You can lose the ability to speak and still grasp the rules of a game. That’s the counterintuitive takeaway from new research at MIT’s McGovern Institute for Brain Research. For centuries, we’ve assumed that language and logical reasoning are tangled up together. You think, then you find words for it. The brain processes structure in speech and structure in syllogisms using the same machinery. But a new study published in PNAS suggests that assumption is wrong.
Logical reasoning does not rely on the language centers of the brain.
Hope Kean, a postdoctoral researcher in the Fedorenko lab, and lead author Evelina Fedorenko found that even people with severe aphasia—who struggle to form sentences or comprehend speech—can solve complex logic puzzles. Brain scans of healthy volunteers confirmed it. When they worked through logic problems, their language regions stayed quiet. Instead, the brain used a different system entirely.
How Logical Reasoning Works Without Words
The connection between language and thought is seductive because they look alike on the surface. Both break down into smaller components. Both stack in hierarchical structures. You can take a thought, dissect it into sub-propositions, and reassemble those atoms into a complex rule. It looks just like grammar.
Hope Kean points out that this structural similarity makes language seem like the foundation of abstract thinking. “Abstract thinking has properties that look like language,” she notes.
But appearance can be deceptive. Just because two things share a skeleton doesn’t mean they grow from the same bone. The researchers questioned whether the same brain circuits must handle both. People need language to hear a problem or say their answer, sure. But what happens in the dark, between the prompt and the response?
Kean argues that logic often demands a precision that natural language simply can’t provide. Language is linear. It moves word by word. Logic? Logic can be spatial. It can evaluate multiple possibilities at once. Thinking in non-linear ways might be impossible in strict syntax but effortless in pure abstraction.
“There are aspects of thinking that go beyond the limitations of language.”
— Hope Kean
Testing the Limits of Aphasia
To prove this, the team had to strip language out of the equation entirely. They partnered with Rosemary Varley at University College London, an expert in acquired language disorders. This collaboration provided a rare opportunity. They recruited two participants who had suffered strokes damaging their language-processing brain regions.
The result? Severe aphasia. They couldn’t talk well. They struggled to understand spoken instructions. But did they lose their minds?
No. The researchers designed tasks that required zero words. One puzzle involved two lists of numbers. The participants had to find the hidden rule. Was the order reversed? Were numbers above a certain value discarded? Once they figured out the pattern, they had to apply it to new data.
Another task used geometric shapes. Participants looked at a matrix of designs and picked the missing piece. No instructions. No talking. Just pure rule discovery.
The results were stark. Both aphasic participants performed just as well as healthy controls. As the puzzles got harder, their accuracy held steady. They communicated their findings through gestures or sketches, proving they understood the abstract relationships even when they lacked the vocabulary to name them.
“It really upends a theory that symbolic rule induction is not possible without linguistics capacities,” Kean says.
MRI Evidence: Separating the Systems
Behavioral data is powerful. It shows what people can do. But it doesn’t show where it happens in the brain. To map the territory, Kean and the team brought healthy volunteers into fMRI scanners at MIT.
They used a dual approach. First, they mapped each person’s specific language zones. Then, they looked at the “multiple demand network”—a distributed system known for handling complex cognitive tasks like problem-solving and working memory.
Inside the scanner, volunteers played logic games. They tackled syllogisms like: “If the ball is red, it is big. The ball is red.” Logic dictates: Is the ball big?
The researchers tweaked the difficulty, pushing the brain harder to see which areas lit up. They compared two types of reasoning:
- Inductive reasoning: Finding a hidden rule from examples.
- Deductive reasoning: Assessing the validity of a logical statement.
The brain scans told a surprising story. Neither type of logic activated the language system. The language centers stayed dormant while the participant reasoned.
Here’s where it gets interesting. The multiple demand network jumped into action during inductive reasoning. It helped identify the hidden rules in the number lists and shape matrices. But? It did not participate in deductive reasoning. Kean is still investigating that split. Why does the brain use different circuits for finding patterns versus evaluating strict logical validity?
The evidence points to a clear separation. The brain handles logic independently of language. This mirrors earlier work from Fedorenko’s lab, showing that object categorization and social reasoning also bypass language circuits.
What This Means for Intelligence and AI
This isn’t just a philosophical win. It changes how we view disability. Aphasia is often misinterpreted as cognitive decline. If you can’t speak, people assume you can’t think.
But this research clarifies the distinction. Linguistic difficulty is not a proxy for intelligence.
People with aphasia can still balance a checkbook. Play chess. Do Sudoku. They understand consequences and rules. Their impairment is in the channel of expression, not the engine of thought.
“We should continue to educate the public… linguistic difficulties… are not indicative of how smart or capable anyone is.”
— Evelina Fedorenko
The findings might also have implications for artificial intelligence. Current large language models, like the ones powering ChatGPT or Claude, learn entirely from text. They generate text. They mimic reasoning because they are trained on human language, which contains logical statements. But they aren’t reasoning in the human sense. They are pattern-matching.
By comparing these models to the human brain—where logic and language operate separately—researchers can see the gaps in AI design. AI isn’t thinking. It’s echoing. Understanding the biological separation might help build systems that actually reason, rather than just simulate it.
The Geography of Thought
For Kean, the work is part of a larger mapping project. We know where we see. We know where we hear. But the geography of abstract thought? Still largely unmapped.
The study suggests that the brain has specialized islands for language and distinct regions for logic. They talk to each other, certainly. But they don’t share the same land. One person can lose the language map and still navigate the logical terrain.
It’s a humbling reminder. We are more than what we say. And perhaps, logic was never tied to words to begin with. It was there all along, waiting to be used.
