Assessing vowel centralization in dysarthria: A comparison of methods

Abstract

Previous literature has consistently reported correlations between acoustic vowel centralization and perceptual measurements of dysarthria. However, the strength of these relationships is highly variable, and many of the techniques used to measure vowel centralization have not been directly compared. This study evaluates methods of assessing vowel centralization and listeners’ perceptions of dysarthria—with the aim of strengthening the relationship between these variables. Sixty-one speakers of New Zealand English (NZE; 17 healthy older individuals and 44 speakers diagnosed with dysarthria) read a standard passage. Metrics of vowel centralization were calculated using first and second formants of the [ɐː], [iː] and [oː] NZE point vowels. The results demonstrate that both the use of a flexible formant extraction point, and changes to the frequency unit in which formants are measured, can strengthen the relationship between acoustic and perceptual measures. Furthermore, applying these formant values to different metrics of vowel centralization, and changing the instructions listeners are given to rate dysarthria, can also reduce levels of unexplained variation in the relationship. In combination, these changes accounted for 18–26% more variance between vowel centralization measurements and listener perceptions of dysarthria in both male and female speakers.