How consonants, dialect and speech rate affect vowel devoicing?

Masako Fujimoto, Seiya Funatsu, Ichiro FujimotoInterspeech 2012

TL;DR

This paper investigates the acoustic and physiological mechanisms behind vowel devoicing in Japanese, revealing that glottal opening patterns are key. They found that consonant types, regional dialects, and speech rate all influence how vowels are devoiced, with distinct patterns observed between Tokyo and Osaka speakers.

What Problem Does This Paper Address?

Vowel devoicing is a common phenomenon in Japanese, where certain vowels like /i/ and /u/ can become inaudible when surrounded by voiceless consonants or pauses. While this is generally understood, the precise factors influencing when and how it happens, and why there's variation between speakers and dialects, remain unclear. This paper aims to unravel these complexities by examining the glottal opening patterns associated with devoicing, considering the influence of surrounding consonants, regional dialects (Tokyo vs. Osaka), and speech rate.

Methods

The researchers employed a combination of photoelectric glottography (PGG) and high-speed video recordings to capture laryngeal movements during speech. PGG measures the degree of glottal opening, while high-speed video provides detailed visual information. They analyzed the glottal opening patterns, categorizing them into "single-phase" and "double-phase" openings. The study also involved acoustic analysis of speech waves and PGG signals, time-aligned to correlate vocal fold vibration with glottal gestures.

Datasets

The study utilized custom-designed word lists containing specific phonetic environments designed to elicit vowel devoicing. These lists included:

  • Typical consonantal environments: e.g., /kike/, /kise/, /sike/ (Type A: St-St, Fr-St, St-Fr without /h/)
  • Atypical consonantal environments: e.g., /sise/, /hihe/, /kihe/ (Type B: Fr-Fr without /h/, Type C: C2 is /h/)
  • Consecutive devoicing environments: e.g., /kikiki/
  • Phrase final position: e.g., the word "masu"

The data was collected from five Tokyo dialect speakers and three Osaka dialect speakers.

Key Insights and Ideas

  • Two Glottal Opening Patterns: Devoicing is consistently linked to either a single-phase or a double-phase glottal opening.
  • Consonant Type Matters: Fricatives have a dual role: they facilitate devoicing before a vowel but suppress it after.
  • Dialectal Differences: Osaka speakers, where devoicing is less frequent, tend to exhibit double-phase openings even in typical consonantal environments, suggesting a different underlying gestural strategy compared to Tokyo speakers.
  • Speech Rate Accelerates Devoicing: Faster speech leads to a merging of glottal gestures, increasing the likelihood of vowel devoicing.
  • Gestural Reorganization vs. Overlap: Single-phase openings in typical environments for Tokyo speakers are attributed to gestural reorganization, while double-phase openings in atypical environments or for Osaka speakers are linked to gestural overlap.
  • Phrase-Final Devoicing Variation: Devoicing at the end of phrases shows significant interpersonal variation, particularly between dialects.

Key Results

  • Typical Consonantal Environments (Tokyo): For Tokyo speakers, typical environments (e.g., /kike/) consistently showed a single, smooth glottal opening pattern, correlating with complete vowel devoicing. This suggests a planned gestural reorganization.
  • Atypical Consonantal Environments (Tokyo): In atypical environments (e.g., /sise/, /hihe/), Tokyo speakers sometimes exhibited double-phase openings, leading to partial or full devoicing, attributed to glottal gesture overlap.
  • Osaka Dialect: Osaka speakers showed double-phase openings more frequently, even in typical environments, and devoicing was generally less pronounced.
  • Speech Rate: In faster speech, glottal openings for consecutive consonants merged, facilitating devoicing.
  • Phrase Final Position: Tokyo speakers showed a higher rate of devoicing for phrase-final /u/ (e.g., in "masu") compared to Osaka speakers, with variations in the degree of devoicing (full devoicing, partial devoicing, or full voicing).

Limitations and Future Work

  • Generalizability: The authors note that further examination is necessary to generalize these findings, particularly regarding the greater interpersonal variation observed in phrase-final devoicing.
  • Unclear Cases: The paper acknowledges that the reasons for some observed glottal opening patterns in specific instances (e.g., certain /kise/ and /sike/ tokens for Osaka speaker O1) remain unclear and could be due to gestural reorganization or overlap.
  • Further Investigation: The authors suggest that a deeper understanding of the interplay between gestural reorganization and overlap, especially across different dialects and speech rates, is needed.

Summary updated

BibTeX
@inproceedings{fujimoto2012how,
  author = {Masako Fujimoto and Seiya Funatsu and Ichiro Fujimoto},
  title = {{How consonants, dialect and speech rate affect vowel devoicing?}},
  booktitle = {Interspeech},
  year = {2012},
}
Figure