MIT Finds Your Brain Doesn't Need Words to Think Logically

Most of us assume we think in words — that the sentence running through your head as you weigh a decision is more or less the reasoning itself. A new study out of MIT's McGovern Institute says otherwise. Researchers there found that logical reasoning and language processing run on almost entirely separate machinery in the brain, to the point that people who have lost the ability to understand or produce language can still solve demanding logic problems just as well as anyone else.

Two Patients Who Couldn't Talk — But Could Still Reason

The study, led by MIT associate professor Evelina Fedorenko and postdoctoral researcher Hope Kean, working with Rosemary Varley's group at University College London, centered on two stroke patients with severe aphasia — language-processing brain regions damaged badly enough to leave them unable to reliably understand or produce speech. Rather than treating that as a dead end for testing higher-order thinking, the researchers built logic games that required zero language at all: inferring a hidden rule that transforms a sequence of numbers, or spotting the pattern that completes a geometric matrix. Patients showed their answers with gestures or sketches instead of words.

The result: both patients solved the puzzles about as well as a control group with no language impairment. Whatever machinery was doing the reasoning, it clearly didn't depend on the machinery that had been knocked out by their strokes. The findings were published August 11, 2026 in the Proceedings of the National Academy of Sciences.

What the Scanner Showed in Healthy Brains

To confirm the pattern in people without brain damage, the team ran fMRI scans on neurotypical adults while they worked through the same kinds of problems. They first mapped each person's individual language network and separately mapped what's known as the "multiple demand network" — a set of regions long associated with effortful, general-purpose cognitive work. While participants solved the logic puzzles, language regions stayed largely quiet. The multiple demand network, meanwhile, activated during inductive reasoning — figuring out a hidden rule from examples — but not during deductive, syllogistic "if-then" reasoning, suggesting even different flavors of logical thought don't route through a single, unified system.

"There are aspects of thinking that seem to go beyond some of the limitations of language," Kean said of the findings. It's a useful reframe: language is linear, one word after another, while a lot of real reasoning — weighing multiple relationships at once, holding several possibilities in mind simultaneously — doesn't naturally fit that one-word-at-a-time format. The researchers note this also has stakes beyond pure neuroscience curiosity, including for understanding aphasia and for questions in AI research about whether language models need something closer to a separate reasoning system layered on top of language prediction.

Why This Matters If You Solve Problems for a Living

It's easy to assume the running commentary in your head — "okay, so if this happened, then that must mean..." — is the actual work of thinking. This study is good evidence that it's often more like a narration track than the engine itself. That distinction matters practically: if you've ever felt like talking a problem through in your head was actually slowing you down or leading you in circles, that's not necessarily a sign you're reasoning poorly. It may just mean you're running a genuinely logical problem through a system — language — that isn't built to carry it efficiently.

This connects to something we've written about before in the context of how elite performers make faster, cleaner decisions — see our piece on the mental models experts use to think more clearly under pressure. A recurring theme there is getting decisions out of vague internal deliberation and into an external, structured format. This new research gives that habit a concrete neurological rationale: structured, non-verbal formats may be working with the brain's actual reasoning circuitry rather than forcing it through a verbal bottleneck.

How to Actually Use This

• When you're stuck, stop narrating and start sketching. If a problem feels tangled the more you talk yourself through it, switch to a diagram, a flowchart, or physical objects you can rearrange. That shift moves the work into the format this study suggests reasoning actually prefers.
• Separate "explaining it" from "solving it." Writing a clear explanation afterward is a different, still-valuable skill — communication — but don't mistake the struggle to put a half-formed idea into words for a struggle to reason it out. The study suggests those can be two separate bottlenecks.
• Protect real thinking time before you start narrating out loud. Since reasoning and language appear to be separable systems, working through the guts of a hard problem in silence — before you try to explain it to a colleague or write it up — may let the non-verbal system do its job first. A dedicated block from our Focus Timer is a simple way to protect that window.
• Use language deliberately, not by default. If you tend to journal or self-talk your way through every decision, this research is a nudge to notice when that's genuinely helping versus when it's just habit. Our Affirmation Generator can help when the goal is motivation or mindset framing — a job language is well suited for — while purely logical problems may be better served by stepping outside language altogether.

The bigger picture is a brain that turns out to be less centralized than the "inner monologue = thought" model suggests. Reasoning, language, and general-purpose effortful cognition look like distinct systems that cooperate rather than one doing all the work. The more precisely researchers can map where each piece actually lives, the easier it becomes to notice which tool you're reaching for on a given problem — and to reach for a better one when the words aren't helping.

Frequently Asked Questions

Jordan Mercer

Brain Performance Research Analyst

12+ years analysing research on cognitive performance, attention science, and evidence-based productivity. Read full bio →