Why would aspiration lower the pitch of the following vowel? observations from leng-shui-jiang Chinese

Caicai ZhangInterspeech 2009

TL;DR

This paper investigates why the aspiration of a consonant can lower the fundamental frequency (pitch) of the following vowel, a phenomenon observed in some languages but not others. By analyzing Leng-shui-jiang Chinese, the study finds that aspiration significantly lowers pitch in certain tonal contexts and suggests that reduced vocal fold closure (measured by closing quotient) is a key factor.

What Problem Does This Paper Address?

The relationship between consonant aspiration and the pitch of the subsequent vowel is a long-standing puzzle in phonetics. While some languages show aspiration raising pitch, others exhibit the opposite effect, and some studies even suggest no direct correlation. This inconsistency suggests that the articulatory mechanisms involved in aspiration and their impact on vocal fold vibration vary significantly across languages and even within speakers of the same language. This paper aims to shed light on this debate by examining the Leng-shui-jiang (LSJ) Chinese dialect, specifically focusing on how aspirated versus unaspirated initial consonants affect the fundamental frequency (F0) of vowels across different tonal categories.

Methods

The study employed a two-part experimental approach:

  1. Acoustic Experiment (Experiment 1): Six native LSJ speakers (three male, three female) recorded a set of words containing syllables with both aspirated and unaspirated initial consonants. These recordings were made using the Praat software at a sampling frequency of 44.1 kHz. The fundamental frequency (F0) was measured at 13 sampling points throughout each syllable. Statistical analysis, specifically a one-tailed t-test, was used to determine if the observed differences in F0 between aspirated and unaspirated initial syllables were statistically significant. The analysis focused on the first ten cycles of the vowel and the entire syllable duration.
  2. Electroglottograph (EGG) Experiment (Experiment 2): To explore the physiological underpinnings of the observed F0 variations, a subset of speakers participated in an EGG experiment. This technique measures the electrical conductivity between the vocal folds, providing insights into their contact patterns. In this experiment, F0 and the closing quotient (Qx) were simultaneously recorded using Laryngograph VoiceSuite. Qx is a measure of the proportion of the vocal fold cycle during which the vocal folds are in contact, serving as an indicator of vocal fold adduction (how tightly they are closed).

Datasets

The study utilized a custom-built corpus of LSJ Chinese words.

  • Experiment 1 Dataset: 307 words (158 with plosive initials, 149 with affricate initials) were recorded from six speakers. These words were designed to contrast aspirated and unaspirated initial consonants within specific tonal contexts.
  • Experiment 2 Dataset: 88 words, specifically minimal pairs across three tonal groups, were selected for EGG recording. This dataset involved two speakers (one male, one female).

The paper does not present a publicly available benchmark dataset but rather a newly collected corpus for this specific study.

Key Insights and Ideas

  • Aspiration Lowers Pitch in LSJ Chinese: Contrary to some findings in other languages, aspiration of initial consonants in LSJ Chinese was found to significantly lower the F0 of the following vowel in two out of the three tonal groups examined. This challenges the notion of a universal effect of aspiration on pitch.
  • Tonal Context Matters: The effect of aspiration on F0 was not uniform across all tones. The most pronounced pitch lowering occurred in tone group 3 (3a vs. 3c), with a notable effect also seen in tone group 1 (1a vs. 1c). Tone group 2 (2a vs. 2c) showed less distinct differences, suggesting that the interaction between aspiration and tone is complex.
  • Closing Quotient as a Key Factor: The study found that syllables with aspirated initials exhibited a lower closing quotient (Qx), indicating less vocal fold closure. This is hypothesized to be the primary physiological mechanism responsible for the observed lower F0. A more abducted glottis (less closed vocal folds) naturally leads to a lower pitch.
  • Timing of Articulatory Realization: The authors suggest that the linguistic targets for unaspirated tones (1a, 2a, 3a) are realized earlier in the syllable compared to aspirated tones (1c, 2c, 3c). The aspiration's intrinsic perturbation appears to delay the realization of the target F0, potentially explaining why the pitch difference persists for a significant portion of the syllable.

Key Results

  • Significant F0 Lowering: T-tests revealed that F0 was significantly lower after aspirated obstruents compared to unaspirated ones in tone groups 1 and 3.
  • Qx Correlation: A correlation was observed between F0 and Qx, with lower Qx values (indicating less vocal fold closure) generally associated with lower F0.
  • Qx Differences: Syllables with aspirated initials (1c, 2c, 3c) consistently showed lower Qx values than their unaspirated counterparts (1a, 2a, 3a), particularly in the first half of the syllable.
  • Tone Group 3: The most significant difference in F0 was observed between 3a and 3c, with 3c having a considerably lower pitch.
  • Tone Group 2: The F0 curves for 2a and 2c were nearly identical, except for the very initial points, suggesting a weaker or absent effect of aspiration in this tone group.

Limitations and Future Work

  • Limited Speakers: The study involved only six speakers, and the EGG experiment was conducted with just two. The authors acknowledge that individual speaker variation exists and that more data is needed to confirm their findings across a broader population.
  • "Normal" Qx Threshold: The criterion of 40% Qx as "normal" is acknowledged as an assumption, and further research is needed to establish definitive thresholds for steady-state vocal fold vibration.
  • Subglottal Air Pressure: While the paper discusses subglottal air pressure as a potential factor, it was not directly measured in this study. Future research could incorporate direct measurements of subglottal pressure to further elucidate the interplay between aerodynamic and laryngeal factors.
  • Cross-Linguistic Comparisons: The authors highlight the need for more research to understand how Qx functions in the realization of pitch variations in other languages, particularly those where aspiration has been linked to higher F0. The study also suggests that further investigation is needed to understand why the correlation coefficient (r) between F0 and Qx varies across different tones and speakers.

Summary updated

BibTeX
@inproceedings{zhang2009why,
  author = {Caicai Zhang},
  title = {{Why would aspiration lower the pitch of the following vowel? observations from leng-shui-jiang Chinese}},
  booktitle = {Interspeech},
  year = {2009},
}
Figure