How Old is Your "Way of Speaking"? Exploring Brain Aging Through Voice: Possibilities and Limitations

You know what you want to say, but the words don't come out immediately. You find yourself pausing longer in conversations than before. Have you noticed such changes in family conversations?

On the other hand, choosing words slowly might be characteristic of someone's way of speaking. Distinguishing between a calm speaking style and health-related changes is not easy.

Researchers are currently exploring whether combining these speech characteristics can provide clues about aging and brain health.

A study published in the scientific journal "Science Advances" and introduced by the German newspaper WELT used a "speech clock" to estimate age from voice. It's essentially an "age clock" based on the way one speaks.

However, it's important to note upfront that this study does not indicate that "people who speak slowly have dementia."

Not Just a Study Measuring "Vocal Youthfulness"

The study involved 2,928 Spanish speakers living in five countries: Argentina, Chile, Mexico, Peru, and Colombia.

The breakdown included 1,504 individuals with healthy cognitive function, 1,068 with Alzheimer's-type dementia, 332 with frontotemporal dementia, and 24 with mild cognitive impairment.

Participants engaged in tasks such as retelling short stories, describing video content, and listing as many animals, vegetables, or words starting with a specific letter as possible within a minute.

The research team analyzed over 700 features from these recordings. The focus wasn't just on pitch or speaking speed but also on syllable length, pauses, word choice, emotional expression, and information content in the speech.

In other words, it's not simply about scoring "youthful voice" or "smooth speaking." It's an attempt to capture both "how one speaks" and "what one speaks about."

Focusing on the Gap Between Actual Age and "Speech Age"

The study developed a computer model to estimate actual age from these speech characteristics.

The focus was on the "Speech Age Gap," which represents the difference between estimated age and actual age. In this article, it's referred to as the "age gap in speech."

To simplify, if someone is 65 years old and the model estimates their speech age as 75, the gap is plus 10 years.

However, this doesn't mean the brain has aged precisely 10 years. It's merely a numerical representation of how the age estimation model evaluated the person's speech.

According to WELT, the average error in age estimation was about 9 years. There's a significant margin for declaring "your brain age is this" for an individual.

The important aspect of this study is not pinpointing each person's age accurately but understanding how this discrepancy relates to other health indicators.

Correlation with Memory, Brain Imaging, and Blood Indicators

In the analysis, participants with healthy cognitive function had the smallest average speech age gap. In contrast, those with Alzheimer's-type dementia or frontotemporal dementia tended to have speech estimated as older than their actual age.

The gap was particularly large in the subtype of frontotemporal dementia where language impairment was prominent.

Additionally, a higher estimated speech age correlated with lower scores on cognitive function tests examining memory and attention, with a notable link to memory.

Speech characteristics were also associated with biological indicators.

In Alzheimer's patients, a higher speech age was linked to higher concentrations of "phosphorylated tau 217 (p-tau217)" in the blood, a marker used to capture brain changes related to Alzheimer's disease.

Furthermore, correlations with age estimated from brain imaging and aging indicators calculated from DNA chemical changes were reported.

These results suggest that information in conversations holds broader potential than the "impression of voice" perceived by the ear. However, having a correlation doesn't mean voice can replace blood tests or brain imaging.

"Current Differences" and "Future Onset Prediction" Are Separate Matters

The most crucial aspect when reading this study is the limitations of the research method.

The study is primarily a "cross-sectional study" comparing participants at a single point in time. It examines how speech and health status are linked simultaneously but doesn't confirm whether a person will develop dementia years later.

For example, the model might have estimated the speech of someone whose language functions have already changed due to illness as older. This doesn't mean it can predict the future of asymptomatic individuals.

Also, even if the estimated age from a single recording is high, it doesn't directly measure that person's aging as "progressing rapidly." To know the progression speed, repeated examinations over time are necessary.

Researchers also explain that the current speech clock is not a diagnostic test for dementia.

There's still a significant gap between the discovery that "health clues are in the voice" and the claim that "diseases can be identified solely by voice."

Reactions on Social Media: "Can It Be Used in Everyday Conversations?"

In a LinkedIn post by researcher Agustin Ibanez introducing this study, alongside congratulations on the achievement, specific questions about practical application were raised.

A museum education professional, who introduced themselves as working with people with dementia, focused on the task of narrating the content of a short video. They pointed out similarities with situations where participants discuss artworks in museums and questioned whether the same indicators apply in natural conversations, not just in set tasks.

Another comment suggested that reproducing results in different languages would be the next major challenge.

These are just some of the public comments and do not represent the overall opinion on social media. Nonetheless, it's interesting that the expectations for the research are not only about "wanting to use it immediately" but also about interest in verifying whether it can be applied in everyday situations.

On the other hand, more skeptical opinions were seen in the reader comment section of the WELT article provided.

There were remarks pointing out that the difficulty in speaking due to dementia is not new and questions about whether vocabulary tasks are influenced by educational background or inherent language ability.

Although this comment section is separate from social media, it raises the issue of how to interpret the numbers. A system is needed to avoid mistaking differences in speaking ability for differences in health.

Cannot Be Directly Applied to Japanese

Japanese readers might be curious whether similar analyses can be conducted in Japanese conversations.

All participants in this study were Spanish speakers. Although the study included five countries, the effectiveness in Japanese has not been confirmed.

Japanese includes expressions that omit subjects, interjections, and honorifics, with phrasing changing depending on the listener and situation. Pausing with "um" can serve roles beyond recalling words.

Therefore, using it in Japanese would require verification with collected Japanese speech. It must also be confirmed whether differences in dialects, educational background, and commonly used languages can be fairly evaluated.

At least from the current results, one cannot conclude that "the faster someone speaks, the younger their brain" or "using more complex words indicates health."

Similarly, the study does not show that increasing speaking speed or intentionally reducing pauses can prevent aging. Imitating characteristics that appear in indicators and improving health are separate matters.

The Hope is for a Supplementary Evaluation That Can Be Repeated Easily

The strength of using voice is that the burden of recording itself is relatively low. It can be collected without large testing equipment and easily recorded repeatedly even in remote areas.

In the future, if sufficient verification is conducted, it may become a supplementary means to understand health status in combination with other tests.

However, it's premature to conclude that "talking to a smartphone can determine the presence of dementia." The conditions of the voice obtained from tasks set in research and casual conversations or phone calls at home differ.

Considering practical application, not only accuracy but also explanations to the person and handling of recordings will be important. Caution is needed for systems where conversations with family are used for health assessments without the person's knowledge.

Before being data for analysis, a voice is something that conveys a person's experiences and feelings.

Rather Than "Judging" Speech, Gently Acknowledge Changes

This study suggests that everyday conversations may contain clues about brain and body conditions.

At the same time, it clarifies that converting speech into an age number does not reveal everything. There are significant remaining challenges, such as age estimation errors, unverified future predictions, and differences in language and culture.

What we should take from this research is not a standard for counting family silences or doubting those who speak slowly.

There may be information in familiar voices that hasn't been fully captured before. Whether we can interpret that information in a useful way while preserving the person's dignity is the question.

The research on "speech age" is just at its beginning.

Sources and References

1. WELT
   Referencing the breakdown of study subjects, speech tasks, over 700 analytical features, and an average estimation error of about 9 years.
   https://www.welt.de/wissenschaft/article6ac21ba685d64113fb06e82c/gehirn-was-sprechtempo-wortwahl-und-pausen-ueber-das-altern-verraten.html

2. Trinity College Dublin: Research Institution Announcement
   Confirming the research summary, that it is not a diagnostic test for dementia, and the need for long-term tracking, multilingual, and natural conversation verification.Trinity College Dublin
   https://www.tcd.ie/news_events/top-stories/featured/voice-may-reveal-how-well-you-are-aging/

3. LinkedIn: Agustin Ibanez's Post Introducing the Research and Public Comments
   Confirming comments questioning the application to natural conversations and reactions regarding reproducibility in other languages. The main points were summarized in Japanese in the text.Agustin Ibanez
   https://www.linkedin.com/posts/agustin-ibanez-b727172b_speech-aging-clocks-ai-models-estimating-activity-7511760127016730625-relF

4. Science Advances: Journal Where the Research Paper is Published
   The paper is titled "Speech clocks decode dementia phenotypes, social exposome, and biological aging." The publication site was confirmed, but the paper itself was not directly obtained, so the research explanation is based on the provided article and research institution announcement.
   https://www.science.org/doi/10.1126/sciadv.aef9864