Thesis Christ: Degraded Speech Perception with Kalin Mattern

Frog. Fog. Thought. Bought. Lost in the medium of print are the crucial differences in pitch and phonics that distinguish these seemingly simple words from one another in everyday speech. Now distort those same words, and the differences become even more slight—in fact, they may not even register as words. Mattern’s thesis deals with perceptions of degraded speech, which is described as “speech that has had a bunch of the qualities that make it sound normal removed.” She is interested in how people learn to perceive this type of speech and generalize their understanding to new words based on their existing knowledge. Adding another dimension, Mattern’s research also involves the impact of musical and foreign language experience on the ability to learn to perceive degraded speech.

Mattern’s own personal experience drove her to study the connections between music and speech perception. “I have always been interested in music. I’ve been playing instruments since I was in fourth grade,” she said, citing her history with music as a key factor behind her interest in auditory perception. Mattern first explored these topics at Reed in PSY 381: Sensation and Perception with Psychology Professor Michael Pitts, who is now her thesis advisor, where she was able to present on sine wave speech, the form of degraded speech used in her thesis.

Sine wave speech removes broadband frequencies, such as pitch distinctions, from speech, while retaining the formants, or the basic resonant frequencies of the vocal tract which determine the quality of vowels. Perception of sine wave speech is closely connected to the ability to distinguish speech from background noise. 

To further investigate these concepts, Mattern conducted a study in which participants were tasked with recognizing and replicating words that have been distorted through sine wave speech. Participants were asked to listen to “speech tokens,” monosyllabic fragments of speech that could be real or invented words, then rate how clear the speech was on a scale from one to four and finally repeat the word they thought they heard aloud into a microphone. These two responses give a subjective and an objective assessment of how well a participant perceives the speech tokens. Throughout the study, participants listen to five rounds of speech tokens, beginning with solely degraded speech, then degraded speech mixed with the original versions of each word, after which previously unheard words are introduced, and finally concluding with exclusively degraded speech tokens including new words.

At this point, Mattern has concluded assessing participants in the study. She worked with linguistics students, who are transcribing participants’ responses into the International Phonetic Alphabet to judge their accuracy compared to the original words. Mattern is currently conducting data analysis and writing her results and discussion. She reflected, “I have some ideas about the ways things might go, and exploring different limitations of the thesis and possible future directions to address those.”

Although Mattern has not yet finalized any definitive conclusions about her data, she has noticed several interesting trends that suggest avenues for further study. “I have been seeing that people do get better at both the clarity ratings and verbal responses over the course of the experiment, but I don’t have much information on the nuances of when those changes are starting to be seen,” she reported.

Several trends are already apparent in Mattern’s data on the relationship between musical or language experience and speech perception. While people who reported musical experience generally had greater accuracy in perceiving the speech tokens, no such correlation was found for foreign language experience. “When it came to the objective measure of verbal response accuracy, people who have higher scores with musical experience, so people who just have more musical experience, overall did better on responding to the words and getting them right, whereas that effect was not seen for [foreign language] experience,” Mattern explained. 

Mattern hopes her thesis will be expanded on in the future, remarking, “I really hope that someone thesising with Michael in the future actually takes this data set and does some more crazy statistics with it.” At the same time, she made clear that she would not be the one to do that research herself. “I’ve enjoyed doing thesis but I know that a career in research is not for me,” she said. “Thesis for me has been more about doing a project from beginning to end and really being confident in my ability to ask a question and answer it pretty much by myself.”

When it comes to the future, Mattern has multiple paths open to her. She currently works with Reed Science Outreach teaching neuroscience topics to elementary schoolers, leading her to consider taking these efforts further with a career in science education. Mattern is also considering a future in social work, which she sees as a way to see the “tangible results of helping people.”

One of the key learning experiences for Mattern over the course of her thesis was gaining more confidence in her ability to solve her own problems. She reflected, “Other people have been very helpful in giving me overviews of stuff that I didn’t know about before… but I have the knowledge to apply it to my own project. It’s been really nice to become an expert in this one thing and really be like, ‘I’m the one who knows about this.’” She advises anyone thesising in the future to have “confidence in your ability to solve your own problems, but don’t be afraid to ask for help.”

Are you a senior who isn’t completely sick and tired of talking about their thesis? Do you want to spill the contents of your brain to a more-or-less captive audience who will take whatever you say about your niche subject of interest as gospel? Email tanforanv@reed.edu to be the next Thesis Christ! Time is running out…