A duration-based account of speech rhythm in Indian English Robert Fuchs, robert.fuchs@uni-muenster.de Englisches Seminar, Westf¨ alische Wilhelms-Universit¨ at M¨ unster/M¨ unster University July 28th, 2012; LabPhon Stuttgart Speech rhythm in varieties of English Native varieties of English (British, American, Canadian, Australian, New Zealand English) have been described as stress-timed. Nativised varieties of English (Indian, Nigerian, Jamaican, Singapore English and others) have been described as more syllable-timed than native varieties (Platt et al. 1984, 136; Wells 1982). Previous research on Nigerian (Gut 2005; Gut and Milde 2002; Gibbon and Gut 2001) and Singapore English (Deterding 1994; Deterding 2001; Low et al. 2000), among others, provided evidence for lower durational variability compared to British English. Phonology of Indian English Phonological features indicative of a tendency for syllable-timing have been attributed to Indian English (IndE): Vowel reduction not as strong as in British English (BrE) Monophthongisation of /ei/ to /e/ and /ou/ to /o/ (monophthongs shorter) Tense-lax contrast not maintained by some speakers of IndE Consonant cluster reduction Research questions Any acoustic evidence for attributing a tendency towards syllable-timing to IndE? What influence, if any, does a speaker’s L1 have on their speech rhythm in IndE? Data: Read and spontaneous speech Reading (two to three minutes) and spontaneous speech (ca. five minutes) for 16/11 speakers of IndE and 10/9 speakers of BrE (read/spontaneous). Mock police interrogation to elicit spontaneous yet highly comparable spontaneous speech (Nolan et al. 2006). Informants play the role of a suspect in a drug trafficking case. All knowledge they possess about the case is displayed on slides. Informants face the double task of cooperating with the police and answering questions truthfully where possible (information displayed in black), while not divulging any incriminating facts (information displayed in red). These cognitive demands make them less self-conscious. 1 you pass through Parkville sister’s house petrol station PARKVILLE DIXON A 40 Boyd Street Hatfield Avenue Westlake Bypass Sociological profile of informants Indian informants are students at a prestigious university in Hyderabad, Andhra Pradesh, are highly fluent in English and for the most part went to prestigious English-medium convent schools. They have either Hindi, Bengali (Indo-European), Telugu or Malayalam (Dravidian) as L1. Interviews were conducted in early 2012 by the author. British informants are students at Cambridge University and speak Standard Southern British English. Interviews were conducted by the DyViS research group (Nolan et al. 2006). Methodology Automatic segmentation with HTK toolkit and P2FA, corrected manually → durations of vocalic and consonantal intervals Segmentation by the same rater for both BrE and IndE, ensuring high comparability Syllabification (Maximum Onset principle) The following metrics were computed for all inter-pausal intervals longer than 4 syllables/4 vocalic intervals, last syllable discarded: Global metrics, not normalised for speech rate: Δ C, Δ V, % V (Ramus et al. 1999) Global, normalised metrics: VarcoC, VarcoV (Dellwo 2006; White and Mattys 2007) Local metrics: rPVI-C, nPVI-V (Low et al. 2000) From the above, speech rate-normalised vocalic metrics have been found to be more realiable (White and Mattys 2007; Wiget et al. 2010). Syllabic metrics: nPVI-VC (Liss et al. 2009), nPVI-S, Rhythm Ratio (Gibbon and Gut 2001), Variability Index (Deterding 1994; Deterding 2001), YARD (Wagner and Dellwo 2004), VarcoS (Rathcke and Smith 2011) Control/Compensation Index: CCI-V, CCI-C (Bertinetto and Bertini 2008) All metrics were calculated individually for every inter-pausal interval. The speaker score is the median of all such intervals from one speaker. Hypotheses The metrics will show lower variability for IndE than for BrE. Normalised vocalic metrics in particular should show this, but also syllabic metrics. Speech rhythm differences between IndE and BrE are more pronounced in spontaneous speech than in reading. CCI will be further towards compensation for BrE, and control for IndE. No differences between speakers of IndE depending on L1. Results: Overview Significantly less variability in IndE than in BrE (repeated measures t-tests on speaker scores, others p>0.1): Read Spontaneous Hypothesis VarcoV p<0.0000 p=0.01 ✓✓ nPVI-V p<0.01 p<0.01 ✓✓ nPVI-VC p<0.05 n.s. ✓✗ nPVI-S n.s. p= 0.06 ✗ (✓) %V n.s. p=0.13 ✗✗ Vocalic variability (reading) Lower variability of vocalic interval durations for IndE (N=16) than for BrE (N=10) speakers. ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 50 55 60 65 70 40 45 50 55 60 Vocalic variability in read speech nPVI-V VarcoV BENG1f BENG3f BENG6m BENG7m HIN10f HIN12f HIN3m HIN4f MAL10m MAL2f MAL7f MAL9m TEL4f TEL5m TEL7m TEL9f GB10m GB11m GB12m GB13m GB18m GB1m GB20m GB21m GB3m GB62m Syllabic variability (spontaneous) Lower variability of syllable and vocalic interval durations for IndE (N=11) than for BrE (N=9) speakers. This is the only syllable- based metric showing near-significant differences. ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 45 50 55 45 50 55 60 65 Syllabic and vocalic variability in spontaneous speech nPVI-S nPVI-V GB10m GB11m GB12m GB13m GB18m GB1m GB20m GB21m GB3m BENG1f BENG3f BENG6m HIN10f HIN12f HIN3m MAL2f MAL9m TEL4f TEL5m TEL7m GB = BrE, BENG = IndE L1 Bengali, HIN = IndE L1 Hindi, MAL = IndE L1 Malayalam, TEL = IndE L1 Telugu Reading vs. spontaneous speech VarcoV and nPVI-V were slightly, but insignificantly, higher for spontaneous speech than for reading in both varieties. This might be due to hesitation phenomena being more prevalent in spontaneous speech. Conclusions Evidence for somewhat more syllable-timed rhythm in educated/acrolectal IndE compared to BrE due to less variability in durations of vocalic intervals in IndE than in BrE. Less educated/basilectal IndE speakers might show stronger tendency towards syllable-timing. Non-normalised and consonantal metrics did not show differences, but are known to be less reliable; Consonant cluster reduction apparently not widespread in educated IndE Only very tentative evidence for less variability in syllable durations for IndE Absence of difference for CCI might be due to questionable status of tense/long vs. lax/short vowels in IndE. L1 has no discernible influence on IndE rhythm, but confident conclusions only possible with larger sample. References Bertinetto, Pier Marco and Chiara Bertini (2008). “On modeling the rhythm of natural languages”. In: Proceedings of Speech Prosody 2008, Campinas, Brazil. Ed. by Pl´ ınio A. Barbosa, Sandra Madureira, and Cesar Reis. ISCA Archive, pp. 427–30. url: http://www.isca-speech.org/archive/sp2008. Dellwo, Volker (2006). “Rhythm and speech rate: A variation coefficient for deltaC”. In: Language and Language-Processing. Proceedings of the 38th Linguistics Colloquium. Ed. by Pawe ˜ l Karnowski and Imre Szigeti. Frankfurt am Main: Peter Lang, pp. 231–241. Deterding, David (1994). “The rhythm of Singapore English”. In: Proceedings of the Fifth Australian International Conference on Speech Sciene and Technology. Ed. by Roberto Togneri. Perth: Uniprint, pp. 316–321. — (2001). “The measurement of rhythm: a comparison of Singapore and British English”. In: Journal of Phonetics 29, pp. 217–230. Gibbon, Dafydd and Ulrike Gut (2001). “Measuring Speech Rhythm”. In: Proceedings of Eurospeech 2001. Aalborg, Denmark, pp. 91–4. Gut, Ulrike (2005). “Nigerian English prosody”. In: English World-Wide 26.2, pp. 153–77. Gut, Ulrike B. and Jan-Torsten Milde (2002). “The prosody of Nigerian English”. In: Proceedings of Speech Prosody 2002. Ed. by Bel Bell and Isabelle Marlien. Laboratoire Parole et Langage. Aix-en-Provence. Liss, Julie M., Laurence White, Sven L. Mattys, Kaitlin Lansford, Andrew J. Lotto, Stephanie M. Spitzer, and John N. Caviness (2009). “Quantifying speech rhythm abnormalities in the dysarthrias”. In: Journal of Speech, Language and Hearing Research 52.5, pp. 1334–1352. Low, Ee Ling, Esther Grabe, and Francis Nolan (2000). “Quantitative characterization of speech rhythm: Syllable-timing in Singapore English”. In: Language and speech 43.4, pp. 377–401. Nolan, Francis, Kirsty McDougall, Gia de Jong, and Toby Hudson (2006). “A Forensic Phonetic Study of Dynamic Sources of Variability in Speech: The DyViS Project”. In: Proceedings of the 11th Australasian International Conference on Speech Science and Technology. Ed. by P. Warren and C.I. Watson, pp. 13–18. Platt, Talbot, Heidi Weber, and Mian Lian Ho (1984). The new Englishes. London/Melbourne: Routledge. Ramus, Franck, Marina Nespor, and Jacques Mehler (1999). “Correlates of linguistic rhythm in the speech signal”. In: Cognition 73, pp. 265–92. Rathcke, Tamara and Rachel Smith (2011). “Exploring timing in accents of British English”. In: Online Proceedings of the 17th International Congress of Phonetic Sciences. Hong Kong, pp. 1666–9. url: http://www.icphs2011.hk/ICPHS_CongressProceedings.htm. Wagner, Petra and Volker Dellwo (2004). “Introducing YARD (Yet Another Rhythm Determination) And Re-Introducing Isochrony to Rhythm Research”. In: Proceedings of Speech Prosody 2004. ISCA, pp. 227–230. Wells, John C. (1982). Accents of English. Cambridge: Cambridge University Press. White, Laurence and Sven L. Mattys (2007). “Calibrating rhythm: First language and second language studies”. In: Journal of Phonetics 35.4, pp. 501–522. Wiget, Klaus, Laurence White, Barbara Schuppler, Izabelle Grenon, Oleysa Rauch, and Sven L. Mattys (2010). “How stable are acoustic metrics of contrastive speech rhythm?” In: Journal of the Acoustical Society of America 127.3, pp. 1559–1569. Robert Fuchs, robert.fuchs@uni-muenster.de M¨ unster University Speech rhythm in Indian English