Using ultrasound tongue imaging to study covert contrasts in second-language learners’ acquisition of English vowels

ABSTRACT

Research attempting to understand the intermediate stages of first-language acquisition and disordered speech has led to the discovery of covert contrast. A covert contrast is a statistically reliable difference between phonemes that is produced by a language learner, but in a way that cannot be heard readily by a listener of the target language. In the present study, we aimed to extend the investigation of covert contrasts to the domain of second-language acquisition. In particular, we used ultrasound tongue imaging to examine whether adult second-language learners of American English produced articulatory distinctions between the target vowels, in addition to producing acoustic differences. We collected speech recordings from 21 speakers (7 Korean speakers, 7 Spanish speakers, 7 English controls) while they produced words illustrating the relevant vowel contrasts in English: /i/-/ɪ/ and /ɛ/-/æ/. Results showed that approximately 36% of our second-language participants implemented various patterns of covert contrast in vowel articulation. Ultrasound measurements revealed that two participants made a covert distinction between two vowels that were perceptually neutralized. For one participant, the anterior part of the tongue was higher for /i/ than /ɪ/, and for the other, it was higher for /ɛ/ than /æ/. Overall, our findings highlight the importance of a learner-oriented approach in studying the acquisition of phonemic contrasts and suggest that ultrasound tongue imaging can be a promising tool to examine the articulatory details of vowel production in second-language learners.

Notes

¹ One of the anonymous reviewers raised the question as to whether eliciting the productions by having the participants read words aloud from the computer screen may have introduced a confounding variable of the sound-grapheme relationship, especially with the Spanish-speaking subgroup of participants. The reviewer noted that because Spanish phonics relates the grapheme “i” with the Spanish high front vowel, and that the spelling of all the target words eliciting /ɪ/ were spelled with the grapheme “i,” there could have been an effect on the productions. As will be discussed in the Stimuli section later, our stimuli consisted of high-frequency words that the L2 participants were familiar with. We would argue that because the participants were familiar with all the target words, it was unlikely that they were providing reading pronunciations, which would have been much more likely if the targets were unknown to the participants.

² Because our stimuli included near-minimal pairs as well, an anonymous reviewer raised the possibility that transcribers could have been biased in favor of a vowel that created a real rather than a nonsense word (e.g., Transcribing chat as [$t\int \text{ae}t$] regardless of the actual pronunciation, as there is no real English word pronounced as [$t\int et$]). Although an interesting possibility, it is not possible to assess whether the transcribers were influenced in any way. All we know at this point is that the transcribers were asked to be faithful to what they heard.

³ The measure anterior tongue height used all coordinate points from the tongue peak to the tip. In addition, we also computed anterior tongue height excluding some end coordinate points (specifically, excluding 5% and 10% of the coordinate points from the tongue tip) and tried out the measures as well because the ends of the tongue contours often showed more variability in a measurement (as well documented in the literature as in Stone 2005). This analysis showed that using any one of these measures did not change the significance of the results.

⁴ Although we did not collect the English language proficiency of the L2 participants, most of them (11 out of 14) were full-time students who met the English language proficiency requirements of the university (e.g., A minimum TOEFL iBT score of 65 required for undergraduate admission and that of 79 required for graduate admission). Of the remaining three, two were ESL students and one was a university staff (lecturer).