Haʻina ʻia mai ana ka puana: The vowels of ʻōlelo Hawaiʻi

Abstract

This dissertation investigates aspects of the pronunciation of the vowels of ʻōlelo Hawaiʻi, the aboriginal Polynesian language spoken in the Hawaiian archipelago. Specifically, this acoustic description concerns the voices of eight elderly speakers recorded in the 1970s on the Hawaiian-language radio program Ka Leo Hawaiʻi. Measurements of the first and second formants (F1 and F2, in Hz) are extracted for all vowel tokens (n=19,599 after exclusions). The plots and statistical analyses that utilize the spectral F1/F2 measurements are interpreted as reflective of tongue position (high–low, back–front) over the course of vowel production. This description confirms many previous observations regarding the pronunciation of Hawaiian vowels, and presents several novel insights. When pairs of short vowels /a, e, i, o, u/ and long vowels /ā, ē, ī, ō, ū/ are compared, the long vowels appear in more peripheral regions of the vowel space; in particular, /ē/ is located notably front and high compared to /e/. Tokens of unstressed /a, i, o/ are pronounced in less peripheral positions than when they are stressed, while unstressed /e/ is raised. Labial consonants /m, p, w/ are found to condition more back productions of adjacent vowels; alveolar consonants /n, l/ condition more front productions; and /ʔ/ conditions lower productions. This analysis also demonstrates that vowel-on-vowel coarticulation is apparent even when the vowels in question are separated from each other by a consonant and thus not directly adjacent. The coarticulatory effects of a following vowel on stressed /a/ are investigated in detail; in these contexts, significant differences are found in both F1 and F2 according to the following vowel. In other words, there are significant differences in /a/ pronunciation that involve both height and backness between words like mala, male, mali, malo, and malu. Significant differences are also observed in terms of both height and backness for the starting points of the aV diphthongs, /ae, ai, ao, au/. These four diphthongs are found to originate in four separate locations, with /ae, ai/ fronter than /ao, au/ and with /ae, ao/ lower than /ai, au/. Within the /ai/ and /au/ vowels, differences are also observed based on word type: compared to content words, pronouns and directionals exhibit shorter, more centralized vowel trajectories. Phonetic variation is also investigated between and within the eight speakers of this sample. While most speakers maintain a consistent distinction between the spectral trajectories of /ei/ and /ē/ and between /ou/ and /ō/, this is not true of one speaker; while most also maintain a distinction between /āi/ and /ae/, two speakers do not. The words laila (‘there’), maikaʻi (‘good’), and family member terms beginning with kaik- are found among most speakers to contain a vowel more similar to /ei/ than /ai/. For most speakers in the sample, the primary stressed monophthong /a/ in maikaʻi is also pronounced more like a typical /e/ than a typical /a/. Most speakers usually pronounce the /o/ in the words ʻaʻole (‘no’), ʻaʻohe (‘none’), and hope (‘after’) as /a/; one speaker is found to usually use /o/ in these words utterance-finally while using /a/ in non-utterance-final contexts. This study represents the first large-scale, multi-speaker acoustic analysis of Hawaiian. As such, the data processed for this dissertation has the potential to reveal many further insights into twentieth-century Hawaiian pronunciation that reach well beyond the spectral analyses presented here. Many of the findings of this dissertation also provide evidence relevant to linguistic theories that have rarely been explored in the context of indigenous, under-documented languages such as Hawaiian. I conclude with suggestions for the enhancement of this dataset, future phonetic and phonological research extensions, and community-facing pedagogical applications for this work.