Premium

A New Clue to Where Speech Originates in the Brain

A new study suggests that part of the brain called Wernicke’s area may help you picture a word before you say it.

A digital illustration of a black and white brain on a light blue background with a cherry red corded phone resting on the brain.
Linking speech with listening in the same part of the brain could help restore voices and help us understand human evolution.
Photo Illustration: NGM Staff. Source Photo for Illustration: Nat Geo Image Collection, Alamy
ByHelen Thompson
Published November 19, 2026

In the 1870s, German neuropathologist Carl Wernicke was stuck on a puzzle. When some of his patients experienced damage to an area in the left hemisphere of the brain just above the ear, they lost the ability to understand speech, as well as the ability to speak. Their words came out as jumbled or in gibberish. Wernicke thought that this part of the brain, now called Wernicke’s area, might contain a sort of junction where auditory signals from our ears are matched to their meaning. But while that might explain difficulties understanding speech, it couldn’t answer the question of why the patients couldn’t speak normally.  

Now a new analysis may have finally solved the century-old mystery.

Experiments conducted through brain surgeries suggest that a subzone of this region contains neurons vital for speech production. The results are exciting as they provide a new clue for how to restore speech in some patients. But they are also significant because they change “the wiring diagram that we now have in our current textbooks,” says Edward Chang, a neurosurgeon at the University of California, San Francisco, who led the work.

A new high-resolution picture of the exact nerve cells that determine how we listen and talk is taking shape.

The classic view of language in the brain is that it’s split between two regions. Wernicke’s area, nestled in the auditory cortex, is responsible for listening and comprehending words. While Broca’s area, in the brain’s motor cortex that controls voluntary movement, handles speech production. The boundaries of each area are blurry, and vary slightly from individual to individual, but the general picture is one of two zones with different functions.

It’s this understanding that suggested an answer to the question of why Wernicke’s patients couldn’t speak. Researchers thought that the speech issues were really listening issues. Perhaps these patients still had some physical ability to form words, but without the ability to understand themselves while talking, they didn’t speak normally.

The study, published November 19 in the journal Science, redraws this picture.

Instead of two parts of the brain working separately on listening and speaking, the new study posits that they are intimately connected. Remarkably, the scientists were able to spot these connections in real time, in real brains undergoing surgery. 

Leading up to certain types of neurosurgery, doctors routinely test Wernicke’s area to minimize language loss. Patients are awake for some of these procedures, allowing doctors to ask them questions as they prod inside their skulls. Chang’s lab piggybacked on these tests in patients with tumors or epilepsy to analyze damaged parts of Wernicke’s, record brain activity, safely turn activity off with electrodes, and decode these signals. When patients were asked to recite a sentence before surgery, the team noticed that neurons in the left posterior superior temporal gyrus (lpSTG) lit up about a quarter of a second before patients would say a word. When the researchers disrupted this signal during surgery, patients also struggled with their words. That’s key. Right before you say a word, this area generates a sort of “blueprint” for that word and sends these instructions to a planning center that connects to the motor cortex, which tells the vocal tract, mouth, jaw, tongue, and lips what to do.

“This timing evidence strengthens the argument that the region is involved in generating upcoming words, rather than merely monitoring speech after it has been produced,” says Müge Özker Sertel, a neuroscientist at Radboud University in the Netherlands who was not involved in the study. Previous studies have hinted that this might be the case, but Chang’s team confirmed its source.

Their analysis also suggests that these neurons encode full words, rather than individual sounds or syllables. The signal was also stronger when the person uttered a more complex word like caterpillar compared to simpler word like cat, as well as when the word was more familiar, says Matt Leonard, a cognitive neuroscientist at the University of California, San Francisco who was also involved in the work.  This higher-level complexity reinforces the lpSTG’s role in providing instructions for an entire word and its importance for planning speech.

The idea of rewriting textbook brain diagrams isn’t totally surprising to Leigh Hochberg, a neurologist at Mass General Brigham in Boston who was not involved in the study. Going back more than a century to Wernicke’s time, scientists have typically assigned a function to a specific part of the brain based on cases where someone had an injury to that area and lost the ability to do something. But as researchers develop better methods to study the brain, “placing a circle around a part of the brain and assigning it a single function, I suspect is going to turn out to be a substantial oversimplification,” says Hochberg.

The results mean Wernicke wasn’t totally wrong about the idea of a junction where sound meets meaning. He just didn’t know that the area could serve more than one role.

The lpSTG can be as big as a nickel and as small as a bean, depending on the person. This variation can explain why not all patients who experience a stroke in this general area develop speech issues.

The next step for Chang’s lab is fixing such speech issues, and perhaps restoring the ability to speak. That’s where another set of the team’s experiments come in. The team analyzed brain activity from two people who were left paralyzed and unable to speak from strokes. From their data, the scientists were able to decode some of what the patients were trying to say. This could provide a new pathway for brain-computer interfaces (BCIs), implants aimed at restoring speech in people who have lost it. Currently, most devices in clinical trials can only help treat those with damage beyond the motor cortex.  

There’s hope that these insights could also help people with severe amyotrophic lateral sclerosis (i.e. ALS), brainstem strokes, or damage to the motor cortex itself. “In these individuals, it might be possible to identify words that the speaker intends to say by decoding electrical signals” from undamaged parts of Wernicke’s area, says Jeffrey Binder, a neurologist at the Medical College of Wisconsin who was not involved in the study.

Connecting listening and talking also get at bigger questions. This word blueprint could be important for other activities that might involve a mental representation of a word. “It remains to be seen whether that's also the case when you read, or when you write, or when you type, or when you're texting on your phone, but it's definitely a possibility,” notes Alex Silva, a graduate student in Chang’s lab, who is keen to investigate those questions next.  

Because some of the same neurons activated when hearing a word as well as speaking a word, the region may be important for the repetition that helps infants learn languages, the team argues. Other primates also don’t seem to have an analogous area in their brains, which could be a factor in why they didn’t evolve complex speech. “This subzone, which turns the words that we hear into words that we say may be part of what makes us human,” Chang says.

Helen Thompson is the Senior Digital Editor for Science at National Geographic. She covers everything from mosquito behavior to dinosaur debates.