Articulating Non-Native Vowel Contrasts

Abstract

The goal of this dissertation is to better understand the targets of vowels in speech production. Three experiments investigate vowel production in non-native speech, probing how new vowel categories, and thus targets, are formed. Three research questions are addressed:

1) Do vowels have articulatory targets (as predicted by Browman & Goldstein (1989; 1990)), as opposed to acoustic targets (as predicted by Liljencrants & Lindblom (1972), Flemming (1996), and Stevens & Keyser (2010))?

2) Do speakers re-use the articulatory gestures (as predicted by Best & Tyler (2007)) or acoustic targets (as predicted by Flege (1987; 2005)) from their native language to produce non-native phones? How does vowel quality, and similarity to the native language sounds, influence speakers’ production strategies?

3) Do naïve speakers, who are hearing certain vowel contrasts for the first time, use different strategies than language learners, who have had more exposure to the language?

The three experiments in this dissertation explore these questions by collecting acoustic, video and sonographic data on non-native vowel production, and are designed to look at situations where articulatory and acoustic theories make different predictions about vowel production.

Experiment 1 examines how English speakers produce Akan vowels to which they have had no prior exposure. Akan vowels are acoustically similar to English vowels, and thus are predicted to be assimilated by English listeners to English categories, but are produced with different vocal tract configurations, using mostly the tongue root rather than tongue body (Tiede, 1996; Kirkham & Nance, 2017). Experiment 2 examines how Spanish speakers produce French front round vowels to which they have had no prior exposure. Because front round vowels are acoustically dissimilar from any Spanish vowels, they are predicted to fail to assimilate to any Spanish vowel category. Experiment 3 examines how native English speakers learning French (as opposed to naïve speakers) produce the vowels of French. For each experiment, the question is asked, do speakers produce the novel phones with the acoustic targets of the native language, the articulatory gestures of the native language, both, or neither?

The results show a complex pattern of L1 transfer, with much individual variation and dependence on vowel quality. For both French and Akan, learners and naïve speakers were more likely to transfer both tongue position and acoustic target for the high vowels than for the mid vowels, with /i/ showing particular stability across speakers and languages, suggesting a special status for the “corner” vowels in both acoustics and articulation. For the front round vowels of French, both learners and naïve speakers transferred lip rounding gestures from the L1, but not tongue body or lip un-rounding gestures, suggesting a special status for a “round” feature or gesture. Across all three experiments, participants did not re-use native language categories for dissimilar phones, suggesting that different strategies are used for categorized and uncategorized “new” vowels. Overall, these results demonstrate that models of L2 category formation must reference both the articulatory gesture and the acoustic-phonetic category.