TL;DR
This study investigates how native Mandarin speakers and Korean L2 learners adjust their speech production (intensity, duration, and pitch) when speaking in noisy environments. The findings reveal that L2 learners exhibit a more pronounced Lombard effect, suggesting they rely more heavily on auditory feedback and are more negatively impacted by noise than native speakers.
What Problem Does This Paper Address?
Producing clear speech in noisy environments is a challenge for everyone, but it's particularly difficult for second language (L2) learners. This phenomenon, known as the Lombard effect, causes speakers to involuntarily increase their speech intensity, alter their pitch, and lengthen their speech sounds to compensate for the noise. While the Lombard effect has been studied in native speakers and L2 learners of English, research on L2 learners of tonal languages like Mandarin has been limited. This paper aims to fill this gap by examining how Korean L2 learners, whose native language is non-tonal, produce Mandarin tones in noisy conditions compared to native Mandarin speakers. Understanding these differences is crucial for developing better speech technologies and improving L2 learning strategies.
Methods
The researchers employed a controlled experimental setup to record participants' speech. They used two types of noise: white noise and babble noise, both presented at a significant intensity of 80 dB SPL, along with a quiet condition for baseline comparison. Acoustic features, specifically mean intensity, duration, and F0 (fundamental frequency) contours, were meticulously extracted from the recorded speech. To analyze these complex acoustic data, the researchers utilized advanced statistical modeling techniques. Linear mixed-effects models (LMMs) were used for intensity and duration, while generalized additive mixed-effects models (GAMMs) were employed for F0 contours. This approach allowed them to account for individual speaker variations and the specific characteristics of each tone. Bonferroni post-hoc tests were used for pairwise comparisons to pinpoint significant differences between conditions and groups.
Datasets
The study did not introduce a new dataset. Instead, it utilized recordings from 52 participants (26 native Mandarin Chinese speakers and 26 Korean L2 learners of Chinese) who produced 28 monosyllabic Mandarin words under three different acoustic conditions (quiet, white noise, babble noise).
Key Insights and Ideas
- L2 Learners' Greater Lombard Effect: Korean L2 learners showed significantly greater adjustments in intensity and duration across all tones in noisy conditions compared to native speakers. This suggests a stronger reliance on auditory feedback for speech monitoring and compensation in L2 learners.
- Differential Impact of Noise Types: White noise had a more significant impact on speech production than babble noise for both groups. This is attributed to white noise's broader frequency range, which can lead to a stronger masking effect.
- Tone 3 as a Challenge: The most notable group difference in F0 contour adjustments was observed for Tone 3. While Korean L2 learners produced Tone 3 similarly to native speakers in quiet, noise significantly disrupted their production, leading to less precise F0 contours. This aligns with previous research suggesting Tone 3 is particularly challenging for non-tonal language speakers.
- Feedback Control Theory in Action: The findings strongly support feedback control theory in L2 speech production. L2 learners, being more prone to speech errors, rely more on auditory feedback. Noise disrupts this feedback loop, leading to greater compensatory adjustments (intensity, duration) but also reduced accuracy in pitch control.
Key Results
- Intensity: Korean L2 learners significantly increased their mean intensity in both noise conditions compared to quiet. Native Chinese speakers showed no significant intensity adjustments.
- Duration: Korean L2 learners significantly lengthened the duration of all four tones in noise. Native Chinese speakers only significantly lengthened durations for Tone 3 and Tone 4 under white noise.
- F0 Contours: Both groups adjusted F0 contours for all tones in noise. However, significant differences emerged for Tone 3, where Korean L2 learners' production was more negatively affected by noise compared to native speakers.
Limitations and Future Work
The authors acknowledge that the study focused on higher-level Korean L2 learners. They suggest that future research should investigate lexical tone production in lower-level L2 learners under noise to understand how proficiency impacts these adjustments. Additionally, the study hints at a potential compensatory relationship among acoustic features (e.g., increased intensity and duration compensating for F0 difficulties), which warrants further exploration.
Summary updated