A computer model that listens to a few minutes of speech can estimate a “speech age” that runs older in people with dementia than in healthy adults, according to a study of 2,928 adults across Latin America published in the journal Science Advances1. The gap between this speech-based estimate and a person’s real age was also linked to brain scans, DNA-based measures of aging, a blood marker of Alzheimer’s disease and lifetime social hardship1. The researchers stress that it is not yet a test for dementia2.
What happened
Scientists use “aging clocks” to estimate whether a person’s body seems older or younger than their calendar age. An aging clock is a statistical model that takes physical or behavioral data and predicts a person’s age; the difference between that prediction and real age is called the age gap1. A positive gap is usually read as faster aging, and a negative gap as slower aging1.
Most established clocks rely on brain scans (MRI) or blood tests that read chemical tags on DNA, known as methylation1. Both are expensive and hard to use in lower-income regions1. The team wanted to know whether ordinary speech could do some of the same job1.
The researchers built what they call a “speech clock,” a machine-learning model, meaning software that learns patterns from examples, that estimates age from sound and language features of a recording3. Outlook India reported that it analyzes hundreds of features of a person’s speech2.
The evidence so far
The study included 2,928 native Spanish speakers in Colombia, Peru, Mexico, Chile and Argentina, with an average age of about 65; 67 percent were women1. Of them, 1,504 were cognitively healthy and 1,424 had a diagnosed brain condition1. That group included 24 people with mild cognitive impairment (an in-between stage of memory and thinking problems before dementia), 1,068 with Alzheimer’s disease and 332 with frontotemporal dementia, a group of disorders that damage the front and side areas of the brain1.
Everyone did seven short speaking tasks in a quiet room. These included describing an animated video, retelling a short story right away and again 20 to 30 minutes later, and naming as many words as possible in 60 seconds that start with a certain letter or belong to a category like animals1.
Healthy participants had the smallest speech age gaps, and the patient groups had larger ones3. The largest gaps were in people with the language-type of frontotemporal dementia, followed by the non-language type and then Alzheimer’s disease3. These differences held after the researchers adjusted for age, sex, education, country, other illnesses and the number of languages a person spoke3.
The speech measure also lined up with other signs of aging:
- People whose speech sounded older than their age tended to have older-looking brains on MRI scans and faster aging on three separate DNA-methylation clocks2.
- Bigger gaps went with poorer thinking skills, everyday functioning and memory test results, including some tasks that did not depend directly on language2.
- In people with Alzheimer’s, the gap was linked to higher blood levels of p-tau217, a protein marker tied to Alzheimer’s-related brain changes2.
- Bigger gaps were linked to a harder “social exposome,” the combined effect of things like education, money, food insecurity, access to health care and early-life experiences2.
The strongest clues in speech were how much people said, the timing of their speech, their pitch and the emotional content of their words3. Slower speech and longer pauses are thought to reflect slower mental processing and trouble finding words1. Changes in pitch and emotional tone may capture the flattened emotional expression often seen in dementia, and shifts in vocabulary may point to the loss of general knowledge and facts1. The combined speech age measure separated the patient groups better than any single speech feature on its own2.
What it does not show yet
The study was mostly cross-sectional, meaning each person was assessed at one point in time rather than followed over years3. So it cannot say whether an older-sounding voice predicts who will later develop memory decline or dementia2. It also cannot show whether a larger gap reflects an aging process that keeps moving forward3.
The model was built only with Spanish speakers from five Latin American countries3. How well it works in other languages, cultures, dialects and in looser everyday conversation is unknown3. Microphones, recording conditions, education, speaking more than one language and the exact speaking prompts may all affect the results3. Pauses and word choice can also be shaped by mood, culture, hearing, illness and the room where the recording is made3.
The researchers said the speech clock is not yet a diagnostic test for dementia, and that longer studies, testing in more languages and cultures, and more natural speech settings are needed before any use in clinics2.
Who is behind it
The research was led by Hernan Hernandez of the Latin American Brain Health Institute at Universidad Adolfo Ibáñez in Chile1. The senior authors were Agustin Ibanez, a professor of global brain health at the Global Brain Health Institute at Trinity College Dublin, and Adolfo M. García, who directs the Cognitive Neuroscience Center at Universidad de San Andrés in Argentina1. Participants were recruited through a research network spanning several countries1. The sources reviewed for this article do not describe who funded the study or whether any author has a commercial stake in speech-based tests.
Ibanez told PsyPost the work grew from his group’s effort to build several aging clocks and from the fact that many established markers are expensive and hard to scale1:
We therefore wanted to test whether natural speech could provide a low-cost, non-invasive measure that captures several dimensions of aging simultaneously.
He was more expansive in comments reported by Outlook India2:
Our voice appears to contain much more information about ageing than we previously recognised.
He added that if the findings are confirmed over time and across populations, speech could become one of the most scalable tools for monitoring healthy and faster aging2. That is a hope about future studies, not something this study measured.
How it fits the field
The appeal is practical. Speech can be recorded remotely and analyzed automatically, without MRI machines, blood draws or special equipment3. That could make it useful for research and possibly for large-scale monitoring where advanced diagnostic tools are scarce3. Behavioral aging measures based on questionnaires and memory exams need trained staff and can be skewed by how well a person reads1.
Researchers are also looking at other everyday signals. In a separate study published in Neurology, 2,183 older adults wore small chest monitors for about 12 days, and those with weaker or more irregular daily rest-and-activity rhythms were more likely to develop dementia over about three years4. That study, too, showed a link rather than a cause4.
MedScriptum’s summary described the speech gap as a promising research marker that could add to clinical checks, blood tests and brain scans rather than replace them3. For more on other ways scientists try to measure aging, see Biohack’s report on 17 blood proteins tied to death risk and faster aging.
What to watch next
- Studies that follow the same people over years, to see whether a growing speech age gap comes before memory decline3.
- Tests in other languages and cultures, and with everyday conversation instead of structured tasks2.
- Checks on how much microphones and recording conditions change the results3.
Until then, the speech clock is a research tool: a sign that the way people talk may carry information about aging, measured once in one language group.
