A newly developed artificial intelligence system capable of analyzing speech patterns may offer insights into an individual’s aging process and their susceptibility to dementia, according to research published Wednesday (Sept. 30) in the journal Science Advances.
The technology examines hundreds of linguistic features to estimate a person’s chronological age based on how they speak. In a study of nearly 3,000 Spanish-speaking adults, researchers found that participants whose estimated age exceeded their actual age were more likely to exhibit signs of cognitive impairment, dementia, and accelerated biological aging.
These individuals also tended to face more challenging socioeconomic conditions compared to other participants. The findings suggest that speech analysis could serve as a non-invasive alternative to existing “aging clocks,” which typically rely on costly and invasive procedures such as brain scans or blood tests.
“Existing aging clocks are very powerful, but many of them rely on brain scans or blood tests that are expensive, invasive or difficult to repeat regularly,” said study co-author Adolfo García, director of the Cognitive Neuroscience Center at the University of San Andrés in Argentina. “All of these limitations can be overcome with speech.”
For the study, scientists utilized data from the ReD-Lat consortium, a major dementia research initiative across Latin America. The cohort included 2,900 adults aged 18 to 88 from Argentina, Chile, Colombia, Mexico, and Peru. Approximately 1,500 participants were cognitively healthy, while the remainder had conditions ranging from mild cognitive impairment to Alzheimer’s disease or frontotemporal dementia.
Participants completed seven tasks, including describing animated videos, naming as many words or animals as possible within 60 seconds, and retelling stories after delays of 20 to 30 minutes. Researchers transcribed these recordings and identified over 700 distinct features, such as pauses, speaking speed, pitch, vocabulary, and emotional tone. A machine-learning model was trained on these features to predict age, allowing researchers to calculate a “speech-age gap.”
Cognitively healthy individuals showed the smallest discrepancies between their actual and predicted ages, whereas those with dementia, particularly language-dominant frontotemporal dementia, exhibited the largest gaps. These larger gaps correlated with poorer performance on memory, language, and attention tests, as well as greater difficulty in daily functioning.
Blood test results further supported these findings. Larger speech-age gaps were associated with higher levels of age acceleration on three epigenetic clocks, which measure biological age through chemical tags on DNA. Among Alzheimer’s patients, greater gaps were also linked to elevated levels of p-tau217, a protein associated with the disease.
Additionally, the study found correlations between larger speech-age gaps and various dementia risk factors, including financial hardship, food insecurity, limited healthcare access, difficult childhoods, and lower educational attainment. García noted that these patterns remained consistent across all five countries despite their diverse social and cultural backgrounds.
Dr. Manisha Parulekar, co-director of the Center for Memory Loss and Brain Health at Hackensack University Medical Center in New Jersey, who was not involved in the study, praised the research as a significant step toward accessible brain health assessments. However, she cautioned that the tool is not definitive.
“The human voice is incredibly sensitive to our overall well-being, which means a person’s speech could sound ‘older’ simply because they are severely depressed, exhausted, or navigating serious life stress,” Parulekar said. “As a clinician, I wouldn’t use this tool to definitively diagnose someone with dementia in isolation.”
Parulekar suggested that the speech measure could instead serve as a prompt for doctors to investigate cognitive health more thoroughly. She also highlighted limitations, noting that the study assessed participants at a single point in time rather than longitudinally. Consequently, it remains unclear whether an older-sounding voice can predict future cognitive decline or if voice changes occur as dementia progresses.
Furthermore, the model was trained and tested exclusively on Spanish speakers from five Latin American nations. Parulekar emphasized that the system cannot simply be adapted for English without significant modification.
“We cannot just take the Spanish model and plug English into it,” she explained, noting that language-specific features must be adapted and tested in other populations.
Researchers plan to validate the approach in additional languages and conduct longitudinal studies to determine if speech changes can signal impending cognitive decline. “I truly believe that there is a very, very strong signal in speech to anticipate dementia and conversion in the future,” García said, “but we need many more diverse studies.”
This article is for informational purposes only and is not meant to offer medical advice.
Great science, but we need to be careful. A voice analysis shouldn’t replace a doctor’s judgment or cause unnecessary panic.
Socioeconomic factors clearly play a massive role here. It’s less about biology alone and more about the toll hardship takes on us.
I wonder how this tool performs on non-Spanish speakers. The study was limited to Latin America, so generalizability is a concern.
Wait, so my speech sounds older because I’m tired and stressed? That’s not helping my self-esteem, honestly.
This is huge for accessibility. Early detection without invasive tests could save countless lives and reduce healthcare burdens globally.