The Phonology of Indian English I: Overview

Indian English Session 3: The Phonology of IndE I BEd Seminar Indian English, summer term 2015, University of Münster Robert Fuchs, robert.fuchs@uni-muenster.de Indigenisation of IndE: Overview • Process of indigenisation/ nativisation • Areas of phonology, morphology, syntax, lexis and pragmatics • Greatest differences between varieties of English usually found in their phonology • Phonological differences are more often categorical (e.g. a phoneme is always realised differently than in another variety) than in syntax/morphology, where differences are often distributional (= frequency) Indigenisation of IndE: Lectal differences • Cline/continuum from basilect over mesolect to acrolect • Research on emerging standards in World Englishes usually focusses on educated speakers? -> Why? Indigenisation of IndE: Lectal differences • The process of standardisation (of a language or variety) involves as a key stage the selection of a variety • All social and regional varieties are not equally likely to be chosen as standard -> the variety used by the social/national elite is usually selected (Deumert and Vandenbusche 2003, Trudgill 1999) • Example: The International Corpus of English relies on speakers with a university degree (or in the process of getting one) Questions on the Texts • Sailaja (2012): section 2 • Sirsa and Redford (2013) Phonology of IndE: Overview • In pairs: Make a list of differences between the phonology of Indian English and British English • Distinguish between differences on the following levels – Phonetic segmental (same phoneme realised differently) – Phonological segmental (e.g. two phonemes have merged) – Suprasegmental (stress, intonation, speech rhythm) References • Deumert, A. and Vandenbusche, W. (2003). “Research directions in the study of language standardization,” in Germanic Standardizations: Past to Present, A. Deumert and W. Vandenbusche, Eds. Amsterdam: John Benjamins, 455–467. • Fuchs, R. (to appear). Speech Rhythm in Varieties of English. Evidence from Educated Indian English and British English. Berlin: Springer. • Sailaja, P. (2012). Indian English: Features and sociolinguistic aspects. Language and Linguistics Compass, 6(6), 359-370. • Sirsa, H., & Redford, M. A. (2013). The effects of native language on Indian English sounds and timing patterns. Journal of Phonetics, 41(6), 393-406. • P. Trudgill (1999). “Standard English: What it isn’t,” in Standard English: The Widening Debate. Routledge, 177–128. Sources of Images • Note: Every effort has been made to attribute the images used in this document to their source. If you are the original creator of any of these images, and would like me to add you to this list or want the image removed, please contact me. • 2: http://static.ibnlive.in.com/ibnlive/pix/sitepix/09 _2011/india_uk.jpg • 4: https://ec.europa.eu/digitalagenda/sites/digitalagenda/files/Standarisation.jpg To appear as Fuchs, R. (2015). Speech Rhythm in Varieties of English: Evidence from Educated Indian English and British English. Berlin: Springer, 12-18. 12 can be their ”mother tongue”. English remains their ”auntie tongue”, an important and often cherished and loved relative, but still not a ”mother tongue”.5 In addition to codification work in the strict sense, making explicit suggestions for a canon of IndE (such as Bansal 1976; Nihalani et al. 1979), there has also been a surge in scholarly and primarily descriptive interest in documenting what the emerging IndE consists of, especially in the last two decades. This includes lexical focus marking (Balasubramanian 2009a,b; Fuchs 2012; Lange 2007, 2012; Parviainen 2012; Sedlatschek 2009), topicalisation, dislocation and clefts (Lange 2012), use of determiners (Davydova 2012; Sedlatschek 2009; Sharma 2005b), verb complementation (Hoffmann and Mukherjee 2007; Koch and Bernaisch 2013; Mukherjee 2010; Mukherjee and Hoffmann 2006; Mukherjee and Schilk 2008; Schilk 2011), extension of the progressive (Collins 2008; Davydova 2012; Sharma 2009), use of past tense and present perfect (Davydova 2011; Sharma 2009; Werner 2013), and copula omission (Sharma 2009). Many of these studies relied on data from educated speakers because this is the group that is most likely to make use of the variety (Trudgill 1999; Deumert and Vandenbusche 2003: 459). A focus on educated speakers also informs research on other emerging varieties of English, and the International Corpus of English, consisting of subcorpora of national varieties of English, only samples language produced by speakers with a university degree or who are about to obtain one. As it is the aim of the present study to investigate aspects of the phonology of the emerging IndE, this focus on educated speakers will also be adopted in the following. As the previous paragraph has shown, the last two decades have seen an impressive amount of empirically founded research on the syntax, semantics and pragmatics of educated IndE. However, empirical and acoustic research on the phonetics and phonology of IndE is rare, according to Sailaja (2012: 361): Acoustic studies are minimal in number. Hardly any work on [IndE] can claim to be a comprehensive account of the phonology. Works that are empirical in nature are severely restricted because the number of their informants is restricted. Most findings can only be generalised to a small population. Acoustic works [that have been published] also cannot be taken to be representative of all speakers of that region, because there are several others with the same whose Indian English speech is at variance with theirs. The present study intends to address this scarcity of empirical research in the area by focussing on one oft-quoted aspect of IndE phonology, its alleged syllable-timed rhythm. While most literature on IndE rhythm ”maintains that it is syllable timed”, it is still an ”extremely fuzzy [area] as far as generalisations are concerned, very little work being done” in this area (Sailaja 2012: 361). 2.2 The Phonetics and Phonology of Indian English This section summarises previous results on phonological and phonetic differences between IndE and BrE, starting with segmental differences, followed by intonation and speech rhythm. This will form the basis of tests of their relevance in the perception of differences between IndE and BrE in Chap. 7 below. If IndE and BrE differ in the number and realisation of their phonemes, their intonation and their rhythm, then these differences might also be perceptually salient for listeners when identifying a speaker as British or Indian (accent discrimination experiment, Sect. 7.2 below). Such differences are also likely to make it easier for listeners to attend to one of two simultaneous speakers if one of them is Indian and the other British (cocktail party effect experiment, see Sect. 7.3). How salient each of these phonological categories is probably depends on how great the differences are. If, for example, almost all the phonemes of IndE and BrE are realised in similar ways, then this difference might not be very salient. On the other hand, if many phonemes differ in their realisation, these segmental differences are also likely to be important for the perception of differences between the two accents. Most acoustic studies of IndE have focussed on segmental characteristics, mostly vowels. Less is known about the intonation of IndE, and empirical evidence on speech rhythm is very limited. Starting with Masica (1972) 5 Far from being a recent phenomenon, the competition between progressive and conservative forces promoting or inhibiting the use of English can be traced back through Indian independence and different phases of the colonial administration, both in official documents (Sailaja 2011) and anecdotal evidence on the attitudes of colonial officials (Chaudhuri 1976). 13 and his description of the ”General Indian English” accent, almost all research on IndE phonetics and phonology relied im- or explicitly on data from speakers with Dravidian and Indo-Aryan L1s. The only exception appears to be Wiltshire (2005), who presented acoustic evidence of considerable differences between General IndE phonetics and phonology and that of speakers with a Tibeto-Burman L1. Speakers of Tibeto-Burman and Austroasiatic languages account for only a tiny proportion of the Indian population, as shown in Sect. 2.1.2 above. The following discussion is therefore restricted to the IndE of educated speakers with Dravidian and Indo-Aryan L1s. 2.2.1 Segmental Characteristics 2.2.1.1 Vowels There are several differences between the vowel inventories of IndE and BrE and the phonetic realisation of some vowels, which will be discussed using Wells’s (1982) lexical sets notation. This information is essential for determining the perceptual importance of segmental differences between IndE and BrE in Chap. 7 below. If most vowels differ in their realisation between IndE and BrE, segmental differences are likely to be easily perceived. A merger has been proposed between the NURSE (BrE /3/), STRUT (BrE /2/) and COMMa (BrE /@/) vowels (Bansal and Harrison 1994[1972]: 16–17; Bansal 1978: 25, 1990: 223; Gargesh 2004; Hickey 2004: 544; Sailaja 2009: 24–25). Acoustic evidence (Maxwell and Fletcher 2009) showed that some, but not all speakers have this merger. L1 Hindi speakers, in particular, distinguished the NURSE and COMMa vowels by duration, but not by quality. The L1 Tamil and Gujarati speakers in Wiltshire and Harnsberger’s (2006) acoustic study maintained differences between the NURSE and STRUT vowels (the COMMa vowel not being included in the study), although they were closer to each other in quality than in BrE. There has also been some discussion about the realisation of the NORTH (BrE /O/) and LOT (BrE /6/) vowels, some suggesting a merger (Barron 1963: 100; Bansal and Harrison 1994[1972]: 16–17; Smith 1975: 61; Bansal 1976: 15; Thundy 1976: 30; Bansal 1978: 25; Trudgill and Hannah 2002: 130), and others that both are realised as [6] but with a difference in duration, with NORTH being long and LOT short (Nihalani et al. 1979: 211; Bansal 1990: 223; Sailaja 2009: 25). In an empirical impressionistic study, Sahgal and Agnihotri (1988) analysed recordings of L1 Hindi speakers from Delhi and found that the NORTH vowel is sometimes realised, as in BrE, as [O]. This occurred more often among educated than among less educated speakers, more in reading than in casual style, and more frequently among older and female speakers than younger and male speakers, which prompted the authors to conclude that realisation as [O] is the prestige form in IndE. However, Maxwell and Fletcher’s (2009) acoustic study found no consistent distinction between the NORTH and LOT vowels, both being sometimes realised as [o]. Wiltshire and Harnsberger’s (2006) L1 Tamil and Gujarati speakers produced the two vowels with very small differences in quality. The L1 Tamil speakers also made a duration distinction, but pronounced the LOT vowel longer than the NORTH vowel, contrary to what Nihalani et al. (1979: 211) had suggested. Perhaps the most conspicuous phonetic difference between the vowels of IndE and BrE is that the GOAT and FACE vowels are not realised as diphthongs, as in BrE, but as monophthongs [o] and [e] ([e] as in German ’Fehler’ and French ’aller’, not as in BrE DRESS) (Bansal and Harrison 1994[1972]: 16–17; Bansal 1976: 17, 1990: 223; Trudgill and Hannah 2002: 130; Sailaja 2009: 25, 2012: 360; confirmed in acoustic studies by Maxwell and Fletcher 2009; and Wiltshire and Harnsberger 2006). In addition, the START vowel, realised as [A] in BrE, has been suggested to be pronounced further to the front as [a] in IndE (Bansal 1978: 23, 1990: 223; Nihalani et al. 1979; Wells 1982; 211 Hickey 2004: 545). However, there is acoustic evidence based on data from speakers of IndE with Hindi and Punjabi as L1s (Maxwell and Fletcher 2009), and Gujarati and Tamil as L1s (Wiltshire and Harnsberger 2006), suggesting that an [A] realisation like in BrE is used, at least by speakers with these L1s. The distinction between lax and tense vowels seems to be at best inconsistent in IndE. This is important for the measurement of rhythm because some the metrics that have been suggested, such as the Syllable Ratio and the 14 Table 2.2: The vowels of educated IndE compared to BrE, using Wells’s (1982) lexical set notation. Alternative IndE pronunciations are separated by ∼; BrE pronunciations are only given where they differ from IndE, and are separated by commas where IndE has a merger. KIT DRESS GOAT, NORTH FOOT STRUT, COMM a, NURSE PRICE MOUTH SQUARE IndE BrE I E o @U, O U 2 ∼ 5 ∼ @ 2, @, 3 aI aU E@ FLEECE FACE LOT GOOSE TRAP CHOICE NEAR CURE IndE i e O u æ OI I@ u(@) BrE EI 6 U@ Control/Compensation Index (CCI), rely on this distinction, as do some methods of syllabification (see Sect. 3.2 below for details). Goffin (1934: 27), Datta (1972/73), Bansal and Harrison (1994[1972]: 21–22) and Gargesh (2004: 234–235) suggested that speakers of IndE with L1 Bengali and some varieties of Hindi do not make a distinction between tense and lax vowels. Bansal (1990: 224) stated that IndE speakers do not consistently pronounce tense vowels longer than lax vowels regardless of their L1, and that vowels before voiceless consonants are not shorter than before voiced consonants, unlike in BrE. The tense-lax distinction has phonemic status in BrE as shown by minimal pairs such as /i/ - /I/ and /u/ - /U/. The acoustic correlates of this feature in BrE and AmE include both quality and quantity, lax usually being shorter than tense vowels (Gopal 1990; Shockey 2013; Wells 1962). There is conflicting evidence on whether most speakers of IndE maintain the distinction, and if so what its phonetic correlates are. Kalashnik and Fletcher (2007) and Maxwell and Fletcher (2009) provided acoustic evidence suggesting that IndE speakers with L1 Hindi do not consistently use duration to mark the tense-lax contrast. However, Wiltshire and Harnsberger’s (2006) L1 Gujarati and Telugu speakers did make some length distinctions: The tense vowels belonging to the FLEECE and FACE sets were on average pronounced twice as long as their lax counterparts in the KIT and DRESS sets, but the difference between the durations of the GOOSE - FOOT and START - STRUT vowels was smaller. Together these results suggest that a considerable number of IndE speakers do not mark the tense-lax contrast consistently, particularly since no acoustic studies with speakers of IndE with L1 Bengali have been conducted. Nihalani et al. (1979: 211) stated that the CURE vowel is pronounced as in BrE, as [U@]. However, Gargesh (2004: 234–236) also reported alternative pronunciations, including a monophthong [u], and Maxwell and Fletcher (2010a) presented acoustic evidence that some speakers with L1 Hindi and Punjabi use this variant. Maxwell and Fletcher (2010a) also found evidence of a previously unreported SQUARE - NEAR merger and variable realisation of the PRICE , MOUTH and CHOICE vowels. However, their study was based on only seven speakers, which might not warrant a generalisation of these findings to the larger population of speakers of educated IndE. This, as Sailaja (2012: 361) pointed out, is a general problem with acoustic studies on IndE, which are rare and usually based on a comparatively small number of speakers. However, the sometimes conflicting findings of studies on the phonology of IndE might, as Sirsa and Redford (2013) argued, be due to varying education levels of the speakers. While there might be stronger differences with respect to L1 between speakers of IndE who have attended so-called regionalmedium (or vernacular-medium) schools, Sirsa and Redford (2013) found very small differences between the IndE of graduates of English-medium schools with four different L1s. In total, these studies suggest that IndE has ten monophthongs and six diphthongs (see Table 2.2), although it has to be stressed that some speakers use variable pronunciations. The vowel inventory of educated IndE is therefore smaller than in BrE, where there are eleven monophthongs and eight diphthongs (Deterding 1997; Hawkins and Midgley 2005; Wells 1982). Given these pronounced phonological and phonetic differences between the vowel systems of educated IndE and BrE, it seems likely that they can be easily perceived. Listeners might rely on them 15 Table 2.3: The consonants of educated IndE where they differ from BrE. Pairs of consonants separated by a comma are separate phonemes, and those separated by ∼ are allophones of the same phoneme. BrE IndE v, w V ∼w THIN T ”th COOL , SCHOOL kh ∼ k= k= TOOL , STOOL th ∼ t= t= ∼ ú= ROLL ô ô ∼ R ∼ r ∼ fiô CAR ∅ ∅ ∼ ô ∼ R ∼ r ∼ fiô VILLAGE , WILL BrE IndE l, ë l THUS D d” POOL , SPOOL ph ∼ p= p= DULL d d∼ã LIP, PILL when identifying a speaker as British or Indian (Sect. 7.2 below), and when attending to one of two simultaneous speakers (Sect. 7.3 below). 2.2.1.2 Consonants While the comparison of the vowel inventories of educated IndE and BrE showed that there are several sets of vowels that have merged in IndE, there is only one potential merger when contrasting the IndE with the BrE consonant inventory. This concerns the BrE labio-dental fricative /v/, which is often realised as a labio-dental approximant [V], and also often merged with /w/ (Barron 1963: 102; Sailaja 2012: 360; Bansal 1976: 18, 1978: 39, 1990: 225; Kachru 1994: 515; Gargesh 2004: 238; Hickey 2004: 544). Sailaja (2009: 20) suggested that it is only that have a complete merger, while standard/educated IndE does not have the merger, where /v/ is pronounced as [V], and /w/ as [w]. However, in an empirical study based on data from speakers with L1 Hindi, Sahgal and Agnihotri (1988) found that [w] is used in less than 30 % of all cases, and is associated with little or no prestige. The plosives /t,d/ are sometimes not realised as alveolar (as in BrE), but as retroflex sounds [ú,ã] (Bansal 1978: 39, 1990: 224; Kachru 1994: 514; Trudgill and Hannah 2002: 130; Hickey 2004: 544). Sailaja (2012: 360) stated that alveolar and retroflex realisation is variable both across and within speakers. Education may be a sociolinguistic and voicing a phonological factor that explains some of the variation, with voiceless /t/ more frequently realised as alveolar [t] and voiced /d/ more often as retroflex [ã] (Sailaja 2009: 21–22). An empirical study showed that educated speakers are more likely to use alveolar pronunciations than less educated speakers (Agnihotri and Sahgal 1985; Sahgal and Agnihotri 1988). Variation between speakers may also be explained by differences between Indian languages in how phonologically retroflex consonants are realised, the degree in retroflexion being greater in the south (Barron 1963: 98; Kachru 1994: 514). The BrE dental fricatives /T,D/ are replaced in IndE by dental plosives [t”,d”], of which the voiceless phoneme is often aspirated [t”h ] due to the influence of spelling (Barron 1963: 101; Bansal and Harrison 1994[1972]: 55, 61–62; Sailaja 2009: 21, 2012: 360; Bansal 1976: 17, 1978: 39; Kachru 1994: 514; Gargesh 2004: 238; Hickey 2004: 544).6 In BrE, the voiceless plosives /p,t,k/ have unaspirated allophones /p= ,t= ,k= / in consonant clusters and sometimes at the end of a word, but are strongly aspirated if they occur initially in a stressed syllable and weakly aspirated in most other contexts /ph ,th ,kh / (Gimson 1980: 153–155; Gimson and Cruttenden 1994: 139–142; Kachru 1994: 514–515; Hughes et al. 2013: 42). In IndE, they are always unaspirated (Bansal and Harrison 1994[1972]: 54– 56; Bansal 1976: 17, 1978: 47, 1990: 224; Sailaja 2009: 23, 2012: 360; Gargesh 2004: 237). Another allophonic contrast that is levelled in IndE is the difference between syllable-initial clear-l [l] and syllable-final velarised-l [ë], where only the former is used (Bansal 1990: 224; Kachru 1994: 514; Sailaja 2012: 360; Gargesh 2004; Hickey 2004: 544). 6 In contrast to other authors, Kachru (1994: 514) explicitly included a variant [dh] for /D/. 16 IndE has been described as rhotic (= post-vocalic /r/ is pronounced) by Barron (1963: 99), Bansal (1976: 18, 1978: 39), Thundy (1976: 30), Nihalani et al. (1979: 210–212), Wells (1982: 3:629), Gargesh (2004: 238) and Lange (2009), and as non-rhotic by Trudgill and Hannah (2002: 130) and Sailaja (2009: 19). However, empirical investigations (Agnihotri and Sahgal 1985; Chand 2010; Sharma 2005a) showed rhoticity to be variable across and within subjects, and Bansal (1990: 222) and Sailaja (2009: 19, 2010) suggested that educated or standard IndE is non-rhotic, while many are rhotic. This conclusion was based on Agnihotri and Sahgal (1985) and Sahgal and Agnihotri’s (1988) analysis of 1980s Delhi IndE, where a lower rate of rhotic pronunciations was found among less educated speakers, among men, and in casual speech, compared to a higher rate among women, and in educated and formal speech. Similar results were presented more recently by Wiltshire and Harnsberger (2006: 99–100) and Chand (2010), although Chand’s data also suggested an increase in rhoticity across the board since the 1980s. Regardless of whether /r/ is pronounced in post-vocalic position, there are different phonetic variants. Trudgill and Hannah (2002: 130) mentioned flaps [R], Bansal (1976: 17) approximants [ô] and flaps, Sharma (2005a) approximants and trills [r], and Chand (2010) approximants, trills and flaps. Wiltshire and Harnsberger (2006: 99–100) found, in addition to the previous three variants, a fricativised approximant fiô used by speakers with L1 Tamil. In summary, most differences between the consonants of educated IndE and BrE are phonetic (see Table 2.3). The only phonological differences are variable rhoticity in IndE, and the frequent (but perhaps not universal) merger of /w/ and /v/. Phonetic differences, on the other hand, are substantial. For the perception of differences between the accents of educated IndE and BrE, both phonetic and phonological differences are important. Listeners might be expected to rely on both kinds of differences between IndE and BrE when identifying a speaker as British or Indian (to be tested in Sect. 7.3 below), and when attending to one of two simultaneous speakers (to be tested in Sect. 7.2 below). Considering both differences between the vowels, and between the consonants of IndE and BrE, in normal discourse at least every second phoneme is likely to involve acoustic differences between the two accents. These differences suggests that they might be perceptually relevant. Whether they really are relevant remains to be shown, and will be tested in Chap. 7 below. 2.2.2 Suprasegmental Characteristics 2.2.2.1 Intonation The intonation of educated IndE has been claimed to differ from BrE in how utterances are divided into intonation phrases, in what kinds of pitch accents are commonly used, and the occasional use of a final rising tone (Bansal and Harrison 1994[1972]: 92; Bansal 1976: 21–22, 1978: 68–70; Nair 1996: 63). Bansal (1976) illustrated the use of multiple pitch accents with the following examples, where in BrE a single pitch accent would be used (’ before a syllable indicates that it is accented): 1. a. b. 2. a. b. IndE: ’Come and ’dine with us. BrE: Come and ’dine with us. IndE: ’Do what I ’tell ,you. BrE: Do what I ’tell you. In an acoustic study, Wiltshire and Harnsberger (2006: 101) estimated that speakers with Tamil and Gujarati as L1 used a higher number of pitch accents in intonation phrases than is common in BrE and AmE, and place a pitch accent on most content words. This gave credibility to earlier observations by Gumperz (1982: 121) that ”almost every content word is highlighted” in IndE. However, in the first empirical investigation of how many stressed syllables are accented, Maxwell (2014) found that speakers with L1 Kannada accented 64 %, and speakers with L1 Bengali 47 % of all stressed syllables. This result suggests that while not ”almost every” content word receives a pitch accent in IndE, half or more of all content words do. In a related matter, Moon (2002) found that (in recordings of L1 Telugu and Hindi speakers) maximum fundamental frequency (f0 ) does not differ between focussed and 17 unfocussed (content) words in IndE, whereas differences were found in AmE using the same measure. These findings are compatible with the explanation that most content words receive a pitch accent in IndE, but function words do not. Regarding the question of which tones are used, Maxwell and Fletcher (2010b), and Maxwell (2010) found in an acoustic study that L1 speakers of Kannada and Bengali frequently use rising pitch accents. When expressing narrow focus, the L1 Bengali speakers used L*+H accents. LH accents were also found to be predominant in the English of Urdu speakers from Pakistan (Jabeen 2010). This result might be generalised to Urdu and Hindi speakers in India, considering that Urdu is spoken in both India end neighbouring Pakistan, and the main differences between Urdu and Hindi concern the lexis of educated registers and the script. Whether IndE speakers have a Dravidian or an Indo-Aryan L1 might influence their intonation in IndE. Wiltshire and Harnsberger (2006: 102) found LH accents to be more frequent in the English of L1 Gujarati speakers (56 %) than L1 Tamil speakers (37 %). On the basis of acoustic evidence, Moon (2002) concluded that Hindi L1 speakers tend to use L+H* accents and L1 Telugu speakers H* accents. However, the f0 contours presented in the study rather suggest early-aligned L*+H accents for the L1 Hindi speakers in all cases, and for the L1 Telugu speakers in two of four cases (with a smaller drop in f0 than for the L1 Hindi speakers). In addition, Pickering and Wiltshire (2000) found pitch-accented syllables to have lower f0 than neighbouring unstressed syllables in the speech of L1 Bengali, Tamil and Hindi/Urdu speakers. Together, these studies suggest that LH and L*H accents are characteristic of IndE, and that these are particularly favoured by speakers with an Indo-Aryan L1 (Hindi, Urdu, Gujarati, Bengali in the studies cited above) and somewhat less so by speakers with a Dravidian L1 (Kannada and Tamil in the studies cited). In summary, the differences between the intonation of educated IndE and BrE are substantial. They differ in which words are accented, how utterances are split into intonation phrases and how many pitch accents are assigned to an intonation phrase, and which pitch accents are commonly used. Differences in intonation between IndE and BrE might be perceptually relevant. However, the mere presence of production differences does not necessarily imply that they are perceptually relevant. If they are, the next question is if they are more, less or equally salient as differences between the phonemes of the two accents (to be tested in Chap. 7 below). 2.2.2.2 Speech Rhythm IndE speech rhythm has been described as syllable-timed (Masica 1972: 8; Trudgill and Hannah 2002: 130; Gargesh 2004: 1001; Hickey 2004: 545; Lange 2009) or more syllable-timed than the stress-timed rhythm of BrE (Sailaja 2009: 34, 2010, 2012). This syllable-timed rhythm is often said to be connected to the absence of weak vowels or reduced vowels from IndE (Bansal and Harrison 1994[1972]: 17,85; Bansal 1978: 25,65; Trudgill and Hannah 2002: 130), and diphthongs are not consistently longer than monophthongs (Kachru 1994: 515). In addition, word stress is considered to be non-contrastive in IndE (Bansal 1978: 55). Due to the lower functional load of word stress, stressed vowels might not be lengthened as much as in BrE, contributing to a more syllable-timed rhythm. While the lower functional load of word stress might contribute to a syllable-timed rhythm, other levels of prominence beyond word stress appear to be marked more consistently through lengthening. Maxwell (2014) found that L1 Bengali and Kannada speakers consistently lengthened focussed syllables more than accented syllables, and accented syllables more than stressed syllables. Another factor that might contribute to the more syllable-timed rhythm of IndE is the inconsistent use of intensity to mark different degrees of prominence. Maxwell (2014) found that while speakers with L1 Bengali and Kannada use intensity to mark focus, L1 Kannada speakers did not always mark accented syllables with higher intensity. To the best of the author’s knowledge, three acoustic studies have so far investigated the speech rhythm of IndE. These were not concerned with the question of whether there is a difference in speech rhythm between IndE and BrE, but comparisons with the results of other studies could be interpreted in this way. Subhadra et al. (2009) 18 recorded eleven speakers of IndE with L1 Hindi pronouncing monosyllablic words in a picture naming task, and determined the variability of syllable durations, a measure similar to the rhythm metric VarcoS, to be presented in Sect. 3.2 below.7 They determined a variability of 35 on the VarcoS scale, which is below the variability reported by Rathcke and Smith (2011) for three of four BrE dialects, and as such consisted with a description of IndE as more syllable-timed than BrE. However, measurements based on individual items from a word list, which is a comparatively artificial task, cannot claim to be representative of read or spontaneous speech (which, in fact, was not the authors’ aim). In addition, evidence in favour of a more syllable-timed rhythm in IndE compared to BrE needs to be based on speech in the two varieties produced under similar conditions (see Sect. 3.3 and Chap. 4 below), something that Subhadra et al.’s (2009) results do not allow. The second study (Krivokapić 2013) investigated rhythmic convergence between four speakers of IndE and four speakers of AmE. Using relative measures of the durations of stressed syllables and feet in monosyllabic and multisyllabic feet, Krivokapić compared how much stressed syllables are reduced in duration (”compressed”) in multisyllabic feet compared to monosyllabic feet. The measurements provided evidence that in AmE stressed syllables are more compressed in duration in multi-syllabic feet than in IndE, which is consistent with a description of IndE as more syllable-timed than AmE. There was also some evidence of convergence when IndE and AmE speakers talked with each other, although another possible explanation is that the IndE speakers might have already converged towards a more stress-timed rhythm in general since coming to the United States. The third study (Sirsa and Redford 2013), based on the reading of a story by 7 speakers with L1 Hindi and seven speakers with L1 Telugu who had attended English-medium schools, reported a proportion of vocalic durations over total utterance duration (%V) of 46.8 %. This is higher than the 41.1 % that Grabe and Low (2002) reported for BrE, and thus supports the description of IndE as more syllable-timed than BrE. However, as Sirsa and Redford noted, the speech rhythm of two varieties of English should with values from different studies can be fraught with danger because of a variety of factors that might influence the results. These are best controlled for when similar material is used and segmented and processed according to a common set of criteria. These three studies have provided some evidence of a more syllable-timed rhythm in IndE than in BrE and AmE, but a greater number of speakers, different L1 backgrounds and different speaking styles need to be taken into account to achieve a more reliable description of the rhythm of IndE. Furthermore, with the exception of Sirsa and Redford (2013), the empirical measures of speech rhythm employed in these studies have not been used by other authors in the rhythm literature or differ in other ways from accepted methods, which makes comparisons with other varieties (such as BrE) difficult. These ways of quantifying speech rhythm will be discussed in the following chapter. 7 Subhadra et al. took mean syllable duration divided by the standard deviation of syllable durations. Therefore VarcoS can be derived from their measurements by multiplying them by 100.