TL;DR
This study investigates how the amount of English language experience affects Mandarin speakers' ability to distinguish between English tense and lax high vowels. The findings reveal that while inexperienced speakers rely heavily on duration, experienced speakers shift towards using spectral cues, mirroring native English speakers' perception strategies. This highlights the crucial role of L2 exposure in shaping auditory perception for non-native sounds.
What Problem Does This Paper Address?
Learning a new language involves not only producing new sounds but also perceiving them accurately. This paper delves into how listeners whose native language (L1) differs from their second language (L2) perceive vowel contrasts that don't exist in their L1. Specifically, it examines the English tense-lax high vowel distinction (/i/ vs. /ɪ/ and /u/ vs. /ʊ/), which is characterized by differences in both spectral quality (formant frequencies) and duration. Previous research suggests that listeners often rely on their L1's phonemic contrasts, leading to different perceptual strategies. For instance, speakers of languages with phonemic vowel duration (like Japanese or Finnish) tend to prioritize duration cues for English tense-lax vowels, while native English speakers primarily use spectral differences. The question this paper addresses is how L1-Mandarin speakers, whose language has tonal duration but not phonemic vowel duration, adapt their perception of these English vowels as they gain more English experience.
Methods
The researchers synthesized speech stimuli to control for specific acoustic features. They created continua of English "heed-hid" (/i/-/ɪ/) and "who'd-hood" (/u/-/ʊ/) vowels. These continua varied in two key dimensions:
- Spectrum: Five steps of formant frequency variations were generated to simulate the spectral differences between tense and lax vowels.
- Duration: Seven steps of vowel duration were created, ranging from shorter to longer.
These synthesized stimuli were then presented to three groups of listeners:
- L1-English speakers: Native English speakers serving as a baseline.
- Experienced L1-Mandarin speakers: Mandarin speakers with significant English language exposure (e.g., undergraduate/graduate students in English-related fields).
- Inexperienced L1-Mandarin speakers: Mandarin speakers with limited English exposure (e.g., students not in English departments).
Listeners performed identification tasks, where they had to categorize the synthesized vowels. The researchers analyzed the proportion of "tense" vowel responses (e.g., "heed" or "who'd") as a function of both spectral and durational variations for each group. Statistical analyses (ANOVAs and post-hoc tests) were used to determine the significance of group differences and the influence of spectral and durational cues.
Datasets
The study did not use pre-existing datasets for training or evaluation. Instead, it generated its own set of synthesized stimuli. These stimuli were based on natural speech tokens of English high vowels and then manipulated using a formant synthesis program (KLSYN) and analyzed with Praat for intonation contours.
Key Insights and Ideas
- L2 Experience Shapes Perception: The most striking finding is how the amount of L2 (English) experience fundamentally alters how Mandarin speakers perceive English vowels. This isn't just about learning new sounds; it's about re-calibrating auditory processing.
- From Duration to Spectrum: Inexperienced Mandarin speakers leaned heavily on vowel duration to distinguish between English tense and lax vowels. This suggests they might be transferring their L1's sensitivity to duration (e.g., in tones) to this new contrast. However, as Mandarin speakers gain English experience, they increasingly rely on spectral cues, moving closer to the native English perception strategy.
- Mandarin's Unique Position: Unlike languages with phonemic vowel duration, Mandarin has tonal duration. This paper explores how this L1 feature influences L2 perception, suggesting that while duration is initially important, spectral cues become dominant with sufficient L2 exposure.
- "Who'd-Hood" vs. "Heed-Hid": The study found some differences in how spectral cues affected the perception of the two vowel pairs. While both groups showed some spectral influence, the inexperienced group's perception of "who'd-hood" was less affected by spectral variation compared to the "heed-hid" contrast.
Key Results
- Heed-Hid:
- L1-English and experienced L1-Mandarin speakers showed a strong reliance on spectral cues, with higher proportions of "heed" responses at later spectral steps.
- Inexperienced L1-Mandarin speakers' responses were less affected by spectral variation, indicating a weaker reliance on this cue.
- For inexperienced speakers, duration played a more significant role, with longer durations leading to more "heed" responses.
- Who'd-Hood:
- L1-English and experienced L1-Mandarin speakers relied more on spectral cues, showing a clear shift in "who'd" responses as spectral steps increased.
- Inexperienced L1-Mandarin speakers were less sensitive to spectral variations for this contrast.
- For inexperienced speakers, duration had a significant impact, with longer durations leading to more "who'd" responses.
- Interestingly, for the inexperienced group, duration was the primary cue for the "who'd-hood" contrast, with significant differences observed between the shortest and longest durations.
Limitations and Future Work
- Synthesized Stimuli: The authors acknowledge that the use of synthesized stimuli, while allowing for precise control, might limit the direct generalization of findings to natural speech. Future research could explore perception with natural speech tokens.
- Why Durational Cues? The paper raises a question for future work: why do both experienced and inexperienced L1-Mandarin speakers seem to use durational cues more than L1-English listeners? While L1 transfer is suggested, the exclusive reliance on duration by inexperienced speakers might be due to other, as yet unidentified, factors.
- Generalizability: The study focused on L1-Mandarin speakers and specific vowel contrasts. Further research could investigate other L1 backgrounds and different L2 sound contrasts to build a more comprehensive understanding of cross-language perception.
Summary updated