Perception of voiceless fricatives by Japanese listeners of advanced and intermediate level English proficiency

Hinako Masuda, Takayuki AraiInterspeech 2010

TL;DR

This study investigates how Japanese speakers with varying levels of English proficiency perceive English voiceless fricatives, particularly in noisy conditions. The findings reveal that intermediate learners struggle the most, and certain confusions are unique to Japanese speakers, while others appear universal, highlighting the persistent influence of native language phonology on foreign sound perception.

What Problem Does This Paper Address?

Understanding how non-native speakers perceive sounds from a foreign language is crucial for improving language learning and speech technology. Previous research has shown that a listener's native language (L1) significantly influences their ability to distinguish between foreign language sounds, especially those that don't exist in their L1. For Japanese speakers learning English, voiceless fricatives like /f/ and /θ/ (as in "thin") are particularly challenging because these sounds are absent in Japanese. This paper aims to quantify this difficulty by comparing the perception of these sounds by Japanese listeners at intermediate and advanced English proficiency levels against native English speakers, and to examine how background noise exacerbates these perceptual challenges.

Methods

The researchers conducted a forced-choice identification experiment. Participants listened to English voiceless fricatives embedded in an /a_a/ context. These sounds were presented under three listening conditions: quiet, multi-speaker babble noise, and white noise, all at a signal-to-noise ratio (SNR) of 0 dB. Participants were asked to identify the consonant from a list of 23 options. The study utilized standard audio recording equipment (Marantz PMD 660, SONY ECM-23F5) and analysis software (Praat). The key novelty lies in the controlled comparison across different proficiency levels and noise conditions, specifically focusing on the challenging voiceless fricatives for Japanese speakers.

Datasets

The study did not use pre-existing datasets for training or evaluation. Instead, they created their own audio stimuli by recording a Japanese-English bilingual speaker producing English voiceless fricatives in an /a_a/ context. The noise conditions (multi-speaker babble and white noise) were sourced from the NOISEX-92 database. The participants' English proficiency was assessed using TOEIC® scores as a primary criterion.

Key Insights and Ideas

  • L1 Influence is Persistent: Even advanced Japanese learners show distinct perceptual patterns compared to native English speakers, particularly in noisy environments, underscoring the deep-seated influence of L1 phonology.
  • Noise Amplifies Difficulties: The presence of background noise significantly degrades the perception of English fricatives for all listener groups, but it disproportionately affects intermediate learners.
  • Universal vs. L1-Specific Confusions: Some consonant confusions (e.g., /f/ and /θ/) appear to be universal, occurring across both native and non-native listeners, suggesting inherent acoustic similarities or general perceptual biases. However, other confusions (e.g., /θ/ with /s/) are more specific to Japanese speakers, directly linked to the absence of /θ/ in their native language.
  • Hypercorrection as Evidence: The finding that Japanese listeners sometimes perceive sounds present in their L1 (like /s/) as sounds absent in their L1 (like /θ/) suggests a phenomenon of "hypercorrection," where learners overcompensate for perceived difficulties.

Key Results

  • Intermediate Learners Lag: Intermediate Japanese English learners consistently scored the lowest in all listening conditions compared to advanced Japanese learners and native English speakers.
  • Noise Impact: Correct response rates dropped significantly for all groups in noisy conditions. For example, in white noise (SNR=0 dB), advanced learners achieved 69% correct, while intermediate learners scored 63%.
  • Specific Confusions: Japanese listeners frequently confused /θ/ with /s/ and vice-versa, especially in white noise. Confusions between /f/ and /θ/ were observed across all groups, suggesting a universal tendency.
  • Advanced Learners' Performance: While advanced Japanese learners performed better than intermediate learners, their performance still showed differences from native English speakers, particularly in noisy conditions.

Limitations and Future Work

The authors acknowledge that the study involved only two native English listeners, limiting the generalizability of their findings. They also note that the SNR of 0 dB might have been too severe, potentially masking subtle differences between advanced Japanese learners and native English speakers. Future research should investigate listening conditions with a wider range of SNRs to better clarify the perceptual differences. Further studies are also needed to explore the impact of other factors, such as the age of acquisition and different types of noise, on the perception of foreign language sounds.

Summary updated

BibTeX
@inproceedings{masuda2010perception,
  author = {Hinako Masuda and Takayuki Arai},
  title = {{Perception of voiceless fricatives by Japanese listeners of advanced and intermediate level English proficiency}},
  booktitle = {Interspeech},
  year = {2010},
}
Figure