Production of Australian English language-specific variability

Abstract

This study presents acoustic and articulatory (EMA) data for the high rounded central vowel at different prosodic boundaries. Data are from three speakers of Australian English. Overall, there is very little variability in production of this vowel according to prosodic boundary, with the two female speakers showing slightly more variability than the male speaker. In particular, variability in F2 is limited, suggesting that the vowel is truly central and cannot be considered a back vowel phonologically. The present results are compared with results presented in Tabain & Perrier from three French speakers producing the high back rounded vowel /u/. The differences in articulatory strategies and in acoustic variability are discussed with respect to the vowel inventory of each language. It is speculated that the back vowel /u:/ has become centralized in several dialects of English due to ease of articulation considerations, as part of a strategy to maintain formant cues indicative of a high rounded vowel.

Keywords:

• Articulatory Prosody
• Vowel Centralization
• Australian English

Acknowledgements

I would like to thank Richard Beare for programming assistance with this study; Linda Cupples and Denis Burnham for their support; and Pierre Badin for useful comments. Earlier versions of this work were presented at the 2005 conference of the Australian Linguistic Society, held in Melbourne; and at the 2005 Australian Language and Speech conference, held in Sydney. This research was funded by an Australian Research Council Discovery Projects grant.

Notes

²At this point, it is worth outlining the role of vowel duration in Australian English. There is a long – short vowel distinction (sometimes described as ‘‘tense – lax’’) in this language. The short counterpart of the vowel ( as in “who'd”) is the high back vowel ( as in ‘‘hood’’). The vowel contrasts with both the high-mid back vowel /o:/ (as in ‘‘hoard’’), whose short counterpart is ( as in ‘‘hot’’); as well as with the rounded, mid vowel (as in ‘‘heard’’). This last vowel is transcribed as by Harrington, Cox & Evans (1997) – i.e. as an unrounded central mid vowel – even though the vowel is believed to be rounded in the speech of many speakers. Throughout this paper, only the system of long monophthong vowels is considered (i.e. the systems of short vowels and diphthongs are ignored). For this reason, the length diacritic will not be marked for the English vowels. It should be noted that in French there is no distinction of vowel length. See Harrington et al. (1997) for the transcription of Australian English vowels.

³There are, of course, some differences in articulatory strategy between speakers in all of the studies mentioned here. For instance, in the Harrington et al. (2000) study, one English speaker raises the tongue while at the same time lowering the jaw at stronger prosodic positions, while in the Tabain & Perrier (2005) study, one French speaker retracts the tongue at stronger boundaries in order to achieve the same acoustic effect as the other speakers. However, it seems to be the case that all English speakers aim to increase or at least maintain sonority, whereas all French speakers aim to enhance the formant structure of the vowel, presumably at the expense of sonority. The reader is referred to the studies cited for further detail and for discussion of articulatory-acoustic relations of /i/.

⁴According to an eta² analysis, all three speakers had a weak effect of prosodic boundary on F3, and a medium effect on F4. F4 was not explored further, since it is generally not thought to be under the active linguistic control of speakers. There was some evidence that speaker MT aimed to control the distance between F3 and F2: however, this was not explored further since F3 is generally not thought to be relevant for English vowels. It is, however, note-worthy that speaker MT is the only speaker who is fluent in languages other than English – including French, which has rounded vowels – and is also the only speaker who was aware of the purpose of the study.

⁵Although there is a slightly greater possibility of labeller bias with hand-labelling of data, a comparison of automatically labelled kinematic data (no manual correction) with purely hand-labelled data carried out on a previous study (Tabain 2003) showed similar mean values using the two labelling methods, but less variability around the means using the hand-labelling method.

⁶In an earlier version of this study, the TB sensor was used for speaker RB. Although the overall statistical results were slightly stronger for this speaker when the TB sensor was used as the main sensor, there was no difference in the pattern or interpretation of his results when a change was made to the TT sensor.

⁷The patterning of the Utterance data between the Intonational and Intermediate data cannot be explained for now. There is the possibility of sampling error for F2 by taking the measurement at the temporal midpoint, and there is also the possibility of some other aspect of the articulation that is not studied here also having an effect on the formant values.

⁸An anonymous reviewer points out that with only three speakers in the present study (due to the issues involved in gathering articulatory data), it is difficult to draw firm conclusions on variation and change for male and female speakers. Further articulatory, acoustic and sociophonetic studies may enhance our understanding of such issues. The reviewer also points out that with one of the speakers being the author (MT), it is possible that the observer's paradox applies for this speaker.

Additional information

Notes on contributors

Marija Tabain1

Marija Tabain, Linguistics Program, La Trobe University, Melbourne VIC 3086