Prosody, prominence and segments in Djambarrpuyŋu

Abstract

This thesis is an investigation of the phonetics of prosodic structure and prominence in Djambarrpuyŋu, an Australian Indigenous language of the Pama-Nyungan language family spoken in northeast Arnhem Land, Northern Territory, Australia. The aim of this study is to provide a phonetic description of aspects of prosody in Djambarrpuyŋu that contributes to the phonological and phonetic understanding of this language, and that will inform phonological and phonetic investigations of Australian languages in the future. Specifically, it provides a phonetic account of the effects of word- and phrase-level prominence on segments, and how information structure is expressed by intonational means. Working with controlled speech data collected with Djambarrpuyŋu speakers on the island community of Milingimbi (Yurrwi), and through a perception study, this thesis is the first substantial quantitative phonetic analysis of Djambarrpuyŋu.

Starting from previously described phonological features of the vowel system, the acoustic characteristics of the vowel length contrast are investigated in a production experiment. The results reveal that contrastively long vowels /ɪː, ɐː, ʊː/ are approximately twice as long as short vowels /ɪ, ɐ, ʊ/. The contrast is enhanced by different fundamental frequency contours between the length categories (similarly to Finnish) and by an inverse durational relationship with the following consonant (in a similar way to Swedish, Washo, and Thai, for example). A forced-choice perception study investigating the vowel length contrast and the effect of consonant duration on listeners’ categorisation of words in the minimal pair of wäŋa /wɐːŋɐ/ “home/place/country” and waŋa /wɐŋɐ/ “talk/speak”, shows that consonant duration is used by listeners particularly when the vowel’s length category is ambiguous.

Word stress patterns previously reported for Djambarrpuyŋu include primary stress on the initial syllable and second stresses assigned to alternating odd syllables through a word or to the first syllable in suffixing morphemes. These patterns are typical of those described for Australian Indigenous languages. The results of a quantitative acoustic analysis partially support the previous description. The initial syllable of words does have the highest fundamental frequency of all syllables within a word. However, other correlates of prominence that are not pitch-related such as longer stressed vowels, higher intensity of stressed vowels, and more peripheral stressed vowel quality, are not found in the data. Furthermore, evidence for secondary stress, considered for both morpheme internal and morpheme initial secondary stresses is not found, suggesting that Djambarrpuyŋu is like other Australian languages where there is one main stress that is strong in terms of phonotactic restrictions and the association of a post-lexical pitch accent but shows little acoustic support for further word-internal strength relations.

The effects of proximity to prosodic constituent boundaries and prosodic prominence on consonants is investigated. Australian languages have been proposed to unusually show strengthening (temporal and articulatory) of consonants following tonic (i.e., accented) vowels. Pre-boundary lengthening commonly described cross-linguistically is also reported, though strong effects of post-boundary lengthening are not as common. An acoustic experiment showed that post-boundary nasals following a pause have shorter duration than nasals following a segment (like findings for English), and that pre-boundary lengthening occurs in Djambarrpuyŋu for nasals preceding pauses. Consistent post-tonic consonant lengthening is not found in these data unlike other Australian languages such as Mawng, Bininj Kunwok, and Walpiri. However, further analysis comparing post-tonic consonants with consonants from non-focal or non-accented words as well as articulatory investigation would prove illuminating.

Considering prosodic structure above the word, the role of intonation in encoding information structure is examined. The proposed intonational phonology of Djambarrpuyŋu is presented. In its tonal inventory, Djambarrpuyŋu is similar to other Australian languages for which there is an intonational analysis. It is shown that focus and topic elements can be marked with an intonational pitch accent associated with, or close to the initial syllable of the word. However, neither gradient nor categorical measures differentiated different uses of focus or topic categories. The contours and placement of pitch accents in questions follow cross-linguistic patterns and support previous impressionistic descriptions of Djambarrpuyŋu. Unusually for Australian languages, some words were also found to be deaccented.

This study reveals that while Djambarrpuyŋu shows striking similarities with other Australian languages across language family boundaries, the results also highlight that there is diversity in acoustic patterns of segments and intonational systems, demonstrating the value in phonetic research of these languages that are often considered similar in their phonological systems.