A new study published in Scientific Reports has found that older adults who talk more throughout their day tend to perform better on tests of cognitive ability—particularly working memory, processing speed, and semantic fluency. The researchers used a novel, unobtrusive method to track speech in real life, offering an objective window into how social interaction and cognition are linked in healthy aging.
The study was driven by a long-standing question in aging research: how do social relationships and communication impact cognitive health in older adulthood? While previous studies have shown that staying socially active can help preserve thinking and memory skills, the methods used to measure social engagement have varied widely. Researchers often rely on broad questionnaires that ask people to recall how often they’ve socialized, attended events, or talked with others—approaches that can be affected by memory errors, personal bias, and inconsistent definitions.
To overcome these issues, the research team wanted a more direct and objective way to measure social activity. They chose to focus on one simple but revealing indicator: how much people actually speak during their day. The idea was that speech—especially spontaneous, everyday speech—is a strong signal of real-time social engagement. The researchers aimed to see whether this measure of speech could be meaningfully linked to specific cognitive abilities and life circumstances.
“We were interested in this topic because social engagement has been shown to be important for maintaining cognitive abilities in older adults, but measuring social activity objectively has been challenging. We wanted to explore using advanced technology to unobtrusively measure older adults’ real-life speech as an indicator of social activity, and examine how it relates to various cognitive abilities,” explained study author Patrick Neff, group and scientific project lead at the University of Zurich.
The study included 83 healthy older adults living in Switzerland, aged 65 and up. Participants were part of a larger research project that monitored daily life using wearable sensors. Each person wore a small custom-built recording device, the size of a button, called the “uTrail.” This device captured 50-second audio snippets every 18 minutes during waking hours over the course of four weeks, producing a detailed sample of each person’s real-world speech patterns.
To analyze this large amount of data, the researchers used a machine learning algorithm called Vocalise, which was trained to identify each participant’s voice using a short, clean sample recorded in a lab. This allowed the team to automatically detect when a participant was speaking and for how long—without needing to transcribe or listen to every recording. They then calculated each person’s total “speech ratio,” or the amount of time they were talking as a percentage of the total recording time.
The participants also completed a series of cognitive tests in the lab at the beginning of the study. These tests covered a range of mental abilities, including working memory (the ability to hold and manipulate information in mind), processing speed, verbal fluency (ability to name words quickly based on prompts), executive functioning, episodic memory, and verbal knowledge. In addition, participants provided detailed sociodemographic information, including age, gender, education, income, partner status, number of people in their household, feelings of loneliness, and a self-rating of their hearing ability.
Using a statistical approach called Elastic Net regression, the researchers examined which variables were most strongly linked to how much people spoke in their daily lives. Three cognitive abilities stood out. People who scored higher on working memory, processing speed, and semantic fluency tests were more likely to speak more often throughout the day.
“Higher levels of working memory, processing speed, and semantic fluency were associated with more speech in everyday life for older adults,” Neff told PsyPost. “This suggests these cognitive abilities may be particularly important for engaging in social conversations.”
“More broadly, our study demonstrates the value of using naturalistic data to gain insights into real-life behavior and cognitive functioning. By unobtrusively recording participants’ speech in their daily lives, we were able to capture authentic patterns of social interaction that may not be apparent in traditional lab-based studies. This approach provides a more ecologically valid picture of how cognitive abilities relate to social engagement in older adults’ everyday experiences.”
Interestingly, some sociodemographic factors also predicted speech behavior. The most powerful predictor was whether someone had a partner. Older adults without a partner—whether single, divorced, or widowed—tended to speak significantly less than those in a relationship. This supports the idea that partners play a key role in daily social interaction during later life.
Another surprising finding was that people who rated their hearing as better actually spoke less. One possible explanation is that those with hearing difficulties may overcompensate by talking more to avoid the challenges of listening, or may find themselves in more one-sided conversations.
“We were surprised to find that better subjective hearing status was associated with less speech in everyday life,” Neff said. “This was unexpected and may indicate that those with poorer hearing engage in more one-sided conversations where they do most of the talking. This could be seen as some kind of compensation strategy of individuals with bad hearing to engage with their (social) environment.”
The study also found no meaningful link between age and daily speech. In other words, within this relatively healthy sample of older adults, age alone did not determine how much someone talked. This may reflect the fact that all participants were cognitively intact and relatively high-functioning, so age-related decline may not have been strongly present.
“The other big surprise was to see that age really did not have any influence on the amount of uttered own speech,” Neff explained. “Here, we might have a bias of very active and healthy older adults participating in the study (selection bias).”
“Our partly surprising results highlight the importance of keeping an open and exploratory mindset when examining real-world data. These findings may reveal complexities in older adults’ social and cognitive functioning that are not captured by conventional research methods. Overall, this study showcases how innovative technologies and naturalistic observation can enhance our understanding of cognitive aging and social behavior in later life.”
Despite the innovative methods and rich dataset, the study has limitations. The sample was limited to 83 people and skewed toward healthy, well-functioning older adults. That means the findings might not apply to those with cognitive impairments or more diverse backgrounds. The machine learning method, while powerful, also comes with technical constraints. For example, the software only detected speech in five-second chunks.
“With the recent developments in AI in the last two years, we already transcended the technical options of the time when this study was conducted and analyzed,” Neff noted. “We’re currently working on a follow-up paper, where we a) further analyze own vs. other speech and their interactions as well as b) further elements of the auditory scene (environmental sounds, noises, media consumption etc.).”
“Our long-term goals are to further refine methods for unobtrusively measuring social activity in daily life and to conduct longitudinal studies examining how cognitive abilities and social engagement influence each other over time in older adults. We also aim to explore potential interventions to enhance social activity and cognitive functioning. The focus here is on real life data collection and application, using techniques like semantic activity analysis and transdisciplinary teams.”
The study, “Cognitive abilities predict naturalistic speech length in older adults,” was authored by Patrick Neff, Burcu Demiray, Mike Martin, and Christina Röcke.
Leave a comment
You must be logged in to post a comment.