# Subtle changes in speech may help predict the risk of dementia

Researchers at Mass General Brigham, in Boston, are using sensors and artificial intelligence to analyze some 130 features of speech and flag who is at higher risk of Alzheimer's before symptoms appear. In early results, the voice did better than today's cognitive tests.

- Section: Longevity
- Published: November 22, 2026
- Address: https://ceoone.com.br/en/article/subtle-changes-in-speech-may-help-predict-the-risk-of-dementia
- Em português: https://ceoone.com.br/artigo/mudancas-sutis-na-fala-podem-ajudar-a-prever-o-risco-de-demencia.md

![An older white-haired man speaks into a microphone on a table while a researcher in a lab coat listens; beside them, a laptop shows the sound wave of his voice](https://pub-b80e49b2737940b3ab5d3f20d0578192.r2.dev/artigos/mudancas-sutis-na-fala-podem-ajudar-a-prever-o-risco-de-demencia-fala.webp)

AI-generated

## Key points

- Researchers at Mass General Brigham, in Boston, analyze some 130 speech features with sensors and artificial intelligence to flag Alzheimer's risk before symptoms.
- In a study of 49 people without symptoms, speech analysis got the risk group right up to 77% of the time, against 47% for today's cognitive tests.
- SpeechDX is gathering three years of data from 2,000 people in the US, Australia and Spain, and the voice is also studied as a sign of depression, Parkinson's and diabetes.

It is not every day that a 76-year-old woman is asked to say nonsense, repeating "ga ba da" and "ka pa ta" as fast as she can. But Deb, who lives near Boston, does just that with sensors attached to her tongue, forehead, chin, jaw and lips, recording the tiniest muscle movements in the name of science. The goal is to analyze changes in people's speech to try to predict how likely they are to develop dementia and to track changes in brain health.

"We are seeing differences that humans cannot see or really detect with their ears, but that are there years and years before cognitive decline", says Marziye Eshghi, the researcher who leads the work at the Mass General Brigham Institute of Health Professions. The project measures some 130 speech-related features in participants, and the data are processed by artificial intelligence algorithms and compared with known biological risk factors for Alzheimer's.

In preliminary results, the speech data did better than cognitive tests at identifying risk in older adults without cognitive impairment. "What we are doing is identifying individuals at higher risk of developing Alzheimer's disease who do not yet show any symptoms", she says.

## What the voice reveals

More and more, researchers are embracing an idea once seen as intriguing but fairly marginal: that speech, and the way it changes over time, can help predict and track cognitive decline. Technological advances are making it possible to collect and analyze many aspects of speech, including how fast people talk, how often and how long they pause, the complexity and variety of their vocabulary, the pitch and cadence of the voice and the coordination of vocal and facial muscles.

"It really is a window into brain health", says Dr. Marilu Gorno Tempini, director of the language neurobiology lab at the University of California, San Francisco. The way we can analyze speech and language today, she says, offers information on the detailed workings of the brain and on cognition that was unimaginable until recently.

Researchers envision a future in which everyday speech, which people can record on their own, is analyzed for gradual markers of brain health. Combined with brain scans and blood tests, it could be used to predict the risk of dementia and track the progression of cognitive impairment.

Dr. Steven Arnold, an Alzheimer's specialist and professor of neurology at Harvard Medical School who collaborates on some of Eshghi's studies, says more research is needed before speech analysis has a role in diagnosing or monitoring neurological diseases. "I still consider this an early-stage area, but a potentially very important one for the field of early dementia detection", he says.

## What the study found

In the study Deb took part in, 49 people without cognitive impairment were classified as low, moderate or high risk for Alzheimer's, based on blood levels of a protein linked to the disease and on the presence or absence of a genetic variant associated with higher risk. Then, without knowing which category each participant was in, the researchers analyzed recordings of their speech and tested whether a computer model could tell each person's group.

The speech-based analysis classified participants correctly up to 77% of the time, well above the 47% accuracy of the cognitive test models in use today. Deb is in the low-risk category. Participants are being reassessed several times to follow how their speech changes, and one of the goals is to find which aspects of it best signal the early brain changes of Alzheimer's.

Eshghi's research combines speech analysis with other assessments, such as brain imaging, blood biomarker measurements and genetic tests. Melissa Lee, director of digital and artificial intelligence initiatives at the Alzheimer's Drug Discovery Foundation, a nonprofit, leads a larger study, SpeechDX, which is collecting three years of speech data, along with blood tests, brain scans and cognitive assessments, from 2,000 people in the United States, Australia and Spain. The study, which includes speakers of English, Spanish and Catalan, will provide data for other researchers to build models that use speech to predict someone's cognitive state.

## An everyday activity

Speech is a mine of information because it involves many elements, from word production to memory, attention, expression, processing speed, awareness of context and muscle movement. It is also an everyday activity that requires no long or costly tests, such as brain scans and lumbar punctures.

"Most people talk, and talking is a very cognitively complex task", says Rhoda Au, professor of anatomy and neurobiology at Boston University. Voice recordings can be stored indefinitely and reanalyzed as new information emerges. A pioneer in studying speech as a marker of cognitive change, she says that years ago nobody listened to her, and that her colleagues react differently today. In the future, she says, privacy will have to be addressed, so that patients' results are not shared without consent. "Speech is extremely promising and very powerful. We just want to be able to use it in ways people give us permission for."

Speech and voice quality are also being studied for signs of other health conditions, such as depression, which can show up in subtle changes in speech rate and pauses, Parkinson's, in which speech can become quieter or slurred, and diabetes, which can come with hoarseness.

## Sources and editorial notes

- O Estado de S. Paulo, Metrópole section, "Mudanças sutis na fala podem ajudar a prever o risco de desenvolver demência", by Pam Belluck, The New York Times, 22/11/2026.

CEO ONE · https://ceoone.com.br

