TL;DR
This paper investigates how the phrase-final accent in Kammu, a language with both tonal and non-tonal dialects, conveys multiple prosodic signals like focus and engagement. By analyzing pitch contours, they reveal how lexical tone influences intonation, leading to restricted pitch ranges in tonal dialects and interesting interactions between tone, focus, and speaker engagement.
What Problem Does This Paper Address?
Prosodic features at the end of phrases are known to carry a rich set of information, including marking phrase boundaries, conveying focus, and even signaling speaker engagement. However, disentangling the contributions of different prosodic elements, especially when lexical tone is involved, is challenging. This paper addresses the question of how phrase-final accents in Kammu, a language with distinct tonal and non-tonal dialects, manage to simultaneously encode focus, phrase finality, utterance finality, and speaker engagement. The existence of dialects with and without lexical tones makes Kammu an ideal candidate for studying the interplay between lexical tone and intonation.
Methods
The researchers collected speech data from both tonal and non-tonal Kammu dialects. They used a combination of elicited speech (reading sentences designed to control for lexical tone and elicit focus) and spontaneous speech (describing pictures and telling stories) to capture natural prosodic variations.
The core of their analysis involved acoustic measurements of fundamental frequency (F0). They used the WaveSurfer program to display waveforms, spectrograms, and F0 contours. F0 values were manually annotated for maximum and minimum points on syllables, particularly focusing on pre-penultimate, penultimate, and ultimate syllables. To allow for cross-speaker comparisons, F0 values were converted to semitones and normalized. Statistical analyses, specifically single-factor ANOVAs, were employed to determine the significance of observed differences in F0.
Datasets
The study utilized speech data collected from 13 speakers of non-tonal Kammu dialects and 10 speakers of tonal Kammu dialects. The data comprised approximately 10 hours of recordings from Laos and 5 hours from Thailand, including both elicited (47 read sentences) and spontaneous speech. The elicited sentences were designed to control for lexical tone and elicit focus in different positions.
Key Insights and Ideas
- Lexical Tone Constrains Intonation: The presence of lexical tones in the tonal dialect significantly restricts the overall pitch range used for phrase-final accents compared to the non-tonal dialect. This suggests that speakers in tonal dialects need to maintain tighter control over absolute pitch for lexical distinctions, limiting their freedom to use wide pitch excursions for other prosodic functions.
- Focus and Tone Interaction: In the non-tonal dialect, focus is clearly marked by higher F0 excursions on focused words. In the tonal dialect, this effect is less pronounced, especially when the focused word has a low lexical tone. However, when a focused word also carries a high lexical tone (as in sentence 4), the speakers can leverage this to achieve a higher F0 maximum, demonstrating an interplay between lexical tone and focus marking.
- Speaker Engagement Overrides Tone: In cases of semantic markedness or unusual content (like the word "badger" in a list of common farm animals), speaker engagement can lead to increased F0 excursions. This effect can even override the typical constraints imposed by lexical tone, causing low-toned words to be realized with higher F0s than high-toned words in similar contexts.
- Truncation Phenomenon: Both dialects exhibit a truncation phenomenon where F0 movement is compressed when the final syllable ends in a voiceless stop, suggesting that the absolute F0 maxima might be more critical than minima for marking phrase endings.
Key Results
- Pitch Range Difference: The non-tonal dialect showed a significantly wider F0 range (6-9 St) in phrase-final accents compared to the tonal dialect (3-4 St).
- Focus Marking in Non-tonal Dialect: In the non-tonal dialect, focal accent on final color adjectives (sentences 2 & 4) resulted in significantly higher F0 maxima (e.g., 3.13 St for sentence 2) compared to non-focal nouns (e.g., 1.73 St for sentence 1).
- Focus Marking in Tonal Dialect: In the tonal dialect, the effect of focal accent was less significant. While sentence 4 showed a higher F0 maximum (1.54 St) than sentence 3 (0.76 St), the difference was not statistically significant for all pairs.
- Tone Influence in Tonal Dialect: A significant difference in F0 was observed between sentence 2 (low tone) and sentence 4 (high tone) in the tonal dialect, potentially due to the combination of focus and high tone.
- Listing Patterns: In lists, the non-tonal dialect showed higher F0 maxima for the first and third nouns in sentence 6 (low tone) compared to sentence 5 (high tone), contrary to expectations. The tonal dialect showed higher F0 maxima for high-toned nouns in the first two positions of the list.
Limitations and Future Work
- Truncation: The paper notes that the truncation phenomenon (compression of F0 movement on voiceless stops) applies to both dialects. However, the authors suggest this might indicate that F0 maxima are more important than minima, a point that could be further explored.
- Speaker Engagement Nuances: While speaker engagement was shown to influence pitch, the specific mechanisms and extent of this influence, especially in relation to different types of engagement, could be a direction for future research.
- Generalizability: The study focused on specific sentence structures and elicited speech. Further research with more diverse spontaneous speech and a wider range of prosodic contexts would be beneficial to confirm the generalizability of these findings.
- Interaction Complexity: The paper highlights the complex interplay between lexical tone, focus, and speaker engagement. Future work could delve deeper into the precise weighting and interaction of these factors in shaping phrase-final prosody.
Summary updated