Having considered the overall composition of the quotative system in the contact ver- nacular spoken by JNU students, I now examine the results for the variable rule analy- ses of the five most frequent quotative types/variants. As the first step, I report results for the three most frequent traditional quotatives – verbs of reporting, zero quotatives, and verbs of mental activity and perception. I furthermore present analyses for the innovative forms be like and okay (fine).3 In what follows, statistically significant factor groups for each variant are given first. These are then accompanied by the presentation of non-significant effects in square brackets. The main (multivariate) analyses of data
3. In these analyses, factor weights report favouring or disfavouring effects of various factors on the occurrence of the linguistic item under analysis. Effects above .50 indicate a positive association between a specific factor and a variant; effects below .50 indicate a weak associa- tion or a lack of such an association. The main difference between mere percentage values and factor weights is that the latter report the probability of occurrence of a linguistic item when all potential factor effects are assessed simultaneously in the mathematical model.
Free ebooks ==> www.ebook777.comLinguistic change in a multilingual setting 313 are complemented with diachronic contrastive analyses exploring the developmental trajectories of individual variants in real time.
5.1 Verbs of reporting
Table 6 presents results of the multivariate analysis of factors conditioning the occur- rence of verbs of reporting in the HCNVE data.
Table 6. Multivariate analysis of factors contributing to the use of verbs of reporting in HCNVE-India, 2007–20114
Corrected mean .32
Log likelihood –186,909
Total N 113/349
Factor
weight % N
1. Grammatical person
1st .69 57% 41/71
2nd .57 44% 11/25
3rd .41 30% 61/198
Range 28
2. Gender
Male .59 41% 60/146
Female .43 26% 53/203
Range 16
3. Lect
Mesolect .72 54% 19/35
Acrolect .49 30% 77/256
Upper-mesolect .40 29% 17/58
Range 32
4. In the datasets presented in Tables 6 through 10, the information concerning the person of the grammatical subject could not always be elicited. This is particularly true of zero quo- tatives but also of other variants, since IndE is essentially a pro-drop variety allowing for omission of subjects (see Footnote 3). The contexts for which the grammatical subject could not be elicited were removed from the analysis. For this reason the token number adds up to a total of 294 tokens for the grammatical subject in all five analyses presented here. Similarly, one proposition featuring quotative say could not be coded for quote type (People in India would always say, ‘This work will…’ HCNVE: IE52), which is why the token number adds up to a total of 348 tokens for the quote type in all five analyses.
(Continued)
www.ebook777.com
Free ebooks ==> www.ebook777.com
314 Julia Davydova
4. Quote type
Non-lexicalised content [.78] 63% 7/11
Indigenous content [.55] 37% 6/16
Story advancement [.50] 32% 64/194
Thought, attitude [.47] 27% 35/127
5. Year
2011 [.52] 30% 59/193
2007 [.47] 34% 54/156
One linguistic and two non-linguistic factors are implicated in the occurrence of verbs of reporting in the dataset. The grammatical person of the subject is an operative constraint as is gender and the speaker’s lect. The verbs of reporting are favoured by male speakers who received their education in the government ver- nacular schools and learned English primarily in formal contexts as opposed to through spontaneous and more informal interactions. The grammatical environ- ment associated with the use of this quotative type is the 1st person of the subject.
That verbs such as say are favoured with the 1st person of the grammatical subject has been noted for native Englishes (see Tagliamonte & Hudson 1999: 161; D’Arcy 2004: 337), so this result is not very surprising. More interesting is the fact that quote type was not selected as a statistically significant factor here. This is in stark contrast to the native varieties of English spoken in England and North America where quote type is an operative constraint with say (see Tagliamonte & Hudson 1999: 161; D’Arcy 2004: 337). Another notable finding is that year was not selected as a significant constraint in this dataset. I interpret this as an indication that this quotative type represents a stable element in the local system that has not been undergoing any robust recent changes in the JNU community. A different picture, however, emerges once the JNU data is set against the ICE-India data, collected a decade earlier.
Figure 1 reveals a sharp decrease in the use of various verbs of reporting in the time period from 1990 to 2011 (χ2 (1) 19.511, p = 0.000), although this quotative type is still the most frequent strategy to introduce quotation in both data pools. This find- ing is generally consistent with those attested for native-speaker English, for which a conspicuous real-time decline for quotative say, perhaps the most prototypical verb of reporting, has been documented, even though this variant is still leading the system as a major quotative marker, at least in some L1 vernaculars spoken worldwide (see D’Arcy 2012: 356–357; Buchstaller 2014: 153).
Table 6. Multivariate analysis of factors contributing to the use of verbs of reporting in HCNVE-India, 2007–2011 (Continued)
Free ebooks ==> www.ebook777.comLinguistic change in a multilingual setting 315
0%
10%
20%
30%
40%
50%
60%
70%
80%
1990s 2007 2011
Figure 1. Relative frequencies of verbs of reporting in ICE-India (N 289/396), 1990–1996, and in HCNVE-India (N 113/349), 2007–2011
5.2 Zero quotatives
Table 7 provides a multivariate analysis of the contribution of various factors to the probability of zero quotatives in HCNVE 2007–2011.
Table 7. Multivariate analysis of factors contributing to the use of zero quotatives in HCNVE-India, 2007–2011
Corrected mean .18
Log likelihood –58,624
Total N 64/349
Factor weight % N
1. Grammatical person
1st .52 4% 3/71
3rd .49 4% 8/198
2nd .47 4% 1/25
Range 5
2. Lect
Mesolect .76 25% 9/35
Upper-mesolect .70 18% 11/58
Acrolect .41 17% 44/256
Range 35
3. Gender
Female [.52] 18% 37/203
Male [.48] 18% 27/146
(Continued)
www.ebook777.com
Free ebooks ==> www.ebook777.com
316 Julia Davydova
4. Quote type
Non-lexicalised content [.68] 18% 2/11
Story advancement [.51] 19% 37/194
Thought, attitude [.50] 18% 23/127
Indigenous content [.32] 12% 2/16
5. Year
2007 [.58] 21% 33/156
2011 [.44] 16% 31/193
The multivariate analysis reveals that one internal and one external factor affects quotative choice: the grammatical person of the subject and lect. In contrast to verbs of reporting and verbs of mental activity and perception, zero quotatives do not exhibit any clear preference for a particular grammatical person of the subject. The effect is therefore diffused, which is in turn reflected in the relative impact of this factor group (range 5). Notice also that quote type is not a significant constraint on the use of zero quotatives. Moreover, speaker gender does not contribute to the use of zero forms. The latter finding is consistent with that reported in Tagliamonte & Hudson (1999: 160) for native varieties of BrE and CanE.
Similar to verbs of reporting, zero quotatives are favoured in the speech of mesolectal speakers. Reportedly, these speakers do not have any exposure to informal, casual English during their secondary school years. Together with say and think, zero quotative markers are quite old forms, having been attested in written records of L1 English that date back to the mid-19th century (see, for instance, D’Arcy 2012). It is then perhaps not surprising that these forms are attested and preferred in the speech of those speakers of IndE who have learnt English primarily through formal instruc- tion, i.e. books.
What is also of interest is that the factor ‘year’ was not selected as a statistically sig- nificant factor during the binominal step-up / step-down procedure, although it looks like zero quotatives have somewhat declined (from 21% in 2007 to 16% in 2011) in real time. This finding is generally consistent with what has been reported for zero quota- tives in accounts of native-speaker English. For instance, D’Arcy (2012: 362) comments that “the null form […] display[s] vacillating waves of use, coming in and out of fash- ion”. In order to show how this is the case in IndE, I extend the scope of comparison to the data from ICE-India.
Figure 2 reveals that zero quotatives indeed increase (χ2 (1): 6.169, p = 0.013) and then decrease its frequency of use in real time. This finding thus confirms the sociolinguistic hypothesis stating that zero forms belong to the cohort of “recycled”
Table 7. Multivariate analysis of factors contributing to the use of zero quotatives in HCNVE-India, 2007–2011 (Continued)
Free ebooks ==> www.ebook777.comLinguistic change in a multilingual setting 317
variants, i.e. linguistic forms that exist latently in the repertoire of a speech commu- nity (Buchstaller 2006: 19, D’Arcy 2012: 362). In other words, these forms are believed to be part of the linguistic reservoir that speakers may resort to in an attempt to con- struct a specific, if only transient, sociolinguistic meaning, and attitudinal evidence is needed in order to ascertain what this meaning is exactly. In a way, this fluctuating diachronic trajectory is also a testimony to the dynamic and fleeting nature of an evolving system.
5.3 Verbs of mental activity and perception
The multivariate analysis of verbs of mental activity and perception reveals similarities with verbs of reporting, as shown in Table 8. The factor group ‘year’ was excluded from the analysis because the tokens of this factor group were skewed across the categories
‘gender’, ‘grammatical person’ as well as ‘lect’. This will be discussed in a separate distri- butional analysis shown in Figure 3.
Both grammatical person and gender are statistically significant constraints underlying the occurrence of verbs of mental activity and perception. Male speakers favour the use of this type of quotatives. In so doing, they associate these quotatives with the 2nd person of the grammatical subject at .82. In contrast, this quotative type is disfavoured with the 1st person at .43. Interestingly enough, the 1st person is favoured with think, a canonical verb of mental activity, in both BrE and CanE, as reported in Tagliamonte & Hudson (1999: 161). While the variable constraint of the grammati- cal subject was transferred into the indigenised variety (since it exerts a significant
0%
10%
20%
30%
40%
50%
1990s 2007 2011
Figure 2. Relative frequencies of zero quotatives in ICE-India (N 46/396), 1990–1996, and in HCNVE-India (N 64/349), 2007–2011
www.ebook777.com
Free ebooks ==> www.ebook777.com
318 Julia Davydova
Table 8. Multivariate analysis of factors contributing to the use of verbs of mental activity and perception in HCNVE-India, 2007–20121, 2
Corrected mean .07
Log likelihood –76,673
Total N 24/349
Factor weight % N
1. Grammatical person
2nd .82 28% 7/25
3rd .48 6% 13/198
1st .43 5% 4/71
Range 39
2. Gender
Male .67 11% 17/146
Female .37 3% 7/203
Range 30
3. Lect
Acrolect [.55] 7% 18/256
Mesolect [.38]5 6% 6/93
Range
4. Type of quote
Story advancement [.53] 7% 15/195
Thought, attitude, non-lexicalised content
[.47] 5% 8/137
Indigenous content no data6 0% 1/16
5. Year
2011 no data
2007 no data
5. Upper-mesolectal speakers did not produce any tokens of this quotative type and are thus not featured in the analysis.
6. Verbs of mental activity and perception do not occur with indigenous content in the data studied here. The factor was therefore removed from the multivariate analysis. Quotes expressing thought and attitudes were merged with those signalling non-lexicalised content because the latter usually renders expressive meaning.
Free ebooks ==> www.ebook777.comLinguistic change in a multilingual setting 319 effect on the occurrence of the quotative type), it has also developed a different pattern of use. Notice also that similar to verbs of reporting and zero quotatives, quote type does not have any discernible effect on the occurrence of verbs of mental activity and perception. This is in opposition to what has been reported for English in Britain and North America, where quote content has been shown to be implicated in the occur- rence of the verb of mental activity think in BrE and CanE (Tagliamonte & Hudson 1999: 163).
But how have verbs of mental activity and perception evolved in the quotative system of IndE over the last two decades? Figure 3 reports the contribution of this quotative type to the overall composition of quotation marking in the 1990s and late 2000s.
0%
10%
20%
30%
40%
50%
1990s 2007 2011
Figure 3. Relative frequencies of verbs of mental activity and perception in ICE-India (N 14/396), 1990–1996, and in HCNVE-India (N 24/349), 2007–2011
Although occurring at a much lower rate than zero quotatives, the developmen- tal trajectory of verbs of mental activity and perception reveals an essentially similar picture. Although more frequent in the late 2000s (this effect is weakly significant, χ2 (1): 4.018, p = 0.045), this quotative type shows a vacillating pattern. Verbs of mental activity and perception thus fluctuate in real time, serving as an alternative resource to introduce quoted material in the discourse.
5.4 Be like
Table 9 reports the results of the multivariate analysis of the contribution of indepen- dent factors to the probability of use of be like in HCNVE 2007–2011.
www.ebook777.com
Free ebooks ==> www.ebook777.com
320 Julia Davydova
Table 9. Multivariate analysis of factors contributing to the use of be like in HCNVE-India, 2007–2011
Corrected mean .16
Log likelihood –137,065
Total N 55/349
Factor weight % N
1. Grammatical person
1st .54 22% 16/71
3rd .53 18% 37/198
2nd .17 4% 1/25
Range 37
2. Gender
Female .61 19% 39/203
Male .34 10% 16/146
Range 27
3. Year
2007 .63 19% 30/156
2011 .38 12% 25/193
Range 25
4. Type of quote
Indigenous content [.59] 25% 4/16
Thought, attitude [.55] 19% 25/127
Story advancement [.46] 12% 25/194
Non-lexicalised sound [.27] 9% 1/11
5. Lect
Acrolect [.52] 18% 47/256
Upper-mesolect [.49] 10% 6/58
Mesolect [.36] 5% 2/35
Three factors are implicated in the occurrence of be like: grammatical person of the subject, gender, and year. In this dataset, be like is associated with both 1st per- son and 3rd person contexts, which is an indication that the variant has begun mak- ing inroads into the quotative system through these linguistic environments. In other words, JNU students effectively use the be like variant to quote themselves and others, as in:
(27) I was like, ‘Oh God, some day I will be there as an interpreter’.
(HCNVE: IE14)
Free ebooks ==> www.ebook777.comLinguistic change in a multilingual setting 321 (28) He is like, ‘What is PhD? I don’t understand, okay. What is PhD?’
(HCNVE: IE35) Whenever be like is used with a 3rd person grammatical subject, it is equally used in both referential and existential contexts (18 tokens vs. 19 tokens, total N 37; χ2 (1):
3.044, p = 0.81), as in
(29) […] we would be coming in a car and it was like, ‘I’ve never seen such a huge university!’ (HCNVE: IE21) (existential context) (30) […] and she was like, ‘Oh my God, everybody is so dressed up!’
(HCNVE: IE04) (referential context)75 Note that quote type is not implicated in the occurrence of be like in the IndE dataset reported here. This finding is striking given that quote type is a consistent predictor for the use of be like in native Englishes (Tagliamonte & Hudson 1999; Cukor-Avila 2002;
Tagliamonte & D’Arcy 2004).
Interestingly, quote type is not a significant factor in Irish English either, as reported in Hửhn (2012: 285). Drawing on data obtained from the Irish component of the ICE corpus (ICE-IRE), Hửhn’s study reveals a statistically significant contribution by the grammatical person of the subject but not by quote type. The emergence of Irish English as a distinct variety involved high levels of adult second-language acquisi- tion in the 19th century (Siemund et al. 2012: 35). Irish English is therefore clearly a second-language variety of English from an historical perspective. IndE, currently a second-language variety, replicates exactly this pattern. Together, these findings raise the question of what is really possible in second-language acquisition, which will be taken up this issue in Section 6.
Notice also that female students clearly favour the use of be like in my data. This is consistent with the sociolinguistic findings on be like reported for native-speaker Englishes in the current literature (cf. Tagliamonte & Hudson 1999: 160; Macaulay 2001: 14; Buchstaller & D’Arcy 2009: 308; Buchstaller 2011: 71; Durham et al. 2012).
Given a more general sociolinguistic claim that women do not “only engage in change but advance it as well” (D’Arcy 2004: 324), it is not unreasonable to suggest that female students may have been responsible for the spread of be like in the Indian community studied here.
Though the factor group ‘lect’ was not selected as a significant factor during the binominal step-up/step-down procedure, a close examination of the distribution of be like across the three speaker cohorts reveal some noteworthy patterns reported in
7. Existential contexts are contexts featuring the inanimate pronoun it as a subject; refer- ential contexts are the contexts that feature the animate pronouns I/we/you/he/she/they or animate noun phrases as a subject.
www.ebook777.com
Free ebooks ==> www.ebook777.com
322 Julia Davydova
Table 9. Be like is most frequently used by acrolectal speakers (18%) followed by upper- mesolectal (10%) and, finally, by mesolectal speakers (5%). Thus, speakers who have most exposure to English in a casual setting and speakers with frequent and regular exposure to (American) native-English mass media, are also the speakers who produce the most tokens. The differences between acrolectal and upper-mesolectal/mesolectal speakers are statistically significant (χ2 (1): 5.411, p = 0.020).
All in all, it appears that be like has undergone linguistic restructuring in the JNU community. Unlike in native-speaker varieties, its use is not constrained by quote type.
Grammatical subject is the only linguistic effect implicated in the occurrence of be like. Furthermore, be like exhibits a propensity to occur in speech of acrolectal female speakers.
Finally, the factor group ‘year’, a statistically significant constraint, shows a trend that the use of the global innovative form be like has dropped in the community. The year 2007 maintains a strong effect on the probability that be like is used (.63) but in 2011 this probability decreases to .38. This finding is somewhat unexpected given robust expansion of be like in native-speaker English since the 1990s (see Table 5). To be able to understand why be like has declined in this dataset over time, we need to consider this variant against the backdrop of the results obtained for the innovative local variant okay (fine), to the discussion of which I will now turn. The diachronic comparisons for both variants are presented in Figure 4.
5.5 Okay (fine)
In the description of the overall distribution of quotative variants, okay (fine) was presented as a superstrate-based local innovation, as illustrated by:
(31) […] he okay, ‘What are you doing these days?’ (HCNVE: IE35) (32) […] it’s nothing that okay (,) fine, ‘This person he doesn’t know English,
then there is something, something that he lacks’. (HCNVE: IE35) (33) […] they have set certain limits (,) okay, ‘This is the grade that you’re
supposed to get, if you get it, we don’t have any problems. How you get it, except for cheating in the exams, how you get it, and how you manage your time and stuff, that’s all your business, you chat all day long, but still you get
a decent grade, that’s okay’. (HCNVE: IE02)
In this case, IndE speakers employ the two English words okay and fine, and imbue them with a new discourse-pragmatic meaning. The analysis suggests that these two items okay and fine have undergone specialisation of function. This process is indica- tive of the ongoing grammaticalisation of okay (fine). This contention is bolstered by the results of the multivariate analysis, shown in Table 10.
Free ebooks ==> www.ebook777.comLinguistic change in a multilingual setting 323 Table 10. Multivariate analysis of factors contributing to the use
of okay ( fine) in HCNVE-India, 2007–2011*
Corrected mean .08
Log likelihood –83,821
Total N 27/349
Factor weight % N
1. Grammatical person
3rd .55 12% 24/198
1st .34 4% 3/71
2nd no data 0% 0/25
Range 21
2. Year
2011 .66 11% 23/193
2007 .30 2% 4/156
Range 36
3. Gender
Male [.51] 4% 7/146
Female [.48] 9% 20/203
4. Quote type
Indigenous content no data 0% 0/16
Story advancement [.56] 10% 20/194
Thought, attitude, non-
lexicalised content [.34] 5% 7/138
5. Lect
Acrolect [.55] 8% 21/256
Upper-mesolect [.50] 8% 5/58
Mesolect [.18] 2% 1/35
* ‘No data’ indicate an empty cell. The factor is thus removed from the multivariate analysis.
‘Year’ is a significant constraint underlying the use of okay (fine) in the data: it is an incipient variant in 2007, while in 2011 speakers are more likely to use it as a quota- tive. I interpret this as a sign of the further grammaticalisation of the lexemes okay and fine into a quotative marker.
This finding furthermore suggests that in contrast to the global innovative quota- tive marker be like, the local variant okay (fine) has been increasingly gaining ground in the community over the last few years. This contention is substantiated by the results of the chi-square test: there is a significant difference in the use of be like and
www.ebook777.com