What is it about?
This study tested whether stuttering is more likely on words that are harder for a listener to predict from the words that came before. We analyzed nearly 35,000 words from spontaneous interviews with 35 adults who stutter and used a large language model to estimate how predictable each word was. We found that less predictable words are more likely to be stuttered, even after accounting for other language-related factors. Standard measures of stuttering severity and overall impact did not predict who was more sensitive to these harder-to-predict words.
Featured Image
Photo by Volodymyr Hryshchenko on Unsplash
Why is it important?
Stuttering can vary from moment to moment, and that unpredictability is often one of the hardest parts of living with it. This study shows that words carrying more information are more likely to be stuttered. It also offers a new way to study stuttering with modern language tools and suggests that researchers and clinicians may need to look beyond severity scores alone, including the communicative pressure speakers feel when trying to express words that they deem carry value.
Perspectives
What I find most compelling about this study is that it helps explain a familiar but poorly understood feature of stuttering: its moment-to-moment variability. The findings suggest that stuttering is shaped not only by the sounds or structure of words, but also by the communicative demands of the moment. I also find it striking that this pattern was not explained by standard severity measures, which suggests there is still much to learn about how speakers experience communicative pressure. I hope this work encourages more research—and more sensitive clinical conversations—about the real demands people who stutter navigate in everyday speech.
Haley Warner
New York University
Read the Original
This page is a summary of: Information Load Predicts Stuttering Events, Journal of Speech Language and Hearing Research, March 2026, American Speech-Language-Hearing Association (ASHA),
DOI: 10.1044/2026_jslhr-25-00833.
You can read the full text:
Contributors
The following have contributed to this page
