Researchers have built a "speech clock" from recordings of 2,928 Spanish-speaking participants across five Latin American countries, and report in Science Advances that the gap between how old a person's speech sounds and their actual age tracks with several measures of dementia and aging. The study was part of the Multi-Partner Consortium to Expand Dementia Research in Latin America (ReDLat), and the participants ranged from cognitively healthy adults to people with mild cognitive impairment, Alzheimer's disease, and two forms of frontotemporal dementia.
The idea rests on a simple observation: aging changes how people talk. Slower processing can lengthen pauses and slow word retrieval, affect can flatten, and vocabulary can either grow richer in healthy aging or deteriorate in dementia. Using standardized tasks such as video description, verbal fluency and story retelling, the team used an automated pipeline to estimate a "speech age" for each person and compare it with their chronological age. A positive speech age gap (SAG) means the speech sounded older than the person's years; a negative one means it sounded younger or better preserved.
The largest gaps appeared in people with language-dominant frontotemporal dementia, followed by non-language-dominant frontotemporal dementia, Alzheimer's disease and mild cognitive impairment. Larger gaps were linked to poorer memory, executive function and global cognition, and they also lined up with brain-imaging age measures, epigenetic clocks and, in people with Alzheimer's disease, higher plasma p-Tau217 levels. Larger gaps were additionally associated with more adverse social conditions such as lower education and food insecurity, though the authors stress this is an association rather than proof of cause.
For caregivers and researchers, the appeal is practical. Many existing aging markers require expensive scans, blood tests or specialized equipment, while speech can be recorded and analyzed automatically. The authors describe the approach as a potential "functional expression of systemic aging," and they point to lower-resource settings with high dementia burden as places where such a tool could matter most.
The study is cross-sectional, so it captures how old speech sounds at one point in time rather than showing that someone is aging faster. The recordings came from different centers under varying conditions, the automated tools may carry biases, and the mild cognitive impairment and language-dominant groups were small. All participants were Latin American and used structured tasks, so the results may not carry over to other languages. The authors say longitudinal studies are needed to test whether these gaps predict future change.