity. During the logic problems they were mostly quiet. Almost nothing.

The work was being done elsewhere. Inductive and matrix reasoning drew on the multiple demand network, a general-purpose system that shows up across all sorts of effortful, goal-directed tasks. Deductive reasoning didn’t recruit that network at all; it engaged a distinct set of frontal and parietal regions. Different kinds of thinking, different roads.

And the two participants with aphasia did well. On the rule-finding problems they scored 76 percent and 98 percent — on par with the uninjured control group. On the geometric pattern task, both performed well above the average for their age group.

“People often assume that complex thoughts take the form of an internal monologue, based in language,” Kean said. “Our findings suggest that, at least for formal logical reasoning in adults, this is not strictly the case.”

The limitations are stated plainly, and they matter. Two participants is two participants; the authors note how rare it is to find people with impairment that profound and specific alongside intact general cognition. Everyone studied was an adult, which leaves open whether learning to reason in the first place requires language, even if the grown brain later files the two separately. And the researchers are careful to say this doesn’t mean language and thought are sealed off from one another in ordinary life — the language network interacts with reasoning constantly, and may work as a kind of scratchpad for holding information mid-task.

What we keep thinking about

The honest caveat first, because it’s a big one: this study is about formal logic. Number rules, syllogisms, grids of shapes. Nobody in it was asked about grief, or the flat feeling that shows up on a Wednesday with no explanation attached. Stretching a result about matrix puzzles into a claim about emotions would be exactly the sort of overreach we try hard not to do.

But the assumption the study was built to test is one a great many people carry into their own inner life. If it isn’t in words, it isn’t really there yet.

That assumption does small, quiet damage. It’s there in the person who opens a journal and finds nothing comes out, and concludes they must be bad at introspection. It’s there when someone is asked how they’re doing and the truest answer available is I don’t know — and they hear that as a failure rather than a description. It’s there in the idea that being unable to explain a feeling means the feeling doesn’t quite count.

What stays with us from this research is the shape of it. The reasoning ran fine without the narration. The words were a separate system, doing a separate job.

There’s a related point in the study we’d want to carry, because it belongs to the participants rather than the finding. Kean referenced an old adage, one the National Aphasia Association has long promoted: aphasia is a disorder of communication, not of intellect. Because language is ordinarily how we show our thoughts to other people, difficulty with language gets mistaken for absence of thought. It isn’t. That’s worth knowing whether or not you ever meet it up close.

Which brings us round to a small design decision. A large part of why Nagi’s check-in is four taps — mood, sleep, energy — is that there’s no text box you have to fill first. No prompt asking you to explain yourself before the day counts. Not because words have no value; often they’re the whole thing. But requiring them turns two seconds into a small piece of homework, and homework is what people abandon on precisely the days worth recording. Nagi doesn’t score you, rank you, or tell you what any of it means — it isn’t a clinician and it doesn’t diagnose anything. It just holds the record, privately: no name, no email, an anonymous ID rather than a name.

The bit we’d underline

If you can’t say what’s wrong today, that isn’t a gap in your self-knowledge.

Sometimes the sentence turns up a week later. Sometimes it never does. The thing was there the whole time either way, and you’re allowed to note that it happened without first working out how to describe it.

If you want a private place to start noticing your own patterns, Nagi is on Google Play and the App Store — or in your browser at app.nagiandme.app. The first 15 messages are free. No name, no email. Just you, checking in.

— The Nagi team

---

Source: Neuroscientists demonstrate that logical reasoning does not require language, PsyPost, 1 September 2026, by Eric W. Dolan. Underlying study: Kean, H., Fung, A., Jaggers, P., Chen, J., Rule, J. S., Benn, Y., Tenenbaum, J. B., Piantadosi, S. T., Varley, R. A., & Fedorenko, E., “Evidence from formal logical reasoning reveals that the language of thought is not natural language”, Proceedings of the National Academy of Sciences. fMRI sample: 29 healthy adults, with each participant’s language network mapped individually. Case studies: two adults with profound aphasia following extensive left-hemisphere damage, compared against 40 adults without brain injury on non-verbal inductive and matrix reasoning tasks. Stated limitations: the aphasia sample is only two people, as such profound and specific impairment alongside intact general cognition is rare; the work covers adults only, leaving open whether language is needed to acquire logical reasoning in childhood; and the authors are explicit that the findings do not mean language and reasoning operate in isolation in everyday life.

Nagi is a mental health companion, not a medical device, and nothing here is medical advice. If you’re struggling, please reach out to a professional or someone you trust.