(© Krakenimages.com - stock.adobe.com)

Natural Speech Pauses Predict Executive Function Changes From Age 18 to 90 In A Nutshell

Researchers analyzed speech patterns from 241 adults aged 18-90 and found that pauses and filler words like “um” correlate with executive function abilities throughout adulthood.

Word-finding difficulty in natural speech predicted performance on cognitive tests better than standard screening tools like the Montreal Cognitive Assessment.

The findings suggest that analyzing everyday speech could provide an accessible, repeatable way to track cognitive changes over time without practice effects that plague traditional testing.

Speech timing features (pause frequency and duration) showed the strongest associations with cognitive abilities, reflecting more than just general processing speed.

The “ums” and “uhs” in your everyday speech may not be just verbal tics after all. Research demonstrates these hesitations correlate with executive function abilities, revealing insights about cognitive control, planning, and mental flexibility across the adult lifespan.

Is it time to start paying more attention when you find yourself pausing more often while describing something? According to a study published in the Journal of Speech, Language, and Hearing Research, word-finding difficulty in natural speech reflects broader cognitive changes.

The research team from the University of Toronto and Baycrest Hospital analyzed speech samples from two groups: 67 older adults aged 65–75, and 174 adults spanning ages 18–90. Each participant viewed two illustrations and described everything visible for 60 seconds per image. Software extracted over 700 features from these recordings, capturing speech rate, pause duration, and filled pause frequency.

Participants completed executive function assessments measuring working memory, inhibition, cognitive flexibility, and verbal fluency. Executive function encompasses mental processes supporting planning, attention management, instruction recall, and multitasking. These abilities typically decline with age, though at varying rates depending on whether someone experiences normal aging or develops dementia.

Speech Timing Predicts Cognitive Performance

Timing-related features of speech emerged across both studies as especially important. Across both studies, timing-related features showed the clearest and most consistent links with executive function. Adults producing more pauses and fillers during picture descriptions scored worse on cognitive control tests. This association held whether researchers examined older adults exclusively or people across the entire adult age range.

Among older adults, word-finding difficulty explained individual differences in both major executive function components identified through analysis. One component reflected inhibition and working memory abilities. The other captured verbal fluency and mental shifting capacity. Greater speech hesitations predicted worse performance on both.

The second study, including younger participants, confirmed these relationships using a different analytical approach. Researchers extracted patterns reflecting natural variation across a normative sample of 773 people. Two components related to speech timing showed the strongest associations with executive function. Adults with more medium-to-long pauses scored worse on verbal fluency and shifting tests. Those with more short pauses and filled pauses also showed weaker executive abilities, though this relationship proved less robust.

It might be time to talk to your doctor if you catch yourself struggling to find the right words regularly. (Photo by fizkes on Shutterstock) Beyond Simple Processing Speed

The connection between speech timing and executive function likely reflects more than general processing speed, though speed plays a role. Executive function tests often depend heavily on response quickness, and faster thinkers tend to speak faster.

Yet research into word-finding failures points to something more specific. The transmission deficit hypothesis proposes that problems arise specifically during the transition from identifying which word to say to retrieving how to articulate it. While general processing speed contributes to word retrieval, a substantial portion of word-finding difficulty appears more targeted.

Consider what happens when someone struggles naming an object in a photograph. The person recognizes the object, understands its meaning, knows they want to say its name, but the sound pattern temporarily escapes them. They might produce fillers while searching their mental lexicon, switch to describing the object instead, or eventually arrive at an approximate term. This process engages executive functions: maintaining goals, inhibiting incorrect responses, flexibly adjusting plans, and monitoring output.

Natural speech may capture aspects of these executive processes that standard clinic tests can overlook, because it represents genuine everyday behavior rather than artificial laboratory tasks. When describing a complex picture, speakers simultaneously plan what to say, select appropriate words, monitor output for accuracy and coherence, and adjust when word retrieval fails. This multitasking draws heavily on executive resources.

Standard Screening Tools Miss Subtle Changes

The research revealed an interesting disconnect between natural speech measures and Montreal Cognitive Assessment scores, a widely used dementia screening tool. Among 155 older adults who completed the MoCA, speech timing showed only modest correlations with total scores, and these relationships didn’t survive statistical correction.

The MoCA screens for dementia risk, and most healthy people perform at or near ceiling. The test is untimed, measuring neither response speed nor reaction time. Items most frequently missed involve episodic memory rather than the executive functions that speech timing appears to tap.

This discrepancy highlights a key limitation of brief dementia screens like the MoCA. These tools can miss subtle shifts in executive function that still fall within the “normal” score range. Speech analysis, by contrast, detects variations in processing speed and word retrieval efficiency within the normal range that still predict meaningful differences in cognitive ability.

Tracking Your ‘Ums’ Could Help You Monitor Your Cognitive Health

Perhaps the most promising application lies in longitudinal monitoring rather than single assessments. Executive function abilities vary widely among healthy adults. A person enjoying normal cognitive health in middle age would need relatively advanced neurodegeneration before falling outside the normal range on a single test.

Tracking how quickly executive function declines within individuals offers more diagnostic potential. Executive abilities decline at certain rates with age in healthy people but may decline more rapidly in those at high dementia risk. Unfortunately, comprehensive executive function testing is time-consuming, invasive, and expensive. Few adults would undergo such evaluation regularly. Moreover, practice effects complicate interpretation because performance typically improves with repeated testing.

Natural speech assessment offers a practical way around many of these hurdles. Picture description is an everyday activity unlikely to improve substantially through repetition. The task imposes minimal burden and could easily repeat every few months. Automated analysis eliminates the need for trained administrators. The approach captures processing speed in a quantifiable but naturalistic manner.

Researchers noted that different feature combination strategies worked better for different age groups and research questions. When examining older adults who might develop dementia, speech features already known to change with disease progression proved most effective. When studying wider adult age ranges, statistical patterns derived from large normative samples better captured meaningful variation.

This flexibility represents both strength and challenge. Speech contains rich information about cognitive function, but specific aspects that matter most may depend on population and clinical question. Future research combining natural speech analysis with detailed executive function test batteries could better separate speed-dependent from speed-independent components of cognitive control.

The study establishes that “ums” and “uhs” littering everyday conversation reflect genuine cognitive processes rather than mere speaking habits. Pause frequency and duration provide a window into executive function abilities that traditional testing often misses. As researchers continue refining automated speech analysis techniques, this window may offer an accessible and sensitive method for tracking cognitive health across the lifespan.

Disclaimer: This article provides general information about scientific research and is not medical advice. Consult qualified healthcare professionals for personal medical concerns or cognitive health questions.

Paper Summary Study Limitations

The research involved cross-sectional comparisons rather than longitudinal tracking within individuals. While results suggest speech timing could effectively monitor cognitive decline over time, this application requires validation through studies following the same people across months or years. The two studies used different executive function assessments, making direct comparisons between cohorts challenging. Sample sizes were modest, particularly for oldest age groups. Participants were relatively highly educated, which may limit generalizability to populations with less formal schooling. The study excluded individuals with MoCA scores below 21, potentially missing earliest stages of cognitive impairment where speech changes might be most clinically relevant.

Funding and Disclosures

This research received support from Mitacs Accelerate program internship grant FR75766 awarded to Hsi T. Wei and Natural Sciences and Engineering Research Council of Canada Discovery Grant RGPIN-2019-06515 awarded to Jed A. Meltzer. Jessica Robin is a full-time employee of Winterlight Labs, Inc., the company that developed the speech analysis software used in the study. All other authors declared no competing financial or nonfinancial interests at the time of publication.

Publication Details

Title: Natural Speech Analysis Can Reveal Individual Differences in Executive Function Across the Adult Lifespan

Authors: Hsi T. Wei (Department of Psychology, University of Toronto; Rotman Research Institute, Baycrest Hospital), Dana Kulzhabayeva (Department of Psychology, University of Toronto; Rotman Research Institute, Baycrest Hospital), Lella Erceg (Rotman Research Institute, Baycrest Hospital; Department of Cell & Systems Biology, University of Toronto), Mira Kates Rose (Rotman Research Institute, Baycrest Hospital; Speech-Language Pathology, University of Toronto), Kiah A. Spencer (Rotman Research Institute, Baycrest Hospital; Krembil Neuroscience Centre, University Health Network), Jessica Robin (Winterlight Labs, Toronto), Ellen Bialystok (Department of Psychology, York University), and Jed A. Meltzer (Department of Psychology, University of Toronto; Rotman Research Institute, Baycrest Hospital)

Journal: Journal of Speech, Language, and Hearing Research, pages 1–19
DOI:10.1044/2025_JSLHR-24-00268

Article History: Received April 25, 2024; Revision received December 20, 2024; Accepted August 14, 2025