The Production of Chinese Affricates /ts/ and /tsh/ by Native Urdu Speakers

Dan Du, Jinsong ZhangInterspeech 2019

TL;DR

This paper investigates how native Urdu speakers produce Chinese affricates /ts/ and /tsh/, sounds absent in Urdu. By analyzing speech data from learners with varying proficiency, the study reveals distinct acquisition patterns influenced by the presence or absence of similar sounds in their native language, offering insights into second language phonological learning.

What Problem Does This Paper Address?

Learning new sounds in a second language is a well-documented challenge, often influenced by a learner's native language (L1) phonology. While many studies have explored L2 sound acquisition, there's a notable gap in research specifically examining native Urdu speakers learning Chinese affricates (/ts/ and /tsh/). These sounds are particularly interesting because they don't have direct counterparts in Urdu, and their production can be complex for L2 learners. Understanding how Urdu speakers acquire these sounds can shed light on universal and language-specific factors in L2 phonological development, contributing to more effective language teaching strategies.

Methods

The researchers utilized a corpus-based approach, analyzing speech data from the BLCU-SAIT speech corpus. This corpus contains recordings of native Urdu speakers reading Chinese monosyllables. Participants were divided into two groups based on their Chinese proficiency, determined by an automated mispronunciation detection system. The study involved manual transcription of the speech samples by experienced phonetics students, followed by an error analysis to identify the types and rates of misproduced affricates. The transcription scheme employed Chinese Pinyin and the PET labeling scheme to capture articulation tendencies.

Datasets

The study used data from the BLCU-SAIT speech corpus. This corpus includes recordings of native Urdu speakers reading Chinese monosyllables, disyllables, sentences, and a passage. For this specific study, the researchers focused on monosyllabic words containing the Chinese affricates /ts/ and /tsh/ combined with various vowels and tones. The participants were ten native Urdu speakers (seven males, three females) aged 20-25, with no reported language, hearing, or speaking impairments.

Key Insights and Ideas

  • L1 Influence is Crucial: The study strongly supports the idea that a learner's native language phonology significantly impacts their ability to acquire new sounds in a second language.
  • Differential Acquisition: Despite both /ts/ and /tsh/ being absent in Urdu, they are acquired differently. This suggests that even subtle differences in how sounds are not present in the L1 can lead to distinct learning trajectories.
  • The "z" Interference: The Chinese character "z" (often pronounced /ts/ or /dz/) presents a unique challenge for Urdu speakers due to a deeply ingrained association with the Urdu sound /z/. This interference can lead to persistent errors, even in advanced learners, illustrating the concept of "fossilization."
  • Error Reduction and Convergence: As learners progress, both the rate and variety of errors decrease. The remaining errors tend to be closer to the target pronunciation, indicating a developmental process where learners refine their production.
  • Error Type Dynamics: The types of errors made also evolve. While initial stages involve a wider range of errors, advanced learners tend to make fewer, more specific errors, sometimes even introducing new error types not seen in earlier stages.

Key Results

  • Error Rate Reduction: The error rate for both /ts/ and /tsh/ significantly decreased from the low-proficiency group to the high-proficiency group. Specifically, the error rate for /tsh/ dropped from over 75% to under 15%, while the error rate for /ts/ also showed a substantial reduction.
  • Differential Error Rate Trends: While both affricates showed improvement, the error rate for /tsh/ demonstrated a more stable and consistent decrease. The error rate for /ts/ was higher than /tsh/ in the high-level group in 60% of cases, which was the inverse of the low-level group.
  • Error Type Reduction: The number of distinct error types for both /ts/ and /tsh/ also reduced significantly. For /tsh/, errors decreased from eight types in the low-level group to two in the high-level group. For /ts/, errors reduced from five types to two.
  • Fossilization of /ts/: The study observed instances of "fossilization" for the /ts/ affricate, where a learner in the high-level group consistently produced the sound incorrectly, suggesting a persistent interlanguage feature.

Limitations and Future Work

  • Limited Sample Size: The study involved a relatively small sample size of ten participants.
  • Focus on Monosyllables: The analysis was primarily based on monosyllabic words, which may not fully represent the complexities of producing affricates in connected speech.
  • "z" Interference Nuance: While the interference of the Chinese character "z" was noted, further research could explore the precise phonetic mechanisms and perceptual factors involved.

The authors suggest that this study is a first step towards a comprehensive understanding of L2 Chinese acquisition by native Urdu speakers. Future work could expand on these findings by investigating connected speech, a larger participant pool, and potentially exploring the perceptual assimilation of these affricates by Urdu speakers.

Summary updated

BibTeX
@inproceedings{du2019production,
  author = {Dan Du and Jinsong Zhang},
  title = {{The Production of Chinese Affricates /ts/ and /tsh/ by Native Urdu Speakers}},
  booktitle = {Interspeech},
  year = {2019},
}
Figure