May Pik Yu Chan will be defending their dissertation proposal, tentatively titled 'Pitch-dependent variation in vowel production and perception: Revisiting the source-filter theory' on Monday, April 22 at 3:30pm. The defense will take place in person in the Linguistics department library, and over Zoom. All are welcome to attend!

The proposal document can be found here, and the abstract is included below.


Title: Pitch-dependent variation in vowel production and perception: Revisiting the source-filter theory

Supervisor: Jianjing Kuang

Proposal committee: Eugene Buckley (chair), Mark Liberman, Meredith Tamminga

Time: April 22 (Monday), 3:30pm-5:00pm

Place: Linguistics department library, 3401C Walnut St, Suite 300 and Zoom


The source-filter theory which underpins much of modern speech science, often assumes the source and filter to be independent. This theory predicts that source-related attributes (namely, pitch) cannot be influenced by the filter-related attributes (e.g. vowels). This long-held assumption has resulted in successful descriptions, syntheses and processing of speech and continues to be upheld in most current studies of speech, despite Klatt and Klatt’s (1990) warning that such assumption is better made for low-pitched male speech than for high-pitched females’ (and likely children’s) speech, “due perhaps to tracheal coupling and source/tract interactions” (Klatt and Klatt 1990, p. 820). One challenge that the theory might allude to concerns cases of vowel production and perception beyond the pitch range of normal speech, where the fundamental frequency is higher than the lowest formants. As the lowest formants are important resonance structures for vowel distinction, one might predict that high pitches may lead to unexpected consequences for vowel contrast. Nonetheless, high-pitched speech is not entirely uncommon. For example, singing as a special occasion of speech production, is typically much higher in pitch than the speaking range. Social or communicative contexts may also trigger high-pitched speech, including shouting, stage acting, and public speaking. Furthermore, young children also typically speak at a very high pitch range. These situations indeed have presented themselves as serious problems for current speech technologies. Do human speakers also face the same challenges? If so, what do human voice users do to resolve or remedy the challenges and issues of interaction between source signals and filter structures?

The overall goal of this dissertation is therefore to better understand the ways in which the voice source and filter structures may interact in human vocal production and speech perception. To achieve this goal, I present acoustic and articulatory analyses of the singing voice as a window to explore the limits of the human vocal apparatus, focusing on how vowel production may be affected by different pitch targets. I also seek to include cases of high-pitched speech to explore how it may interact with vowel production and perception, including speech at different levels of vocal effort, and children’s speech, if possible. Lastly, my work also explores how pitch excitation structures may interact with the filter function in human perception. Together, these experiments will better our understanding of core phonetic theories of vowel production and perception, as well as source and filter interactions.