Can Your Voice Reveal How Your Brain Is Aging?

The way a person speaks may reveal much more than their chronological age. According to a new international study, certain characteristics of the voice and language are associated with brain aging, cognitive abilities and several biological markers linked to age.

Published in Science Advances, the study involved 2,928 Spanish-speaking participants from Argentina, Chile, Colombia, Mexico and Peru. Researchers led by Agustin Ibanez of Trinity College Dublin developed an artificial-intelligence system capable of estimating a person’s age by analyzing hundreds of acoustic and linguistic features of speech.

The most notable measure is the “speech age gap”: the difference between a person’s chronological age and the age estimated by the model from their speech. When the system assigns a speech age older than a person’s actual age, the gap indicates that the person’s vocal and linguistic profile appears older than expected.

The study found that this gap was associated with several independent measures of biological aging, brain health, cognitive performance and dementia. It was also associated with social conditions accumulated over the course of life, suggesting that speech may simultaneously reflect biological, neurological and environmental dimensions of aging.

Rather than looking for a single sign of aging, the researchers combined hundreds of features using machine-learning algorithms. These included speech rate, the frequency and duration of pauses, pitch, emotional content, vocabulary, semantic precision, and the amount and organization of verbal output.

The model uses these features to estimate chronological age and calculate the difference between estimated and actual age. The researchers then examined whether this “speech clock” was associated with other measures of aging collected independently of the speech analysis. The sample included both cognitively healthy adults and people with mild cognitive impairment, Alzheimer’s disease and different forms of frontotemporal dementia.

Comparing speech with the brain and epigenetic clocks

The inclusion of different clinical groups allowed the researchers to compare speech-based age estimates with biological, neurological and cognitive measures.

People whose speech appeared older than expected for their chronological age also showed signs of accelerated aging across several biological and clinical systems. One association involved brain age, estimated using structural and functional neuroimaging. The gap between chronological age and speech age was associated with measures derived from brain imaging, suggesting that speech may capture, at least in part, processes that can also be observed through very different diagnostic tools.

The researchers also compared speech age with three independent epigenetic clocks based on DNA methylation. These tools estimate biological age through chemical modifications of DNA associated with aging. Here too, accelerated speech aging was associated with epigenetic measures.

This does not mean that speech directly measures DNA methylation or that it could replace molecular testing. Rather, the finding suggests that speech characteristics and different biological markers may contain overlapping information about the aging process.

Speech and cognitive performance

Another set of findings concerned cognitive abilities. A larger speech age gap was associated with lower performance on measures of global cognition, executive function, functional abilities and several forms of memory.

The associations were not limited to language-based tests. Speech-estimated age was also associated with performance on cognitive tests that did not rely on language, strengthening the possibility that the system captures information beyond the ability to speak itself.

The model was also used to compare cognitively healthy participants with those affected by dementia. The former showed smaller gaps between chronological age and speech age, while larger gaps were observed among participants with Alzheimer’s disease and different forms of frontotemporal dementia.

The combined measure of acoustic and linguistic characteristics distinguished the clinical groups better than individual speech parameters considered separately. This highlights the value of an integrated approach in which the machine-learning system uses information about both how people speak and what they say.

Among people with Alzheimer’s disease, speech age was also associated with higher plasma levels of p-tau217, a blood biomarker associated with Alzheimer’s pathology. The same measure was associated with cognitive and clinical functioning.

The association with p-tau217 provides another link between an everyday behavior, speaking, and a biological measure obtained through a blood test. It does not, however, demonstrate that the system can independently identify Alzheimer’s disease or replace clinical assessment.

“Our voice seems to contain much more information about aging than we previously recognized,” Ibanez said. The senior author, a professor of brain health at the Global Brain Health Institute and the School of Medicine at Trinity College Dublin, added that speech reflects “both the passage of chronological time and signals from cognition, the brain, the body’s biology and even the social environment accumulated over a lifetime.”

The “social exposome”

The social dimension is one of the distinctive aspects of the research.

The authors examined the relationship between speech age and what they describe as the “social exposome” — the collection of conditions and experiences that shape a person’s living environment. Factors considered included education, economic circumstances, food insecurity, access to healthcare and experiences during early life.

Among both cognitively healthy participants and those with Alzheimer’s disease and other forms of dementia, more accelerated speech aging was associated with a less favorable social exposome.

The finding suggests that differences in speech may reflect conditions accumulated over the course of life, rather than only neurological or biological processes. The study does not, however, establish a causal relationship between individual social factors and speech characteristics.

The research was conducted in five Latin American countries, a region that has historically been underrepresented in dementia research. This allowed the researchers to develop and evaluate the system in a Spanish-speaking population spanning different national contexts.

The possibility of extracting information about aging from simple voice recordings could be particularly relevant in countries and communities where access to MRI, molecular testing and specialist assessments is limited. Speech recordings can be collected remotely, repeated over time and obtained without invasive procedures.

These characteristics could make the approach suitable for more accessible monitoring systems, provided that its reliability and clinical usefulness are established.

The study’s limitations

The researchers stress that the system is not yet a diagnostic test for dementia.

The main limitation is the study’s predominantly cross-sectional design. Researchers compared speech characteristics with participants’ clinical and biological measures but did not systematically follow them over time. It is therefore not possible to determine whether a person whose speech appears older than their chronological age will subsequently develop cognitive decline or dementia.

Before the approach could be used clinically, longitudinal studies, validation across additional languages and cultures, and testing in more natural conversational settings will be needed. Researchers will also need to establish how reliable the estimates remain under different recording conditions and in populations other than those included in the study.

The team therefore considers the method a promising research tool, rather than a system ready to provide individual diagnoses.

For Ibanez, the broader significance of the findings lies in the convergence of information normally obtained through very different procedures. Chronological age, brain aging, epigenetic markers, cognitive abilities, Alzheimer’s-related pathology and social conditions all show, to some extent, relationships with the way a person speaks.

If the findings are confirmed in longitudinal studies and in different populations, speech analysis could become a widely accessible tool for monitoring some dimensions of healthy and accelerated aging.

For now, the study shows that integrated analysis of voice and language can detect associations with multiple dimensions of aging. It does not show that dementia can already be predicted or diagnosed from a simple voice recording.

Condividi su Facebook

Tweet