Science

Wernicke’s area plans your next word a quarter-second before you say it

Nadia Okonkwo
Add us on Google

Every word you speak begins in a patch of brain that textbooks assign to listening. A small zone inside Wernicke’s area, above and behind the left ear, switches on about a quarter of a second before a word leaves your mouth and holds a plan of the whole word, a team at the University of California, San Francisco, has found.

That overturns one of the oldest maps in neuroscience. For a century and a half, Wernicke’s area has been taught as the brain’s decoder for other people’s speech, while speaking was credited to Broca’s area in the frontal lobe. The new work answers a puzzle that Carl Wernicke’s own patients posed: why damage to a hearing region leaves people talking fluently but with the wrong words.

The zone, which the researchers identify as the left posterior superior temporal gyrus, is roughly the size of a small coin. Its activity grows with the length of the word about to be spoken and rises for rarer words. It sends its signal straight to the motor cortex, which drives the lips, tongue and larynx. “Every word that you have said today began in that small part above your left ear about a quarter of a second before you said it,” Edward Chang, the UCSF neurosurgeon who led the study, said at a news briefing.

The finding matters beyond textbooks. Brain-computer interfaces that give a voice to paralyzed people read signals from the motor areas, the last stop before the muscles. In people with aphasia, the loss of language that often follows a stroke, the planned word usually never reaches those motor areas intact. A region that holds whole words earlier in the chain could widen the group of people such devices can help.

How the UCSF team caught a word before it was spoken

The obstacle has always been speed. Functional MRI, the standard way to watch a living brain, tracks changes in blood flow that unfold over several seconds, far too slowly to show whether Wernicke’s area fires before a word or only after it. One popular theory held that it fires after: the region would act as a proofreader that hears your own voice and flags errors.

To get timing in milliseconds, the researchers worked with 28 people being treated for epilepsy, who had dense grids of electrodes placed directly on the surface of the brain as part of their care. The participants repeated words and sentences and listened to them while the grids recorded. A distinct group of sites inside Wernicke’s area became active before speech began, and its signal could be told apart from the neighboring responses to the sound of each person’s own voice.

Recordings show correlation, not cause, so the team added two more tests. Among hundreds of patients who had tumors removed from the temporal lobe, those whose surgery damaged this zone were much more likely to have trouble speaking afterward. Then, in 13 participants with healthy Wernicke’s areas, a brief electrical pulse delivered just before they spoke made them produce word errors like those of Wernicke’s aphasia, in which “shoe” can come out as “foot” and “leg” as “peg”.

Finally, the team asked whether the zone could feed a speech device. Two people who had lost the ability to speak after strokes in the brainstem already carried implants for a brain-computer interface. Deep-learning models decoded the words they silently tried to say, from a set list, using activity in Wernicke’s area alone. The same signals also decoded words the two heard.

Listening and speaking share the same cells

The most striking result came from comparing the two tasks. Many of the same recording sites that fired before speech also responded while people listened to speech. “That really sort of struck me,” Matthew Leonard, a UCSF neuroscientist and co-author, told reporters, adding that producing and perceiving speech “really do seem to be fully integrated.”

The researchers suggest the overlap could help explain how people learn to turn words they hear into words they can say. Wernicke himself proposed something close to it, an “auditory image” of each word that guides speech, before the region came to be filed under comprehension alone. Alexander Silva, a UCSF medical researcher and the study’s lead author, described the new signal in the same terms: “This is really an ‘auditory image’ of the sounds in the word that you’re planning to say.”

What changes in the textbook model

In the classic model, taught since the late nineteenth century, language runs on two hubs linked by a bundle of nerve fibers. Wernicke’s area in the temporal lobe understands, and Broca’s area in the frontal lobe speaks. Patients seemed to confirm it: damage near the temple left people struggling to produce any speech, and damage above the ear left them unable to follow others.

The UCSF data move a step once credited to the frontal lobe into the temporal lobe. The zone passes its word plan directly to the motor cortex, a handoff that was attributed to Broca’s area. Adeen Flinker, who studies language at New York University and was not involved in the study, summed up the gap it closes: “We already knew that damage to Wernicke’s area can lead to speech production errors, but this publication answers why.”

The location also matters in the operating room. Neurosurgeons map language sites before they remove tumors or epileptic tissue, and the researchers say the findings could change which sites they test, lowering the risk of cutting into the machinery that builds words.

What the study doesn’t settle

Every participant had a brain condition, epilepsy or a tumor, and long-term disease can shift where language sits in the brain. The recordings came from 28 people, the stimulation tests from 13 and the decoding from two. Those are small numbers next to brain-imaging studies, though typical for work that needs electrodes on the brain itself.

The anatomy stays fuzzy too. In a commentary published with the paper, Flinker wrote that “the precise anatomical boundaries of Wernicke’s area may remain ever elusive,” even as the study shows the region “contains neural populations critically engaged in the planning of speech itself.”

The decoding result is a proof of concept. It used a set list of words in two people whose paralysis came from brainstem strokes, and nobody with aphasia has been tested. For patients whose own damage includes this zone, there is no signal left to read. The work also leaves Broca’s area in an awkward place. Broca’s area “is still clearly important,” Leonard said. “In some ways, it’s sort of the next mystery.”

Common questions about Wernicke’s area

What does Wernicke’s area do?

Wernicke’s area is a region of the left temporal lobe, above and behind the ear, long known for understanding spoken language. The new study shows that a small zone inside it also plans whole words about a quarter of a second before a person says them.

What is Wernicke’s aphasia?

Wernicke’s aphasia is a language disorder caused by damage to Wernicke’s area. People speak fluently and with a normal rhythm but swap words or sounds and struggle to understand others, which can turn their speech into what clinicians call word salad.

What is the difference between Broca’s area and Wernicke’s area?

In the classic model, Broca’s area in the frontal lobe produces speech and Wernicke’s area in the temporal lobe understands it. The UCSF findings blur that split: part of Wernicke’s area plans words and sends them straight to the motor cortex, which leaves the exact job of Broca’s area as an open question.

Could this help people who cannot speak?

The team decoded the words two paralyzed people tried to say from this zone, which points to a new target for brain-computer interfaces, including for some people with aphasia. No one with aphasia has been tested yet.

The UCSF group presented the work at the Society for Neuroscience meeting in Washington, D.C., on November 15, and the paper appeared in Science on November 19, alongside Flinker’s commentary. Silva sees the zone as a way to reach people whose language loss comes from damage to the cortex, a far larger group than the patients today’s speech implants serve. “We really think that this is an area that could expand the number of people that could benefit from the technology in the near future,” he said.

Reference: Silva et al., “A neural code for generative speech production in human Wernicke’s area,” Science, 2026. DOI: 10.1126/science.aef8963

Tags: , , , , ,

Add us on Google

Discussion

There are 0 comments.