Overview
- The study, which was published on Sept. 30, 2026 in Science Advances, trained models on voice tasks from 2,928 Spanish-speaking adults in Argentina, Chile, Colombia, Mexico and Peru to predict chronological age.
- Researchers extracted more than 700 acoustic and linguistic features—including pitch, pauses, speech rate, vocabulary richness and semantic precision—to build a composite 'speech clock' that yields an individual 'speech age gap.'
- Positive speech age gaps, where a person’s voice sounds older than their true age, were larger in people with mild cognitive impairment, Alzheimer’s disease and frontotemporal dementia and tracked poorer memory, executive function and daily abilities.
- Speech age gaps correlated with independent biological measures: MRI-derived brain aging, three DNA-methylation epigenetic clocks and higher plasma p‑tau217 levels in Alzheimer’s patients.
- Authors say the approach could offer a low-cost, scalable screening signal for under-resourced settings but stress it is investigational and needs prospective testing across languages, natural speech settings and diverse populations before clinical use.