Date Thesis Awarded

5-2021

Access Type

Honors Thesis -- Open Access

Degree Name

Bachelors of Science (BS)

Department

Linguistics

Advisor

Anya Hogoboom

Committee Members

Kaitlyn Harrigan

Daniel Parker

Zach Conrad

Abstract

The speech of patients with schizophrenia has been characterized as being aprosodic, or lacking pitch variation. Recent research on linguistic aspects of schizophrenia has looked at the vowel space to determine if there is some correlation between acoustic aspects of speech and patient status (Compton et al. 2018). Additional research by Hogoboom et al. (submitted) noted that measurements of Euclidean distance (ED), which is the average distance from the center of the vowel space to all vowels produced, and vowel density, which is the proportion of vowels clustered together in the center of the vowel space, were significantly correlated for patients with schizophrenia, but not for controls; this correlation was primarily due to a subset of 13 patients. In addition, they found that ED independently was a weak predictor of patient status, but that both density and ED, when used together, were predictors of patient status. This previous study utilized Prosogram (Mertens 2014), a tool that relies on acoustics to sift through the sound files and identify the vowels, which showed unstable reliability in detecting vowels.

Therefore, this research aims to reassess the relationship between the vowel space and patient status by gathering more reliable measurements of the vowels from Hogoboom’s dataset by using the forced aligner FAVE (Rosenfelder et al. 2014). We seek to determine if there is a stronger correlation between vowel space usage and patient status than previously found—one that was previously masked by incomplete vowel measurements. Our current research finds that ED is a strong predictor of patient status (p

Share

COinS