Devices & Hardware

13 views

/ɐlC/-/ɔlC/ Sound change in Australian English: Preliminary res[ɔ]lts

/ɐlC/-/ɔlC/ Sound change in Australian English: Preliminary res[ɔ]lts
of 4
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Share
Transcript
  /   lC/-/   lC/ Sound change in Australian English: Preliminary res[   ]lts  Eleanor Lewis, Deborah Loakes   School of Languages & Linguistics, The University of Melbourne, Australia elewis;dloakes@unimelb.edu.au Abstract This paper provides a preliminary investigation of an apparent sound change in AusE, whereby the STRUT  vowel is produced the same way as the LOT  vowel when preceding consonant cluster /l/ (resulting in minimal pairs such as cult/colt   and  gulf/golf   which become homophones). Results provide some clear evidence of a sound change in progress, but also show that there is considerable interspeaker variation.  Index Terms : phonetics, sound change, Australian English 1.   Introduction 1.1.   The pre-lateral environment The pre-lateral context is a known catalyst for sound misperception and change, due to the strong coarticulatory force of coda /l/ on a preceding vowel [1]. Dark (or “velarized”) /l/, which occurs syllable -finally (and perhaps also syllable-initially [2]) in AusE, is highly resistant to coarticulatory effects from adjacent vowels [3]. As a result, this variant has strong coarticulatory influence over neighboring vowels. Acoustically, dark /l/ has substantially lower second formant frequencies than light /l/, as well as slightly higher first formant frequencies [3] [4], corresponding with greater retraction and lowering of the dorsum (F2) and less apical raising (F1). Vowels preceding dark /l/ generally have second formant frequencies that are significantly lower than those of vowels preceding light /l/ [5], indicating retraction of the tongue body in anticipation of the required position for the following segment. Vocalized /l/ is another lateral variant known to occur word-finally in AusE (e.g. [6] and [7]). This variant lacks the apico-alveolar contact involved in the production of both light and dark /l/ [6], and is therefore vowel-like in sound. The coarticulatory effects of vocalized /l/ on preceding vowels have not, to our knowledge, been investigated experimentally. 1.2.   Pre-lateral sound change in AusE In AusE, the pre-lateral environment has been shown to have significant effects on the first two formants of vowel targets, F2 more so than F1. This effect is strongest on front vowels, where all AusE front monophthongs, and the central / ɜː /, have  been found to be more retracted and somewhat lower when realized pre-laterally [8] [9]. High central / ʉː / displays the most extensive retraction in this context, moving almost to the  back of the vowel space (e.g. [9]). It has also been found that /i ː / becomes bisyllabic pre-laterally, due to the conflicting high degrees of constraint involved in the production of both this vowel and dark /l/. The vowel reaches its target (a  peripheral point as seen in Figures 1 and 2 below) earlier in this context, before passing through schwa space on its way to the lateral target [9]. The strong coarticulatory effect of the pre-lateral context in AusE renders this environment a likely site for vowel variation and sound change. Pre-lateral sound changes documented thus far for AusE include tense vowels that “collapse” into the corresponding lax vowel (i.e. deal  / dill   and  fool  /  full   becoming homophones) [10], and a regionally-defined neutralization of the DRESS / TRAP  distinction, observed in the state of Victoria, e.g. [11], [12]. 1.3.   /   lC/->/   lC/ merger? In his 1985 study of the effects of post-vocalic /l/ [8], Bernard mentions the existence of another pre-lateral vowel merger, whereby words containing a STRUT  vowel which precedes coda cluster /l/ (e.g.  gulf  , cult  ), are pronounced using the LOT  vowel, making  gulf   homophonous with  golf  , for example. He notes this to be a South Australian pronunciation. / ɐ lC/->/ ɔ lC/ appears not to have been investigated in any other studies of Australian English, however auditory impressions of the authors suggest it occurs frequently in Melbourne English, and  perhaps AusE as a whole. This idea is supported by a number of pop-cultural or folk-linguistic resources. For example, a cinema in Melbourne hosts a regular “Cult Vault”   event  [ 13], a title which appears to use cult   and vault   as rhyming words. This usage suggests that, in this context, both cult   (transcribed by the Macquarie Dictionary using the STRUT vowel) and vault   (transcribed with both THOUGHT  and LOT listed as possible vowel variants) are being pronounced using the LOT vowel. A Melbourne-based (but Sydney-born) cryptic crossword author David Astle (“DA”) has also used  gulf   and  golf   as homophones in his clues, most recently in April 2011. This was a topic of debate on crossword dis cussion website “The DA Trippers”, however, with many people objecting to the use of the two words as homophones (e.g. ‘ [Clue] 24A exceedingly lame. Golf and gulf are a world away.’ —   posted  by “RobT”)[14 ]. In terms of folk-linguistic reference to this change in AusE, a now-deleted Wikipedia page about this merger suggested that words like bulk  ,  pulse , culture  and multiply  are often realized using the LOT  rather than STRUT  vowel [15]. The page was removed in 2006 due to a lack of published research on the phenomenon. 2.   Aims The current paper aims to investigate the production of / ɐ lC/ in Australian English, specifically: i.   Is there evidence of / ɐ lC/-/ ɔ lC/ merger in AusE?  ii.   If yes, what is the nature of acoustic variability across / ɐ lC/ productions? 3.   Method Two experiments were designed to investigate the realization of / ɐ lC/-/ ɔ lC/ in AusE. The first uses a citation speech corpus, while the second makes use of a corpus of read sentences (i.e. words in connected, but not spontaneous, speech). Given the gradient acoustic, auditory, and articulatory differences between dark and vocalized /l/ variants, allocating one of the two labels to a token is often extremely difficult (e.g. [6]). For this reason, the present study focuses only on the vowel realization, with a plan to investigate /l/ quality in future work on the merger. 3.1.   Experiment One: Melbourne speakers Participants In experiment one, data from five female participants were analyzed. All were relatively young (aged 21-38 at the time of recording), and all were raised and residing in Melbourne. Corpus and recording procedure Participants were recorded reading a wordlist containing various stimuli, including the minimal pairs cult/colt   and  gulf/golf  , and all AusE vowels in /hVd/ and /hVl/ contexts. An H4N Zoom digital recorder was used, and the device’s inbuilt microphones were used with a sampling rate of 44kHz with 16-bit resolution. Analysis The two minimal pairs involving / ɐ lC/ and / ɔ lC/ sequences were extracted from the recordings. Words containing the AusE monophthongs in /hVd/ and /hVl/ contexts were also extracted to measure speake rs’ baseline productions. Words were segmented using Praat (v.5.2.18) and labeled following the guidelines outlined for the ANDOSL corpus. Formant steady states were also located for all vowels. Annotations were then converted for use with the EMU Speech Database System (v.2.3), where spectrograms were further inspected, and formant tracking errors manually corrected. Vowel targets, taken at the midpoint of vowels’ steady states, were  plotted using R (v.2.12) (averages for the /hVd/ and /hVl/ monophthongs, and individual points for the minimal pair words). Duration data were also analyzed at this stage (using a t-test in R), to assess whether the two phonemes are distinguished using durational cues in addition to formant frequencies. Because of the gradient transition often seen  between vowels and following laterals, the duration of the rhyme was used here. 3.2.   Experiment Two: ANDOSL data Participants In experiment two, six young female speakers  —   those classified as speakers of the “general” AusE sociolect —   were selected from the ANDOSL (Australian National Database of Spoken Language) corpus, recorded in Sydney in the 1990s. These speakers, aged 18-30 years, were chosen as the group most comparable to the present-day Melbourne group used in the first experiment. Analysis Six sentences, containing a total of seven tokens of / ɐ lC/, were located in the corpus: 88. She had scarcely divulged the scandal before it was  splattered over the front pages of the tabloids 90. We were plunged into darkness as the clouds engulfed the moon 147. The chef was in dispute with his colleagues as to whether barley was a pulse 157. It's in vogue to make films which explore the relationship between culture and faith 161. The writer of the obituary tried not to indulge his  prejudice against the bourgeoisie 163. The multi-bulbed light fitting in the lounge was an improvement on the neon tube Control words containing the AusE monophthongs in /hVd/ context were also located for the same young female general (YFG) speakers (/hVl/ monophthongs were not available for these speakers). Words were segmented, labeled, and prepared for EMU/R use following the same procedure outlined for experiment one, above. Vowel target plots were again created using R (averages for /hVd/ monophthongs, and all points for / ɐ lC/ tokens). 4.   Results 4.1.   Experiment One: Melbourne speakers Figures 1 and 2 show the five Melbourne speakers’ vowel targets, produced in cult/colt   and  gulf/golf  , as compared to  baseline /hVd/ and /hVl/ monophthongs. As seen in Figure 1, speakers’ productions of / ɐ lC/ display a large amount of variation, ranging from relatively close to baseline STRUT  targets, to higher and more retracted than baseline LOT . Figure 1-  Melbourne females' productions of /  ɐ   lC/ and /  ɔ  lC/    For some speakers there is a separation of / ɐ lC/ and / ɔ lC/, while for others there is substantial acoustical overlap in  productions of the two phonemes. When productions are  separated according to speaker, these differences become apparent, as seen in Figure 2 (below), which shows averaged values for /hVd/ and /hVl/, and individual results for / ɐ lC/ and / ɔ lC/. The bar between the values connects each speaker’s two tokens of a phoneme. Figure 2-  Melbourne females' productions of /  ɐ  lC/ and /  ɔ  lC/ (separated by speaker)  Speakers ML03, ML12 and ML14 (shown in yellow, orange & red in Figure 2) produce their / ɐ lC/ vowels in approximately the same acoustical space as their / ɔ lC/ vowels,  both occurring close to baseline LOT  productions. Speaker ML01 (seen in blue) keeps her productions of the two contrasts separated, with / ɐ lC/ tokens that pattern lower and further forward (i.e. closer to control STRUT ), while her / ɔ lC/ tokens pattern with LOT . Speaker ML13 (in green) realizes her two / ɐ lC/ tokens differently; one (  gulf  ) sits in the same vowel space as LOT , while the other ( cult  ) is halfway between  baseline / ɐ / and / ɔ / in the vowel space. For this token, the  percept for the authors is a STRUT vowel.  No statistical difference was found between the duration of rhymes containing the STRUT and LOT vowels (t (15)=0.24, p=0.82). 4.2.   Experiment Two: ANDOSL data Turning now to the read speech data from the ANDOSL database, Figures 3 and 4 show the speakers’ productions of the / ɐ lC/ vowel in connected speech (7 tokens x 6 speakers). Figure 3 (below) shows a wide ellipse, indicating that this gro up’s productions of / ɐ lC/, like those of the Melbourne corpus, display substantial variation within the vowel space. This once again suggests that some speakers are conserving a variant close to baseline STRUT , while others produce a much higher, more retracted / ɐ lC/ vowel which is acoustically more similar to the control-context LOT  vowel. Figure 3-  ANDOSL YFG speakers’ productions of /  ɐ   lC/ The interspeaker variation within the ANDOSL speakers’ / ɐ lC/ realizations is shown in Figure 4 (below), which plots / ɐ lC/ tokens by speaker. Figure 4-  ANDOSL YFG speakers’ productions of /  ɐ  lC/ (separated by speaker)   Speakers S010, S018 and S020 (plotted in orange, yellow & red in Figure 4) all produce raised and retracted tokens of the / ɐ lC/ vowel which are perceptually and acoustically more similar to baseline LOT  than baseline STRUT . Speakers S033 & S034 (in light and dark blue) produce lower, fronter tokens than speakers S010, S018 and S020. These vary somewhat in height (i.e. some are as low as baseline STRUT , while others are halfway between the heights of baseline STRUT  and LOT ),  but all are perceptually closer to baseline STRUT . The   productions of the remaining speaker, S042 (in green), sit  between the two other groups: this speaker consistently  produces tokens which are in between baseline LOT  and STRUT  in the vowel space. 5.   Discussion The vowel target plots generated for Melbourne speakers in Experiment One provide some evidence of an / ɐ lC/->/ ɔ lC/ change in progress in AusE: three out of the five speakers are  producing cult/colt   and  gulf/golf   as homophones, while one additional speaker produces  gulf/golf   the same way (but keeps cult   distinct from colt  ). Looking at the ANDOSL data, recorded almost twenty years ago, we see further evidence of the / ɐ lC/ vowel being produced closer to baseline LOT  than STRUT  vowels for many speakers. This suggests firstly that any / ɐ lC/->/ ɔ lC/ merger is not regionally restricted to Melbourne English, and secondly that it has been occurring in AusE (outside of South Australia, where it was first mentioned in [8]) for some time. Future analysis of the middle-aged and older ANDOSL speakers’ productio ns of this variable will investigate whether the merged variant has evolved relatively recently in AusE, or is a more longstanding coarticulatory effect of variable /l/ realizations. In both the present-day Melbourne data, and 1990s ANDOSL data, considerable variation is seen in the realization of the STRUT  vowel in this context (i.e. /__lC). This high level of variability is what one might expect from a vowel in the process of changing in the vowel space (e.g. [16]). In both sets of data, speakers can be found to pattern differently to one another with respect to their realization of the variable. There are speakers who realize / ɐ lC/ and / ɔ lC/ the same way, others who completely preserve the distinction  between the two, and then speakers whose productions routinely fall somewhere between the positions in the vowel space occupied by baseline / ɐ / and / ɔ /. The “in -  between”  productions of this last group may be the source of the listener misperception driving this sound change (i.e. listeners hear a high, retracted token of / ɐ lC/, and misperceive this as [ ɔ lC], changing their perception of what constitutes / ɐ lC/, following [16]). An important consideration for the current work is why   this variant of STRUT  occurs in this particular context, while speakers’ productions of the same vowel in the control word hull   usually remain close to baseline STRUT . While it very much remains to be investigated, one might hypothesize that it is related to /l/ vocalization. Vocalization of coda /l/ has been found to occur more frequently in South Australia than other states, e.g. [10] (although it does also occur in AusE more generally), and in terms of phonetic context, more vocalization has been found to occur in coda clusters than simple codas [7]. Both of these factors have also been noted to promote STRUT - LOT  merger, which suggests a possible link between the two. While the coarticulatory effect of vocalized /l/ on  preceding vowels has not been investigated experimentally, it has been suggested that dark /l/ can perhaps be thought of as the second element of a diphthong, because of its vocalic nature (e.g. [4]). Given that vocalized /l/ is also quite vowel-like, we might expect this to also be the case for this variant. If this is so, it can be suspected by extension that the preceding vowel must in some way adapt to being the first part of a diphthong. This may well be what is occurring in AusE, and warrants further investigation. 6.   Conclusions Analysis of production data from both the present-day and 1990s provide evidence for a sound change in AusE, whereby / ɐ lC/ and / ɔ lC/ are both realized as [ ɔɫ C] by some speakers. A large amount of interspeaker variation is seen in production of the STRUT  vowel in this context, however, with some speakers maintaining the distinction between the two phonemes, other speakers completely merging the two, and others still whose realizations of / ɐ lC/ are acoustically midway between baseline STRUT  and LOT  productions. Further work is required on this change, including analysis of a wider variety of AusE speakers (including middle-aged/older speakers, and speakers of different sociolects),  perception studies to test whether the phoneme contrast is maintained in perception, and investigation into the link  between / ɐ lC/-/ ɔ lC/ merging and /l/ quality. 7.   References [1]   Labov, W. Principles of linguistic change. Volume 1: Internal factors. Oxford: Blackwell. 1994 [2]   Wells, J. Accents of English 3: Beyond the British Isles. Cambridge University Press. 1982 [3]   Bladon, R. A. W. & Al- Bamerni, A. “Coarticulation resistance in English /l/”, Journal of Phone tics. 4: 137-150. 1976. [4]   Espy- Wilson, C. Y. “ Acoustic measures for linguistic features distinguishing the semivowels /w j r l/ in American English ”, in JASA, 92(2): 736-757. 1992 [5]   Lehiste, I. Acoustical characteristics of selected English consonants. The Hague: Mouton & Co. 1964 [6]   Hardcastle, W. & Barry, W. “ Articulatory and perceptual factors in /l/ vocalisations in English”,  JPhon, 15: 3-17. 1989 [7]   Borowsky , T. “The vocalization of dark /l/ in Australian English”, in D. Blair & P. Collins (eds.) English in Aust ralia. Amsterdam: Benjamins, 69-87. 2001 [8]   Bernard, J. R. L. “Toward the acoustic specification of Australian English”, Zeitschrift f  ür Phonetik 2/3: 113-128. 1985. [9]   Palethorpe, S. & Cox, F. “ Vowel modification in pre-lateral environments”, p oster presented at Sixth International Seminar on Speech Production. Sydney, Australia, December 2003 [10]   Bradley, D. “ Regional characteristics of Australian English: Phonology”, i n E. W. Schneider et al. (eds.) A handbook of varieties of English: A multimedia reference tool. Volume 1: Phonology.    New York: Mouton de Gruyter, 645-655. 2004 [11]   Cox, F. & Palthorpe, S. “ The border effect: Vowel differences ac ross the NSW / Victorian border”, in C. Moskovsky  (ed.) Proceedings of the 2003 Conference of the Australian Linguistics Society, 1-14. 2004 [12]   Loakes, D., Hajek, J., & Fletcher, J. “ /æl/-/el/ transposition in Australian English: Hypercorrection or a competing sound change?” , in Proceedings of ICPHS XV11, Hong Kong: ICPhS, 1290-1293. 2011 [13]   Palace Cinemas Westgarth, Cult Vault advertisement http://www.palacecinemas.com.au/events/cultvaultatpalacewestgarth/ [14]   The DA Trippers. DA Debate for 22/23 April 2011. http://datrippers.com/2011/04/25/da-debate-for-the-2223rd-of-april-2011/ [15]   Wikipedia. Articles for deletion: Golf/gulf merger http://en.wikipedia.org/wiki/Wikipedia:Articles_for_deletion/Golf-gulf_merger [16]   Ohala, J. “The phonetics of sound change”, i n C. Jones (ed.) Historical linguistics: Problems and perspectives . London: Longmann, 237-278. 1993
Advertisement
Related Documents
View more
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks