Research Summary: looking inside the vocal tract during stuttering
A new study has been published which aims to understand stuttering by recording images of the vocal tract during stuttering.
The original journal article is available here. If you cannot access it, please contact me!
An interview with the lead author - Yijing Lu - is available below.
What did this study aim to do?
This study aimed to find out what happens inside the vocal tract during stuttering. By looking at the speech movements, the researchers hoped to identify any movement patterns or characteristics relating to the beginning, middle and end of a stutter.
How did they do it?
In this study, a single participant repeated made-up words (for example “mabshibe” or “mabteebeebee”) whilst the movements of the vocal tract, including the lips, tongue and soft palate, were recorded using vocal tract magnetic resonance imaging (MRI). This is the same MRI machine that is commonly used in research and hospitals to look at the brain and other areas of the body.
Here's an example image of the vocal tract, taken using Magnetic Resonance Imaging (MRI). Here, the speaker is facing to the left. The image has been cropped to maintain confidentiality, so the eyes and the tip of the nose have been hidden.
For this study, areas of the vocal tract were defined using rectangles or circles. They are the lips (LIP), Tongue Tip (TT), Tongue Body (TB) and the soft palate or velum (VEL).
Have a look at this other blog post to see an example of a movie taken using MRI of the vocal tract.
Try an example: If you run your tongue tip from behind your teeth, back across the roof of your mouth, you will hit a soft patch. This is the "soft palate" or "velum" and it moves up or down to direct air though the mouth or the nose. Try putting a finger under your nose, whilst your hum. You'll feel the air passing though your nose! This means that the velum is lowered, allowing air into the nasal cavity!
What did the study reveal?
The results showed that, for this person:
1) Stuttering was more likely to occur on the first sound of the word and stuttering always occurred on a consonant (i.e. m, b, t, sh etc) during the experiment.
2) During a stutter, the speech muscles were in the right position and ready to make the intended sound, but they constrict the vocal tract to a greater extent than expected. For example, producing a ‘b’ sound requires the lips to close. During dysfluent utterances, the lips were more tightly closed compared to fluent utterances. The same thing occurred with the tongue tip forming the ‘t’ sound in ‘tie’ and ‘tee’– the tongue closed the airway to a greater extent than for fluent trials. This forms part of the tension that is sometimes associated with stammering and it’s difficult to separate this learned response to stammering from what might be a characteristic of stuttering to start with.
This graph shows the amount of constriction for each sound and speech muscle. The higher the dots, the more constriction. Here, you can see that the black dots (dysfluent sounds) are more likely to be higher (more constricted) than the grey dots (fluent sounds).
3) During a stutter, this increased closure stayed the same or became stronger, but did not get less over time.
This graph shows the constriction (amount of closure) of the lips during a dysfluency. The lip muscles oscillate and the closure increases over time (arrow), before being released.
4) Stuttering was not caused by difficulties forming a combined consonant-vowel sound. When we say a sound like “ma”, with a consonant (m) and a vowel (a), our lips have to close for the “m” but at the same time, our tongue gets in the right position for the “a”. During a stutter on “m”, the tongue was forming the expected shape in preparation for the “a”. And yet, the difficulty appears to be the release of the overly constricted lips.
Try an example: Your tongue position controls vowel sounds (a, e, i, o, u).
Pretend you're about to say "Me". Pay attention to the shape of the vocal tract: your lips are closed and the tongue is quite high in the mouth. When you start to say "Me" the tongue doesn't move much - it's already in position right from the M sound!
Do the same with "Man". This time, the tongue will be low.
What were the limitations of the study?
Here, the authors of the study characterised stuttering for one person as they said made-up words. As we know, stuttering is a highly variable and unique experience from day-to-day and between people. Stuttering varies in the context of real language and social experiences. Restricting the utterances to made-up words helps researchers to control the experiment, but means it’s not a realistic context. The characteristics that are described here might not apply to other people in more natural situations. That’s the next job!
Vocal tract MRI is *very cool* and gives us an amazing look at the length of the vocal tract as someone is speaking … BUT … it captures images at quite a slow rate compared to other methods. It's about the same speed as a movie taken on your phone - 33 images per second. This is fast enough to see most speech movements, but very quick movements, for example those performed by the tongue tip as you say “tattle” will look a bit blurry or will be missed. Recently, researchers have increased the image rate up to 100 images per second, so the technology is always improving.
This new study is the first to look at what happens inside the vocal tract during stuttering using MRI. The results from this case-study revealed some patterns of movement that might help to explain why a stutter starts and what happens during a stutter. This is an interesting first step for this line of research! Hopefully there's lots more to come from Yijing's work.
About the Authors
Yijing Lu is a fifth-year PhD student in Linguistics at the University of Southern California.
"My primary research interest is in speech production, especially the control of speech articulators in both typical and atypical speech. Most of my research involves examining articulatory data collected using real-time MRI."
See Yijing's Website
Get in touch with Yijing here: firstname.lastname@example.org
Louis Goldstein is Yijing's PhD supervisor and is a professor at the Department of Linguistics at the University of Southern California.
The team from Oxford: Me, Kate Watkins (Head of Lab), Mark Chiew (MRI physicist) also collaborated with this work and are happy to answer any questions!
This data was collected as part of the INSTEP project run by the Speech and Brain Lab at the University of Oxford. A huge thank you to all of the participants who took part in the INSTEP study, and in this case, special thanks to the participant who contributed to this work!
An interview with the lead author - Yijing Lu.