Summary:
A multicenter study of nearly 3,000 people published in Science Advances shows that machine-learning models can estimate a person’s biological age and brain health from voice recordings. The study introduces a “speech clock” that produces a “speech age gap” — the difference between speech-predicted age and chronological age — which strongly correlates with neuroimaging measures of brain age, epigenetic DNA-methylation clocks, Alzheimer’s blood biomarkers (plasma p-tau217), and indicators of lifelong social adversity. This approach offers a scalable, low-cost, noninvasive window into systemic biological aging and dementia risk.
Key Facts:
- The “Speech Age Gap”: Machine-learning models extract hundreds of acoustic and linguistic features — including pitch, pause duration, speech rate, semantic precision, and syntactic complexity — to predict a speaker’s age. The gap between predicted and chronological age flags accelerated or decelerated biological aging.
- Multimodal Biological Convergence: A larger speech age gap is linked to MRI markers of brain atrophy and disrupted connectivity, faster epigenetic aging across multiple DNA-methylation clocks, and higher plasma p-tau217 in Alzheimer’s disease patients.
- Underrepresented Cohort & Social Determinants: The research analyzed 2,928 Spanish-speaking participants from five Latin American countries and found that lifelong social adversity (lower education, food insecurity, reduced healthcare access) is associated with faster speech aging.
Source: Trinity College Dublin / Global Brain Health Institute
Measuring how quickly the human body and brain age usually requires costly or invasive technologies such as high-resolution MRI, PET scans, epigenetic methylation arrays, or fluid biomarker assays. Those methods are accurate but are often inaccessible in low-resource or rural settings. By contrast, speech is universally produced and reflects the integrated function of respiratory mechanics, neuromuscular control of the vocal folds, auditory feedback, and higher-level cognitive and language networks.
An international team led by the Global Brain Health Institute (GBHI) and Trinity College Dublin developed a computational speech clock that estimates biological aging and dementia-related signals from simple voice recordings. Published in Science Advances, the study shows that the speech age gap provides a cross-system measure linking neurodegeneration, epigenetic aging, and environmental adversity.
“Our voice carries far more information about aging than we realized,” said senior author Agustín Ibáñez, Ph.D., Professor in Brain Health at Trinity College Dublin’s School of Medicine and GBHI. “Speech captures chronological time as well as cognitive, neural, systemic biological signals and the cumulative effects of social experience. If validated further, speech clocks could complement more expensive aging measures.”
Analyzing Nearly 3,000 Individuals Across Latin America
To address the historic bias toward high-income cohorts in dementia research, the team evaluated 2,928 Spanish-speaking participants from Argentina, Chile, Colombia, Mexico, and Peru. The cohort spanned cognitively healthy older adults, people with mild cognitive impairment (MCI), and patients diagnosed with Alzheimer’s disease or frontotemporal dementia (FTD) variants.
Instead of focusing on single speech metrics, researchers used supervised machine-learning models trained on hundreds of simultaneous acoustic and linguistic features to predict chronological age. Key feature categories included:
- Acoustic Dynamics: Fundamental frequency (pitch), speaking cadence, pause length, vocal timbre, and emotional prosody.
- Linguistic Structure: Vocabulary richness, lexical density, semantic coherence, syntactic complexity, and overall verbal output patterns.
These extracted features were combined into a composite index that estimated an individual’s speech-predicted age and produced the speech age gap (SAG) for each person.
A Unified Signal of Brain, Epigenetic, and Molecular Decline
Participants with an accelerated speech clock — whose voices appeared older than their chronological age — showed consistent evidence of accelerated aging across multiple biological domains:
- Structural and Functional Brain Age: Larger speech age gaps were associated with MRI-derived indicators of brain atrophy, especially cortical thinning, and with weaker functional connectivity patterns.
- DNA Methylation Clocks: SAG correlated with accelerated epigenetic age measured by three independent DNA-methylation clocks.
- Plasma Biomarkers: In Alzheimer’s patients, greater speech age gaps aligned with higher plasma levels of phosphorylated tau 217 (p-tau217), a sensitive marker of Alzheimer’s pathology.
- Global Cognition and Daily Function: Bigger gaps reflected worse executive function, reduced working memory, and decreased independence in daily activities, including on non-language cognitive tests.
The speech clock also discriminated diagnostic groups: healthy controls tended to have the smallest speech age gaps, while progressively larger gaps appeared in MCI, Alzheimer’s disease, and frontotemporal dementia subtypes.
The Social Exposome Leaves Its Mark on Speech
Speech aging was influenced not only by biology but also by social environment. Across healthy and clinical groups, people who experienced adverse lifelong social conditions — such as limited education, food insecurity, inadequate medical care, or financial instability — showed significantly accelerated speech age gaps. This suggests that speech-based measures capture the cumulative impact of social determinants on biological aging.
The authors emphasize that the speech clock is an investigational research tool, not a ready-made diagnostic test. The study was cross-sectional, so longitudinal research is needed to determine whether an older-appearing speech profile predicts future cognitive decline before clinical symptoms emerge.
“Information typically obtained through diverse and costly measures appears to converge, at least in part, in the way we speak,” said Prof. Ibáñez. “If these findings hold up in longitudinal and broader population samples, speech-based biomarkers could become a scalable means for monitoring healthy and accelerated aging.”
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- Journal paper reviewed in full.
- Additional context added by staff.
About this genetics and neurodevelopment Research:
- Media Contact: Thomas Deane
- Source: TCD (Trinity College Dublin)
- Image Credit: Image credited to Neuroscience News
- Original Research is Open Access: Science Advances (Sept 30, 2026). Title: “Speech clocks decode dementia phenotypes, social exposome, and biological aging.” Authors: Hernan Hernandez; Lizeth Katherine Pedraza; Hernando Santamaria-Garcia; Sebastian Moguilner; Agustina Legaz; Pavel Prado; Jhosmary Cuadros; Lucía Amoruso; Liset Gonzalez; Damián Dellavale; Juan Pablo Espinoza–Puelles; Javier Palma Espinosa; Cecilia Jarne; Fabio Mattiussi; Matías Caccia; Alejandro Sosa Welford; Nicolás Pelella; Jeremías Inchauspe; Franco J. Ferrante; Gonzalo Pérez; Marcelo Adrián Maito; Guido Rocatti; Maria Eugenia Godoy; Joaquin Migeot; Paulina Orellana; Ariel Caviedes; Martin Bruno; Leonel Takada; Andrea Slachevsky; Maria I. Behrens; Bárbara Bruna; David Aguillon; Lina Zapata; Jose Alberto Avila-Funes; Nilton Custodio; Bruce Miller; Maria Luisa Gorno-Tempini; Stefanie Pina Escudero; Pablo Reyes; Kun Hu; Maira Okada de Oliveira; Carlos Coronel-Oliveros; Josephine Cruzat; Juan Felipe Cardona; Michael Corley; Irene B. Meier; Vaibhav A. Narayan; Enzo Tagliazucchi; Sandra Baez; Claudia Duran-Aniotz; Adolfo M. García; and Agustin Ibanez.
- DOI: 10.1126/sciadv.aef9864
Abstract
Speech clocks decode dementia phenotypes, social exposome, and biological aging
Biological aging clocks estimate aging and dementia risk but often lack scalability. We present a large-scale, cross-national speech clock derived from 2,928 individuals across five Latin American countries, including healthy controls (HCs), mild cognitive impairment (MCI), Alzheimer’s disease (AD), and frontotemporal dementia subtypes (non-language-dominant and language-dominant FTD). Multimodal acoustic and linguistic features were trained with supervised models to estimate chronological age, producing speech age gaps (SAGs) as cross-sectional markers of deviation from chronological age. Positive SAGs indicate relatively older-appearing speech.
SAGs distinguished diagnostic groups (HCs < patient groups, with AD < non-language-dominant FTD < language-dominant FTD), and aligned with clinical and cognitive measures. SAGs correlated with phosphorylated tau (p-tau217) in AD and with the social exposome in HCs and AD. Brain-derived age measures (structural, functional, combined) associated with SAG in AD and FTD groups. Epigenetic age measures correlated with SAGs in HCs and AD across Hannum, Retroclock, and OMICmAge, while language-dominant FTD showed more selective associations. These findings suggest that speech age gaps capture multilevel aging-related variation and could serve as a scalable, culturally adaptable biomarker for research in underrepresented settings.
