The Vowel System of Korebaju

Jenifer Vega RodríguezInterspeech 2019

TL;DR

This paper provides a detailed acoustic and articulatory description of the vowel system of Korebaju, a Western Tukanoan language. It clarifies previous discrepancies by establishing a phonemic inventory of 15 vowels (6 oral, 6 nasal, 3 glottal) and identifies specific allophones for high vowels, challenging prior assumptions about the presence of a [w] phoneme.

What Problem Does This Paper Address?

The phonological inventory of the Korebaju language has been a subject of debate in previous linguistic studies. Specifically, there has been disagreement regarding the phonemic status of the high central unrounded vowel [i] and the high back unrounded vowel [w]. Additionally, the nature of glottalization (whether it's a suprasegmental feature or a segmental glottal stop) and nasalization (suprasegmental or segmental phoneme) has remained unclear. This paper aims to resolve these ambiguities by providing a data-driven analysis of Korebaju's vowel system.

Methods

The study employed a combination of acoustic and articulatory phonetic methods. Data were collected from six native speakers using synchronized audio and electro-glottography (EGG) signals. A Zoom H6 recorder and Shure WH30 microphone were used for audio capture, while a Glottograph EG2-PC captured EGG data. Recordings were made using Audacity, and data were segmented using Sound Forge. Analysis was performed using Phonedit, Praat, and Winpitch software. A corpus of 275 words was created for acoustic analysis, with 182 words recorded with synchronized EGG data. Each word was repeated three times by each speaker. A carrier sentence was used to ensure semi-authentic speech contexts, with slight variations for male and female speakers due to morphological differences. Formant measurements (F1, F2, F3) were analyzed using Lobanov normalization to visualize vowel distribution. Statistical analysis, including ANOVA and Bonferroni tests, was used to assess differences in vowel duration.

Datasets

The study utilized a custom corpus of 275 words for acoustic analysis and 182 words for synchronized sound-EGG data. This corpus was specifically designed for the study and is not a pre-existing benchmark dataset.

Key Insights and Ideas

  • Clarification of [w] and [i] status: The paper definitively shows that Korebaju does not have a phonemic [w] vowel. Instead, [w] is an allophone of the high back vowel /u/ when it follows a palatal consonant. Similarly, the high central vowel /i/ has an allophone [1] in the same context. This resolves a long-standing point of contention in the literature.
  • Confirmation of a central vowel: The acoustic data confirms the presence of a central vowel, aligning with previous descriptions of [i] in Korebaju.
  • Nasal harmony absence: Contrary to some expectations for languages with nasal vowels, this study found no evidence of nasal harmony in Korebaju.
  • Glottalization as a feature: While glottal vowels /a?, e?, o?/ are contrastive and occur with non-high vowels, the study also identified random occurrences of glottal stops and creaky voice between vowels. These are proposed to be suprasegmental features ([constrictive glottal]) rather than contrastive phonemes, aligning with analyses of other Tukanoan languages.
  • Vowel system complexity: The Korebaju vowel system is more complex than previously thought, comprising 15 distinct phonemic vowels.

Key Results

  • The phonemic vowel inventory of Korebaju consists of 6 oral vowels (/i, e, a, o, u, i/), 6 nasal vowels (/ĩ, ẽ, ã, õ, ũ, ĩ/), and 3 glottal vowels (/a?, e?, o?/).
  • The high back vowel /u/ has an allophone [w] after palatal consonants.
  • The high central vowel /i/ has an allophone [1] after palatal consonants.
  • Acoustic analysis confirmed the formants for the central vowel correspond to the close central unrounded vowel [i].
  • An ANOVA test showed no significant difference in duration between glottal and their corresponding oral vowels.

Limitations and Future Work

  • Random glottal occurrences: The study notes random occurrences of glottal stops and creaky voice between vowels, which are interpreted as suprasegmental features. However, the paper acknowledges that these specific productions are not contrastive in the language, suggesting further investigation into their precise phonetic realization and potential subtle functional roles.
  • Dialectal variation: While the study focused on a single Korebaju village to avoid dialectal influence, the existence of dialectal variations within Korebaju is mentioned. Future work could explore these variations to provide a more comprehensive understanding of the language's phonological landscape.
  • Comparison with other Tukanoan languages: The paper draws parallels with other Tukanoan languages regarding glottalization. Further comparative studies could illuminate broader patterns and historical developments within the language family.

Summary updated

BibTeX
@inproceedings{rodriguez2019vowel,
  author = {Jenifer Vega Rodríguez},
  title = {{The Vowel System of Korebaju}},
  booktitle = {Interspeech},
  year = {2019},
}
Figure