Researchers have developed a machine-learning speech clock published in Science Advances that estimates chronological age from hundreds of acoustic and linguistic features. The difference between actual age and speech-predicted age correlates with independent markers of biological aging, brain health, cognition, social adversity, and dementia.
Acoustic and Linguistic Speech Features
Scientists analyzed 2,928 Spanish-speaking participants from Argentina, Chile, Colombia, Mexico, and Peru. The cohort included healthy adults alongside individuals diagnosed with mild cognitive impairment, Alzheimer’s disease, and various forms of frontotemporal dementia. Rather than relying on a single voice property, machine-learning models evaluated hundreds of characteristics. These included speech rate, pauses, pitch, emotional content, vocabulary, semantic precision, and the amount and organization of verbal output.
Biological and Cognitive Correlations
Participants whose speech appeared older than their actual chronological age displayed accelerated aging across multiple clinical systems. Researchers found that the speech age gap correlated with brain age measurements taken from structural and functional neuroimaging. It also aligned with epigenetic aging determined by three independent DNA-methylation clocks.
Greater speech-age acceleration corresponded to poorer global cognition, executive function, functional abilities, and memory performance. These relationships extended beyond purely linguistic tests, correlating with non-linguistic cognitive measures as well.
Clinical Differentiation and Biomarkers
The speech clock successfully differentiated healthy individuals from people with dementia. Healthy participants recorded the lowest speech age gaps, while progressively larger gaps appeared across Alzheimer’s disease and frontotemporal dementia groups. Agustin Ibanez, Professor in Brain Health at the Global Brain Health Institute and School of Medicine at Trinity College Dublin and senior author of the study, noted that the voice captures chronological time alongside signals from cognition, systemic biology, and the social environment.
In cases of Alzheimer’s disease, the speech-derived measure associated with higher levels of plasma p-tau217, a key blood biomarker of Alzheimer’s pathology. Accelerated speech aging among healthy individuals and dementia patients correlated with an adverse social exposome, encompassing lifelong factors such as education, financial conditions, food insecurity, healthcare access, and early-life experiences.
Remote Assessment Potential and Study Limitations
Traditional biological aging markers often require MRI scanners, blood samples, molecular assays, or specialized clinical evaluations. Speech recordings offer a remote, repeatable, non-invasive, and low-cost alternative that could benefit regions with limited access to advanced diagnostic technologies. Because the study focused on five Latin American countries, it provides evidence that sophisticated aging biomarkers do not rely exclusively on high-resource settings.
Researchers emphasize that the speech clock is not yet a diagnostic test for dementia. Because the research was primarily cross-sectional, it cannot establish whether an older-appearing speech profile predicts future cognitive decline or dementia. Establishing clinical utility will require longitudinal studies, validation across additional languages and cultures, and testing in naturalistic speech environments.