From left, Márton Sóskuthy, Molly Babel, and Suyuan Liu, all enjoying a congratulatory poster.
Former doctoral candidate and current doctorate holder Suyuan Liu successfully defended her dissertation on April 22. The dissertation is titled Where Ideology Meets the Signal: Language Standardness, Voice Similarity, and Intelligibility in Bilingual Speech Processing. It was supervised by Molly Babel. The other doctoral committee members were Márton Sóskuthy, Alexis Black, and Jian Zhu. The abstract is below.
Congratulations, Dr. Liu!
Where Ideology Meets the Signal: Language Standardness, Voice Similarity, and Intelligibility in Bilingual Speech Processing
Languages bridge the processing of signals and the formation of thoughts. Conversely, how do socially grounded evaluations and signal-based structures jointly shape speech processing within the same group of listeners? This dissertation investigates how perceived language standardness and listener-talker voice similarity interact with intelligibility in Mandarin-English bilinguals through the creation of the Mandarin–English Language Interview (MELI) corpus. The MELI corpus is a 30-hour open-access bilingual speech corpus consisting of interviews and read sentences from 51 talkers, with all recordings fully transcribed and force-aligned. Participants in MELI then completed a series of tasks that examine (1) their definitions and evaluations of language standardness, (2) their perceptual and acoustic representations of voices, and (3) how standardness and listener–talker voice similarity affect intelligibility.
For MELI participants, “standardness” is not a fixed linguistic property but a socially organized construct whose meaning differs across languages. Through reflexive thematic analysis and perceptual rating tasks, the results show that English standardness is frequently articulated in relation to intelligibility and communicative clarity, whereas Mandarin standardness is more strongly tied to regional and social evaluative dimensions. Voice qualities, however, can be evaluated through both perceptual sorting and acoustic modeling. These approaches reveal that the relative freedom of perceptual sorting tasks leads to less consistent estimates of voice similarity than those derived from acoustic analysis. Ultimately, both socially grounded standardness and signal-based voice similarity facilitate intelligibility: listeners are more accurate at transcribing speech that is perceived as “more standard” and spoken in a voice similar to their own. However, their effects differ in magnitude across languages: similarity operates consistently in both Mandarin and English, while standardness exerts language-specific influence aligned with participants’ ideological articulations.
Together, these findings demonstrate that speech processing is shaped by the interaction of signal-based structure and ideology-based orientation. Signal-based voice similarity reflects structural properties of perceptual organization, while ideology-based standardness operates through socially grounded evaluation within language-specific contexts. Speech understanding emerges not solely from acoustic input, but through the interplay of form, ideology, and interpretation.
