3.1 Introduction

Recent variationist studies of language use in London have observed a range of linguistic innovations at different levels of the grammar, including among others: radical diphthong changes in face, price, goat and mouth (Cheshire et al. Reference Cheshire, Fox, Kerswill and Torgersen2008, Reference Cheshire, Kerswill, Fox and Torgersen2011; Fox Reference Fox, Llamas and Watt2010, Reference Fox2015); the development of topic-marking who (Cheshire et al. Reference Cheshire, Adger and Fox2013); the recruitment of man as a pronoun (Cheshire Reference Cheshire2013); the acquisition of the new quotative variant this is + speaker (Cheshire et al. Reference Cheshire, Fox, Kerswill and Torgersen2008, Reference Cheshire, Kerswill, Fox and Torgersen2011; Fox Reference Fox, Buchstaller and van Alphen2012); and new additions to the pool of response elicitor variants such as (do/if) you get what I mean/I’m saying and you get me (Andersen Chapter 1; Torgersen et al. Reference Torgersen, Gabrielatos, Hoffmann and Fox2011). The emergence and spread of these innovations has been attributed to the complex sociolinguistic landscape of inner-city London, notably the effects of group second language acquisition and multiple language contact as well as the widespread existence of multi-ethnic friendship groups (Cheshire et al. Reference Cheshire, Fox, Kerswill and Torgersen2008, Reference Cheshire, Kerswill, Fox and Torgersen2011). Together, the aforementioned innovations constitute core features of Multicultural London English (MLE), an ethnically-neutral variable repertoire of language forms (Cheshire et al. Reference Cheshire, Kerswill, Fox and Torgersen2011). This chapter continues the investigation of linguistic innovations in MLE with an analysis of reduced negative-polarity interrogative tags (neg-tags), specifically innit, the variant at the forefront of the innovations described in this chapter.

Though by no means a new feature of London or British English (see, inter alia, Andersen Reference Andersen2001; Cheshire et al. Reference Cheshire2005; Erman Reference Erman and Haukioja1998; Krug Reference Krug1998; Pichler Reference Pichler2013 for previous accounts of innit usage across varieties of British English), innit is rapidly innovating in contemporary London English. Andersen’s (Reference Andersen2001) analysis of innit in COLT, the Bergen Corpus of London Teenage Language recorded in the mid-1990s, has shown that innit is not limited in use to the syntactic-semantic context and discourse-pragmatic functions associated with its claimed source form isn’t it; its sociolinguistic distribution in COLT and a comparable adult corpus was shown to be consistent with a change in progress whereby innit is gradually levelling across the inflectional paradigm and developing new discourse-pragmatic functions. Pichler’s (Reference Pichler2014) and Pichler and Torgersen’s (Reference Pichler and Torgersen2013) more recent analyses of the canon of neg-tag variants in LIC, the Linguistic Innovators Corpus recorded in London in the mid-2000s (see further Section 3.2), reveal ongoing, rapid changes in the use of innit; clause-final innit continues to level across the paradigm and, together with a small number of other neg-tag variants, is developing new functions not reported in Andersen (Reference Andersen2001) or elsewhere. Moreover, in LIC, the occurrence of innit and, to a far lesser extent, that of other neg-tag variants is not restricted to the canonical position of neg-tags. As shown in the examples in (1), it extends to positions beyond the right periphery (RP) of main clause anchors and beyond the ‘follow-up’ position described in Andersen (Reference Andersen2001) (see further Section 3.3).

(1)

1. a.

   That’s how it was. Now, if you got beef, innit, take it to the yard.

   (Alan, 16-F-N-Hk)Footnote ¹

2. b.

   Whatever, innit.

   (Kelly, 16-F-N-Hk)

3. c.

   I dunno what’s gonna happen to her if she’s not back. Innit, it shouldn’t be like this.

   (Tina, 18-F-N-Hk)

4. d.

   Yeah London, innit. Like London ain’t London no more.

   (Gary, 17-M-A-Hv)

5. e.

   Cos down these ends, innit, like, ‘Yeah, what are you on?’

   (Rufus, 19-M-N-Hv)

6. f.

   I mean, the sister, innit, she’s about five times bigger than you, innit, Mark.

   (Tina, 18-F-N-Hk)

7. g.

   People around, innit, just say who’s won.

   (Gary, 17-M-A-Hk)

In what follows, I will examine in detail the occurrence of innit and, where applicable, other neg-tag variants in the non-canonical positions illustrated in (1), with a view to: (i) continuing Andersen’s (Chapter 1) discussion of methodological imperatives for identifying discourse-pragmatic innovations at an early stage of their development; (ii) supporting Waters’s (Chapter 2) call for an empirically- and theoretically-grounded but flexible approach to defining discourse-pragmatic variables and their variable contexts. Unlike Andersen (Chapter 1) who is concerned with identifying innovative discourse forms, the focus of this chapter is on capturing innovative uses of well-established forms. The neg-tag variable at the centre of the analysis and its variants, including innit, are introduced together with the data in Section 3.2. Section 3.3 outlines the range of positions neg-tags and in particular the variant innit occupy in the LIC data. The outline demonstrates that attention to neg-tag variants’ positional distribution and scopal domain is crucial to identifying potential innovations in their use. The following description in Section 3.4 of the discourse functionality of neg-tags in non-canonical positions shows that qualitative data analysis is a compelling means to gauge the status of previously unreported neg-tag uses as an interactionally-motivated practice rather than a series of random performance errors. Section 3.5 provides real- and apparent-time evidence to support the hypothesis that the innit uses illustrated in (1c)–(1f) represent a discourse-pragmatic innovation at an early stage of its development. The discussion in Section 3.6 begins with a comparison of the results presented in Sections 3.3–3.5 with those presented in two previous studies of innit in the same dataset (Palacios Martínez Reference Palacios Martínez2015; Torgersen et al. Reference Torgersen, Gabrielatos, Hoffmann and Fox2011). I will argue that these studies failed to uncover the most dramatic recent innovation in innit (or neg-tag) use because they paid insufficient attention to the methodological imperatives outlined in this chapter. The discussion then moves on to underscoring the value of Waters’s (Chapter 2) call for bespoke variationist analyses of discourse-pragmatic features. I will argue that rigid adherence to previously postulated definitions of the neg-tag variable and variable context would have hampered, if not altogether foiled, attempts to identify ongoing changes in neg-tag use; it would also complicate attempts to situate these innovations in relation to the larger linguistic sub-system of which they become a part. Section 3.7 concludes the chapter.

3.2 Data

The analysis is based on the 1.4-million-word Linguistic Innovators Corpus (LIC) collected in 2005–2006 for a large-scale research project investigating linguistic innovation and change in contemporary London English (Kerswill et al. Reference Kerswill, Cheshire, Fox and Torgersen2007). The corpus comprises recordings of ninety-eight adolescent and sixteen elderly speakers from: Hackney, a multi-ethnic and multi-cultural inner-city borough of London; and Havering, a less diverse and more affluent outer-city borough of London. In addition to locality, the adolescent speaker sample is stratified by ethnicity, with speakers broadly divided into Anglo and non-Anglo, i.e., a relatively homogeneous group of speakers of white British heritage and a more heterogeneous group of speakers from ethnic minority backgrounds. All older speakers are white British. Adolescent recordings were made in local community colleges with friendship groups of two or three speakers; older speakers were recorded in their homes, usually in pairs. In both contexts, fieldworker input was minimal and, as far as possible, interactions were left to develop naturally.

The analysis presented here developed from an investigation by Pichler and Torgersen (Reference Pichler and Torgersen2013) into the syntactic-semantic distribution in LIC of the canon of neg-tags which they defined – on the basis of their prototypical position and shared derivation – as the set of constructions ‘appended to main clause anchors and derived from the following linear string of components: (auxiliary) + (negative clitic) + (pronominal subject)’ (see also Pichler Reference Pichler2013: 179).Footnote ² In addition to the clause-final neg-tag tokens relevant to Pichler and Torgersen’s analysis of paradigm levelling, the semi-automated data extraction of neg-tags from LIC yielded tokens that, though derived from the linear string of components schematised earlier, fall outside the variable context; they are not appended to main clause anchors but occur in the positions illustrated in (1) which have not traditionally been associated with neg-tags. The occurrence of these tokens in LIC must not be ignored since it may be indicative of ongoing change (see Himmelmann’s [Reference Himmelmann, Bisang, Himmelmann and Wiemer2004: 32] ‘syntactic context expansion’). However, testing this hypothesis requires ‘pushing the envelope’ of variation (Aaron Reference Aaron2010) and relaxing the positional criteria previously used for defining the neg-tag variable and variable context. I therefore include in the present analysis all constructions derived from the string (auxiliary) + (negative clitic) + (pronominal subject) irrespective of their position of occurrence. Inclusion in the analysis of all neg-tag tokens in LIC makes it possible to identify the locus of any innovations in their use and to situate any such innovations in the context of established neg-tag uses. Hereafter, neg-tags that are appended to main clause anchors will be referred to as neg-tags in canonical position; turn-initial neg-tags with scope over a previous speaker’s turn will be referred to as follow-up neg-tags; all other neg-tags will be referred to as neg-tags in non-canonical positions. To establish whether it is only innit or indeed a wider range of neg-tag variants that are undergoing change, the variant inventory will be broadly divided into the following categories: canonical neg-tags such as doesn’t she, can’t you, haven’t we; canonically-derived and phonetically-reduced neg-tags such as ain’t it, in’t they, dunnit; and the variant form innit which, according to Andersen (Reference Andersen2001: 197), derived through phonetic attrition from the canonical tag variant isn’t it. At times, neg-tag variants from the latter two categories will be jointly subsumed under the label phonetically-reduced neg-tags.

3.3 Positional and scopal properties of MLE neg-tags

English neg-tags typically occur at the right clausal periphery and typically take wide leftward scope over their preceding anchor clause, as shown in the examples from LIC in (2). Although Ziv (Reference Ziv, Bolkestein, Groot and Mackenzie1985: 198) argues that neg-tags are ‘restricted to sentence-final position’, neg-tags are also, on occasion, parenthetically inserted in clause-medial post-finite position, as illustrated in (3). When they occur either between or within syntactic constituents, medial neg-tags may take narrow scope over the immediately preceding phrasal constituent (Cullicover Reference Cullicover1992: 206; Dehé and Braun Reference Dehé and Braun2013: 136; McGregor Reference McGregor, Hasan and Fries1995: 110; Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 811). Moreover, neg-tags may be appended to elliptical anchors whose ellipted subject and finite verb can be fairly straightforwardly inferred from the surrounding linguistic context. For example in (4), the words in strikethrough font were ellipted but are recoverable from the shared co-text to produce complete propositions. Beyond the limited flexibility illustrated in (2)–(4), however, English neg-tag usage is generally characterised as lacking positional and scopal variability. There is strong consensus in the literature that the occurrence of neg-tags is restricted to post-finite position and that their scope always extends leftwards over the preceding clause or, occasionally, over one of its phrasal constituents (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 208; Knowles Reference Knowles1980: 382).

(2)

1. a.

   You have to have nice shoes, don’t you.

   (Lewis, 16-M-A-Hv)

2. b.

   He was gonna post it to him, weren’t he.

   (Mark, 18-M-N-Hk)

3. c.

   And you already started it at the beginning of the year, din you.

   (Jake, 16-M-A-Hk)

4. d.

   You take it out to the street, innit.

   (Alan, 16-M-N-Hk)

(3)

1. a.

   I mean, I’m a glutton, aren’t I, for historical programmes on the television.

   (Elsa, >65-F-A-Hv)

2. b.

   We had a book, didn’t we, once on rhyming slang.

   (Joe, >65-M-A-Hk)

3. c.

   Cos the boys made carts, didn’t they, out of wood.

   (Joan, >65-M-A-Hk)

4. d.

   It’s a sin, innit, in front of God.

   (Bisa, 17-F-N-Hv)

(4)

1. a.

   How old is she? She’s two years older than me, in’t she. She’s seventy-two now, in’t she.

   (Frances, >65-F-A-Hk)

2. b.

   It’s not this week. It’s next week, innit.

   (Jake, 16-M-A-Hk)

3. c.

   SF: What have you done?

   Rufus: I dunno. I’ve done bare things, innit.

   (Rufus, 19-M-N-Hv)

Although about 93.9% (N = 2501/2663) of all positionally and scopally unambiguous neg-tag tokens in LIC occupy the canonical neg-tag positions illustrated in (2)–(4),Footnote ³ the LIC data do not fully support the positional and scopal restrictions in neg-tag use postulated in the literature. Firstly, we find in the LIC data neg-tag tokens that fit Andersen’s (Reference Andersen2001: 139–50) description of follow-ups (see also Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 812). As shown in (5), these neg-tags occur turn-initially, constitute stand-alone utterances, and scope not over the current but over the previous speaker’s proposition. They generally signal either surprise at or alignment with previous speakers’ propositions. According to Andersen (Reference Andersen2001: 204–5), follow-up uses of innit have developed from innit uses in canonical position.

(5)

1. a.

   Isabella: Yeah, she weren’t meant to go out.

   Bisa: Weren’t she!

   (Isabella, 17-F-N-Hv; Bisa, 17-F-N-Hv)

2. b.

   Sophie: Shayne. He’s got to win!

   Charlotte: In he!

   (Sophie, 16-F-A-Hv; Charlotte, 16-F-N-Hv)

3. c.

   Maria: It’s like me and my sister go back to kids again.

   Emily: Innit! It’s like Christmas day, it’s like your child- childhood.

   (Maria, 16-F-A-Hv; Emily, 16-F-A-Hv)

In addition, the LIC data contain innit tokens that occur in positions previously only associated with invariant lexical tags, i.e., tags such as yeah and right which, unlike innit, are not derived from the linear string of components characterising neg-tags (see Section 3.2). In (6), for example, innit occurs after and has scope over a subordinate (rather than main) clause, a position shared by yeah in Welsh English (Jones Reference Jones1990: 187–8). In (7), innit occurs after the formulaic construction I know which is used to signal agreement with the preceding speaker’s proposition. While the use of the lexical invariant tag right sounds grammatical when I know is used as a non-compositional agreement marker in this context (I know, right; see also Stenström et al. Reference Stenström, Andersen and Hasund2002: 176), the use of canonical neg-tag variants would seem grammatical only when I know is used compositionally to signal familiarity with a following NP (e.g., I know her mother, don’t I.).

1. (6)

   Like, when you’re going yard, innit, it’s like you don’t wanna say nothing.

   (Alex, 16-M-N-Hk)

2. (7)

   Katie: Every time her phone rings, ‘Hello mum.’

   Laura: I know, innit.

   (Katie, 18-F-A-Hk; Laura, 18-F-A-Hk)

Moreover, close auditory analysis of every neg-tag token in LIC revealed that some phonetically-reduced tokens are positioned at the left periphery (LP) of the clause that they scope over. As crudely indicated by the typographical means employed to replicate prosodic features from the audio files (‘:’ for syllable lengthening; ‘(.)’ for a short pause; ‘<text>’ for increase in speech tempo; ‘,’ for continuing intonation contour; ‘.’ for final intonation contour), the LP innit tokens in the examples in (8) are preceded rather than followed by a marked tone unit boundary (see Cruttenden Reference Cruttenden1997: 33–5). Thus, because they are prosodically bound to the right, they take rightward scope over the following proposition. (This is in contrast to the RP tokens in (2) which are prosodically bound to the left and hence take leftward scope over preceding propositions.) The occurrence in LIC of innit and other variants at the LP contradicts the syntactic descriptions of neg-tags in the literature reviewed at the beginning of this section, and it refutes Izutsu and Izutsu’s (Reference Izutsu, Izutsu, Degand, Cornillie and Pietrandrea2013: 230) recent claim that innit does ‘not tolerate the initial position’.

(8)

1. a.

   I’ve e:h <innit, they’re> supposed to give you a fine or something. Like, eighty pound or a hundred pound or something.

   (Ahmed, 19-M-N-Hk)

2. b.

   Cos they’re such pricks. Overreacting idiots. Innit, that’s a waste of police time. Innit, they should be out catching real criminals.

   (Tina, 18-F-N-Hk)

Finally, the LIC data contain (largely phonetically-reduced) neg-tags which are appended to and take narrow scope over: lone noun phrases (NPs) or lone prepositional phrases (PPs), as in (9); left-dislocated (LD) NPs, as in (10); and subject NPs, as in (11). These neg-tag uses are different from those in (4) where the NPs are constituents of elliptical anchors whose missing elements can be recovered from the surrounding context. In (9), the anchor of innit is an independent phrasal unit that cannot easily be extended to a full clause; in (10), the anchor of in he is a detached lexical constituent of the clause following in he; in (11), the anchor of innit is the subject NP of the clause in which innit is parenthetically inserted. Thus, because the neg-tags illustrated in (9)–(11) occur after lone NPs/PPs, between LD NPs and their co-referential pronouns, or between subject NPs and verb phrases (VPs), they categorically have phrasal rather than clausal scope. They will henceforth be referred to as ‘phrasal neg-tags’. The positions between LD NPs and their co-referential pronouns and between subject NPs and VPs have previously been associated with the use of the invariant lexical tags yeah and right (Jones Reference Jones1990: 188; Stenström et al. Reference Stenström, Andersen and Hasund2002: Ch. 7). The occurrence in LIC of innit (and other phonetically-reduced neg-tag variants) after lone NPs/PPs refutes Stenström et al.’s (Reference Stenström, Andersen and Hasund2002: 173) proposal that innit – due to its assumed derivation from isn’t it – ‘always has to follow material that constitutes a complete [clause] […] or that can be enriched to a [clause]’.

(9)

1. a.

   And his head though as well, innit.

   (Dale, 17-M-A-Hv)

2. b.

   What about the other time? Down {road name}, innit, where two woman drivers pulled out (?), didn’t they?

   (Dale, 17-M-A-Hv)

1. (10)

   And now this step dad, in he, he’s like proper strict.

   (Lewis, 16-M-A-Hv)

1. (11)

   So obviously one person, innit, was a bit weak, yeah.

   (Rufus, 19-M-N-Hv)

The preceding outline of the range of positions occupied by neg-tags in LIC reveals that the use of at least some neg-tag variants extends beyond the canonical position of neg-tags at the RP of main clause anchors or the follow-up position illustrated in (5). As shown in the examples from LIC in (6)–(11), some neg-tag variants occur in positions not previously reported for neg-tags in the literature, i.e., after subordinate clauses and formulaic constructions; at the LP of clausal anchors; after lone NPs/PPs; and after LD or subject NPs. The frequency of neg-tags in these non-canonical positions is relatively low; they account for only 2.8% (N = 74/2663) of all positionally and scopally unambiguous neg-tags in LIC.Footnote ⁴ Moreover, almost 85.1% (N = 63/74) of all neg-tags in non-canonical positions have the form innit (see further Section 3.5). Yet despite their relative infrequency and bias towards one variant form, the neg-tags in non-canonical positions deserve closer investigation since their occurrence may be a reflex of ‘syntactic context expansion’ whereby constructions come to be used in positions where they could not be used previously (Himmelmann Reference Himmelmann, Bisang, Himmelmann and Wiemer2004: 32). Before I present real- and apparent-time evidence to test the hypothesis that what we are witnessing here is an incipient stage of a language change, it is important to examine the functionality of neg-tags in non-canonical positions. Qualitative analysis of these tokens will establish whether they are random and meaningless insertions in discourse or whether they are systematically used to perform functions compatible with neg-tags’ core function of seeking corroboration.

3.4 Functional properties of MLE neg-tags in non-canonical positions

The following description of the functionality of neg-tags in non-canonical positions focuses on LP neg-tags and phrasal neg-tags attached to lone or LD NPs/PPs (see (8)–(10)). These are the most frequently attested categories of neg-tags in non-canonical positions in LIC (N = 30 respectively). The modest frequency with which LP and phrasal neg-tags occur made it possible to analyse in detail every single instance of their use. At the same time, their repeated occurrence helped establish the functional characteristics shared across all neg-tag tokens in each category which is an important requirement to produce faithful descriptions of each category’s respective functional profile. (neg-tags in other non-canonical positions are not considered here because their infrequency makes it impossible to determine which functional properties each of them share.)

3.5 Social properties of MLE neg-tags in canonical, follow-up and non-canonical positions

The close analyses in Sections 3.3 and 3.4 of neg-tags in non-canonical positions showed that they are preferentially located at the LP of clausal anchors or appended to LD and lone NPs/PPs where they seek corroboration of following propositions or of referent activation (and identification). I argued earlier that these findings demonstrate that neg-tags in non-canonical positions are not idiosyncratic performance errors but strategically-employed interactional markers. The question remains whether these neg-tag uses are transient phenomena or early attestations of a discourse-pragmatic innovation. As pointed out in Section 3.3, the frequency of neg-tags in non-canonical positions in LIC is fairly modest. Out of the 2663 neg-tag tokens that could unambiguously be coded for position and scope (see Footnote footnote 4), the vast majority (93.9%, N = 2501) occurred in canonical position, 3.3% (N = 88) occurred in follow-up position, and only 2.8% (N = 74) occurred in the non-canonical positions illustrated in (1) and (6)–(11). The fairly low proportional frequency of neg-tags in non-canonical positions is in line with previous observations that linguistic changes start out slowly (Bailey Reference Bailey1973: 77; Kroch Reference Kroch1989: 202–3; Labov Reference Labov1994: 65, 2001: 449). However, real- and apparent-time evidence is required to substantiate any inferences of ongoing change. In the following, I will provide such evidence.

LP and phrasal neg-tags, as illustrated in (8)–(10) and (12)–(17), have not been reported in previous studies of neg-tags in British English (see, inter alia, Algeo Reference Algeo1988, Reference Algeo, Ricks and Michaels1990; Allerton Reference Allerton, Rohdenburg and Schlüter2009; Cheshire Reference Cheshire1981; Childs in prep.; Hoffmann Reference Hoffmann2006; Kimps et al. Reference Kimps, Davidse and Cornillie2014a,Reference Kimps, Davidse, Cornillie, Davidse, Gentens, Ghesquière and Vandelanotteb; Moore and Podesva Reference Moore and Podesva2009; Pichler Reference Pichler2013; Tottie and Hoffmann Reference Tottie and Hoffmann2006). However, the non-mention in these studies of LP and phrasal neg-tags may not be convincing proof of their recency (see Denison Reference Denison2011). The authors of these studies may have unwittingly circumscribed their analyses to neg-tags in the canonical positions postulated for them in the literature (see beginning of Section 3.3); and some of them have conducted their analyses without consideration of neg-tags’ prosodic phrasing (see, for example, Tottie and Hoffmann Reference Tottie and Hoffmann2006) which is crucial to differentiating RP and LP neg-tag tokens (see Section 3.3). Crucially, though, recent studies of innit that specifically set out to uncover innovations in its use and paid close attention to its prosodic phrasing and positional distribution do not report its occurrence in non-canonical positions either. For example, Andersen’s (Reference Andersen2001) systematic and comprehensive analysis of innit in the mid-1990s COLT data resulted in the discovery and description of many innovations in innit use and distribution, including those in the follow-up position illustrated in (5c) but not including those in the non-canonical positions illustrated in (6)–(11). Stenström et al.’s (Reference Stenström, Andersen and Hasund2002: Ch. 7) detailed analysis of innit, yeah, right and okay in the same dataset identified tokens of yeah occurring after LD and subject NPs, but despite their attention to the syntactic distribution of these tag forms, the authors did not identify any tokens of innit in non-canonical positions. In fact, they argued that innit can only occur in post-clausal position (see Section 3.3). The non-mention in these studies of innit tokens in non-canonical positions therefore suggests that their occurrence is a very recent development in London English. This view is supported by the available apparent-time evidence.

The apparent-time framework hinges on the notion that the speech patterns recorded for different age groups at a single synchronic point in time represent different diachronic stages in language change (Bailey Reference Bailey, Chambers, Trudgill and Schilling-Estes2002). Hence, we may infer evidence of ongoing neg-tag change from fluctuations in the frequency of different neg-tag positions across the two age groups represented in LIC (see Section 3.2). Moreover, because previous research on London English has shown that it is young non-Anglo speakers who tend to be the frontrunners in the use of linguistic innovations (Andersen Reference Andersen2001; Cheshire Reference Cheshire2013; Cheshire et al. Reference Cheshire, Fox, Kerswill and Torgersen2008, Reference Cheshire, Kerswill, Fox and Torgersen2011; Kerswill Reference Kerswill, Auer, Hilpert, Stukenbrock and Szmrecsanyi2013; Kerswill et al. Reference Kerswill, Cheshire, Fox, Torgersen, Schreier and Hundt2013), we may assume that their use reflects a slightly more advanced stage of neg-tag development than that of young Anglos. To establish changes in the positional distribution of neg-tags in the LIC data, Table 3.1 reports the frequency of neg-tags in each position listed on the left (canonical, follow-up, non-canonical) as a proportion of all neg-tags produced by the social group listed at the top (old Anglos, young Anglos, young non-Anglos). For example, of the 555 neg-tags produced by old Anglos, 98.9% occurred in canonical position, 0.7% in follow-up position, and 0.4% in non-canonical positions. (The numbers in the non-shaded columns are provided for comprehensiveness and transparency; for each age group, they give the raw, non-normalised frequency of neg-tag tokens across the three positions on the left.)Footnote ¹¹ The results reveal that in all social groups, neg-tags in canonical position account for the vast majority of neg-tags used while neg-tags in non-canonical positions constitute only a minor part of the neg-tag system. However, comparison of older and younger speakers’ proportional rates of neg-tags in different positions also reveals a slight but steady apparent-time increase in the proportion of neg-tags in follow-up and non-canonical positions, with a concomitant apparent-time decrease in the proportion of neg-tags in canonical position. Moreover, young non-Anglos, the reported leaders of linguistic innovations in MLE, have slightly higher proportions of neg-tags in follow-up and non-canonical positions than young Anglos. In fact, the two most prolific users of neg-tags in non-canonical positions, Tina and Alex, share all the social characteristics previously associated with linguistic innovators in MLE: they are young non-Anglos from Hackney with highly multi-ethnic friendship groups (see Cheshire et al. Reference Cheshire, Fox, Kerswill and Torgersen2008). Despite these speakers’ lead, neg-tags are used in non-canonical positions by both non-Anglo and Anglo adolescents from Hackney as well as Havering and with ethnically diverse and non-diverse friendship groups, suggesting that such neg-tag uses are not idiosyncratic uses but an emerging feature of MLE.

In sum, then, the real- and apparent-time evidence provided supports the view that neg-tags in non-canonical positions represent an incipient stage of a discourse-pragmatic change which has been in progress for a very short period of time.Footnote ¹² In the LIC data, innit, canonically-derived and canonical neg-tag variants respectively account for 85.1% (N = 63), 10.8% (N = 8) and 4.1% (N = 3) of neg-tags in non-canonical positions; two of the three canonical neg-tag variants that occurred in non-canonical positions were produced by old Anglos. These distributions demonstrate that not all neg-tag variants participate in the distributional changes to the same extent; among young Londoners, the positional flexibility of neg-tags is virtually limited to phonetically-reduced neg-tag variants. At the same time, though, the occurrence of variants other than innit in non-canonical positions highlights the importance of analysing discourse-pragmatic variants in the larger context of their co-variants.

3.6 Discussion

The findings outlined in this chapter demonstrate that the use of innit in MLE is characterised by positional and scopal flexibility which extends, to a limited degree, to its co-variants in the neg-tag system (as circumscribed in Section 3.2), particularly canonically-derived and phonetically-reduced variants such as in he or din she. London adolescents from diverse ethnic backgrounds use innit and some of its derivationally-equivalent co-variants in small and fluctuating numbers in positions beyond the RP of main clause anchors, most notably at their LP and appended to LD and lone NPs/PPs. In these positions, neg-tags are consistently used by young Londoners to monitor hearer involvement: LP neg-tags uniformly seek interlocutor attention and corroboration of following propositions; phrasal neg-tags categorically invite confirmation of the activation (and identification) of preceding referents while also guiding referent retrievability and marking referent topicality. Real- and apparent-time evidence supports the view that these uses are very recent innovations in the neg-tag system of MLE. Crucially, these innovations could only be uncovered by scrutinising neg-tags’ positional, scopal, functional and social properties. Scrutiny of each neg-tag token’s position and scope was required to detect the sudden innovations in their positional distribution; functional and social analyses of tokens in non-canonical positions were needed to dismiss the possibility that they are random or idiosyncratic discourse phenomena. The importance of scrutinising the full details of a variable’s or a variant’s use in analyses of discourse-pragmatic change and innovation is best illustrated by comparing the results presented in this chapter with those reported in two previous studies of innit in the same dataset. These studies paid insufficient attention to the form’s variable position and scope and, as a result, failed to uncover the most dramatic recent innovation in innit use.

In their corpus-based, comparative analysis of language use in two specially constructed and comparable sub-corpora of COLT (recorded in 1993) and LIC (recorded in 2005–2006), Torgersen et al. (Reference Torgersen, Gabrielatos, Hoffmann and Fox2011: 96) examined the use of innit and other linguistic forms that ‘indicate speaker-sanctioned places in the discourse where the interlocutor can comment’. To uncover changes in the extent of use of innit and its functional co-variants right, ok, (if/do) you know (what I mean/I’m saying), you get me, they compared each variant’s normalised frequency (= N per million words) and spread (= proportion of speakers using a variant) across the two sub-corpora as well as across social factors in the more recent data. The analysis revealed that the popularity of innit was not affected by competition from its newly emerging, functionally comparable co-variant you get me; in contrast to most of its functional co-variants, the usage levels of innit were not significantly different across the two sub-corpora or across the social factors in the more recent data. Unlike Torgersen et al., the present analysis was not concerned with establishing changes in the frequency of use of innit (or its derivationally-equivalent co-variants),Footnote ¹³ and it cannot therefore assess the reliability of Torgersen et al.’s results regarding the quantitative robustness of innit in London English. (The Ns reported in Table 3.1 are raw frequency scores (for all neg-tag variants) and are hence not reliable indicators of fluctuating frequencies in innit or neg-tag use across age and ethnic groups (see Pichler Reference Pichler2010: 593).) However, my analysis, which was designed to identify developments in how the form innit and other neg-tag variants are used, strongly rejects Torgersen et al.’s (Reference Torgersen, Gabrielatos, Hoffmann and Fox2011: 107) conclusion that the ‘use [of innit] seems to have stabilised’. As demonstrated in the preceding sections, there is compelling evidence that innit and some of its derivation-based co-variants are undergoing changes in their positional distribution and, concomitantly, their discourse functionality. Thus, my in-depth analysis of the positional, scopal, functional and formal properties of the system of neg-tags in LIC has shown that the use of innit, while potentially robust in terms of its frequency and spread of use, is far from stable in terms of its positional distribution and interactional use. Innit is vigorously innovating in MLE.

In a more recent comparative analysis of innit (and isn’t it) in COLT and LIC, Palacios Martínez (Reference Palacios Martínez2015) went beyond comparing frequencies and spreads of innit use to conduct a cross-corpora analysis of the form’s linguistic distribution and interactional use. However, because of its problematic framing and presentation, Palacios Martínez’s analysis and the claims he makes about the changing nature of innit use are unconvincing. This applies to his claims about changes affecting innit in its canonical position as well as his claims about its positional flexibility and occurrence in non-canonical positions.Footnote ¹⁴ Palacios Martínez (Reference Palacios Martínez2015: 383) posits in the abstract to his paper that innit ‘show[s] a high degree of flexibility in the sentence, occurring not only in final but also in initial and medial positions’. However, he does not elaborate on this point in the remainder of the paper nor cite any examples from LIC to confirm the occurrence of innit in these sentence positions. Where he explicitly addresses the positional flexibility of innit (e.g., in his Table 5), he describes its well-established variable turn-position rather than any of the newly emerging non-RP and non-follow-up positions illustrated in examples (6)–(11) in Section 3.3 of this chapter. Thus, despite his supposed focus on the form’s sentence distribution, Palacios Martínez failed to uncover such innovations even though they are clearly attested in LIC, as shown in this chapter. His failure to do so might at least in part be due to the fact that he analysed the orthographically transcribed speech data without consideration of tokens’ prosodic phrasing and intonational realisation (see Palacios Martínez Reference Palacios Martínez2015: 388). As pointed out in Section 3.3, consideration of such details is crucial to identifying individual tokens’ scopal and positional properties (see also Denis and Tagliamonte Chapter 4), which, in turn, is crucial to identifying potential changes in innit or neg-tag usage.

Comparison of my results with those reported in Torgersen et al.’s (Reference Torgersen, Gabrielatos, Hoffmann and Fox2011) and Palacios Martínez’s (Reference Palacios Martínez2015) studies of innit in the same dataset demonstrates that discourse variation analyses which focus on frequency alone or pay insufficient attention to variables’ or variants’ positional distribution are potentially highly inadequate for assessing their stability of use. As pointed out in Pichler (Reference Pichler2013: 21), discourse-pragmatic features are variable and changeable on multiple levels. The preceding discussion is a useful reminder that multiple dimensions of discourse-pragmatic variables’ use need to be analysed to yield accurate accounts of their use, variability and change.

Crucially, identification of neg-tags’ sudden positional flexibility and concomitant changes in their interactional use was only made possible by adopting the empirically- and theoretically-grounded but flexible approach to defining discourse-pragmatic variables advocated in Waters (Chapter 2). Previous variationist analyses have tended to define neg-tags and their variable context in terms of their derivation from the string (auxiliary) + (negative clitic) + (pronominal subject) as well as their position at the RP of clausal anchors (see, for example, Pichler Reference Pichler2013: 179; Pichler and Torgersen Reference Pichler and Torgersen2013). Rigid and uncritical adherence to derivational and positional criteria in the extraction of neg-tags from LIC would have left the neg-tags in non-canonical positions illustrated in (1) and (6)–(11) unaccounted for. I was only able to uncover innovations in the position and function of innit and some of its co-variants because I relaxed the positional criteria stipulated elsewhere for defining neg-tags and their variable context, and included in the analysis all tokens derived from the linear string of components schematised earlier irrespective of their position.

Adjustments to how the variable and the variable context is defined are also required to contextualise the neg-tag innovations described in this chapter in relation to the larger linguistic sub-system in which they become embedded, and establish how they interact with any co-variants in that system (see Weinreich et al. [Reference Weinreich, Labov, Herzog, Lehmann and Malkiel1968] on the ‘embedding problem’). I will illustrate this point with reference to phrasal neg-tags, described earlier as: neg-tags seeking participants’ corroboration of the activation (and identification) of preceding LD NP or lone NP/PP referents, specifically ones that are topical, inferentially accessible and locatable in common ground (see examples (9)–(10) and (15)–(17)). Preliminary observations of the LIC data reveal that this description also fits some phrasal yeah tokens (see also Stenström et al. Reference Stenström, Andersen and Hasund2002: 173). In (20), for example, Rufus has just told the interviewer SF and his interview partner Talal that he thinks it is older rather than younger people who are racist. When he mentions his sixty-year-old father in this context, Rufus tags the LD NP referent my dad with yeah before providing more information about the referent in the following clauses. To Talal, Rufus’s dad might be inferentially accessible from the context of ‘older racist people’ and uniquely identifiable due to previous encounters. Thus, like the phrasal innit tokens discussed in Section 3.4.2, yeah in (20) could be argued to seek confirmation of SF’s and Talal’s activation (and identification) of the semi-active and topical referent my dad.

1. (20)

   Rufus:

   That’s one thing I’ve realised. (..) Cos I know, my dad, yeah, he’s about sixty and I can tell he’s he’s racist. I know my dad is. […]

   (Rufus, 19-M-N-Hv)

In example (21), taken from a much earlier point in the same LIC interview, Rufus also tags the LD NP referent my dad with yeah. However, in this instance, the tagged referent my dad has a different activation status. Because it occurs in response to SF’s question What do your mum and dad do?, my dad is already activated and in focus when Rufus starts providing information about his occupation. In this example, then, yeah does not function to seek SF’s activation of a semi-active referent. Like other phrasal innit or yeah tokens, though, it marks the discourse prominence of the preceding NP referent.

1. (21)

   SF:	What do your mum and dad do?
   Rufus:	My dad, yeah, he’s a thingy. (.) My dad works thingy for (..) {name of DIY store}, yeah, but (..) what he does is like, he goes to every {name of DIY store} in England. Take thingy records whatever for some next thing like the things (.) how much things they sold that year and everything. So he does like he goes to a different {name of DIY store} every week. […]

   (SF = interviewer; Rufus, 19-M-N-Hv)

This very preliminary analysis of phrasal yeah suggests that although innit and some of its neg-tag co-variants are making inroads into the positional space previously only associated with invariant lexical tags such as yeah (see Stenström et al. Reference Stenström, Andersen and Hasund2002: Ch. 7), their usage is subtly different. While both innit and yeah signal high topicality of the phrasal referents they tag, the occurrence of yeah does not seem to be limited to LD or lone NP referents that are semi-active, accessible from the discourse context and locatable in common ground. Moreover, innit and yeah may be differentially distributed across NP referent types: in LIC, the former occurs most frequently with lone NPs, followed by LD NPs, and only rarely between subject NPs and VPs; the latter, by contrast, seems to occur most frequently between subject NPs and VPs, less often after LD NPs and never after lone NPs. And, of course, other tags such as right may occur in these positions and with similar functions too. Thus, in order to establish how the innovative, phrasal neg-tags discussed in this chapter become embedded in the existing system of post-NP referent tags, we would need to redefine the variable and variable context in terms of form, position and function to comprise all neg-tag variants and all invariant lexical tag variants that are appended to NP referents to signal their discourse prominence. Alternatively, we could follow the approach adopted by Cheshire (Reference Cheshire2005) in her analysis of discourse-new markers, i.e., take as the starting point of the analysis the specific function associated with phrasal neg-tags and investigate the range of other linguistic forms used by speakers to achieve this particular function. This analysis would include: the phrasal neg-tag tokens discussed in Section 3.4.2; some phrasal yeah tokens (e.g., that in (20)) but not others (e.g., that in (21)); and a range of other linguistic forms that occur in post-NP referent position, quite possibly including a zero variant. Whatever the starting point of the analysis, a new definition of the variable and the variable context is needed to answer the embedding question.

3.7 Conclusion

In this chapter, I have described recent positional and functional innovations in the use of innit and other neg-tag variants in MLE. Identification of these innovations required close scrutiny of the positional, scopal, functional and social properties of all neg-tag tokens in the LIC data, crucially including those that fall outside the canonical neg-tag variable context, defined in terms of variants’ derivational equivalence and positional restriction to the clausal RP. Albeit based on limited data, my thorough account of innovative neg-tag uses in MLE has the potential to improve our understanding of the life cycle of discourse-pragmatic change. I have caught the ongoing change at an early stage where neg-tags in non-canonical positions still retain the interrogative and corroboration-seeking meaning associated with neg-tags in canonical position (see Hopper’s [Reference Hopper, Traugott and Heine1991: 28–30] ‘persistence’ principle) and where the innovations have not yet spread across all members of the community (see Bailey’s [Reference Bailey1973: 77] and Labov’s [Reference Labov1994: 65, 2001: 449] ‘S-curve model’). This makes it possible to trace their emergence from neg-tags in canonical positions and explore the sociolinguistic mechanisms that have triggered their positional mobility and functional shift (see Pichler [Reference PichlerMS] for details). Moreover, the account given in this chapter provides a diachronic benchmark for future studies of neg-tags in MLE and British English more generally. If neg-tags, and in particular innit, continue to undergo context expansion, studies of innit in corpora yet to be collected will be able to trace any diachronic shifts in its use by comparison with its description in this chapter.

4.1 Introduction

The diachronic literature has documented a multitude of discourse-pragmatic changes in many languages over several centuries (see, for example, Brinton [Reference Brinton1996] on hwæt, I gesse and other Old and Middle English pragmatic markers; Diewald [Reference Diewald2011] on denn and other German modal particles; Traugott [Reference Traugott1995b] on the English discourse markers indeed, in fact, besides). However, to shed light on the interaction between social and linguistic factors in language change, and in particular in cases of linguistic innovation (the actuation of a change), it is necessary to closely examine consecutive generations in a single speech community (Janda Reference Janda2001: 318). These relatively narrow time-spans are the type most often examined in variationist sociolinguistics (see, inter alia, Labov Reference Labov2001b; Tagliamonte Reference Tagliamonte2012). In recent years, several discourse-pragmatic features have come under intense investigation using this method, fuelling the recent surge of interest in discourse-pragmatic variation and change (see, for example, Cheshire Reference Cheshire2007; D’Arcy Reference D’Arcy2005; Pichler Reference Pichler2013; Pichler and Levey Reference Pichler and Levey2011; Tagliamonte and D’Arcy Reference Tagliamonte and D’Arcy2007; Tagliamonte and Denis Reference Tagliamonte and Denis2010).

A common thread to these variationist investigations is the examination of the role of grammaticalisation in the development of discourse-pragmatic features. This is particularly true of variationist work on general extenders (GEs). Some argue that the synchronic distribution of GEs is the result of phonetic reduction, decategorialisation and semantic-pragmatic shift, the most common mechanisms of grammaticalisation (Aijmer Reference Aijmer2002; Cheshire Reference Cheshire2007). However, these arguments are frequently based on data with no time dimension (Cheshire Reference Cheshire2007) or general comparisons of adolescents and adults across studies (see, for example, Cheshire Reference Cheshire2007: 162, Table 2; Tagliamonte and Denis Reference Tagliamonte and Denis2010: 349, Figure 2). When the apparent-time construct has been used, the development of GEs exhibits little or no evidence for ongoing grammaticalisation (Pichler and Levey Reference Pichler and Levey2011; Tagliamonte and Denis Reference Tagliamonte and Denis2010; see Palacios Martínez Reference Palacios Martínez2011). Rather, Tagliamonte and Denis (Reference Tagliamonte and Denis2010) argue that the development of and stuff in Canadian English exhibits signs of relatively abrupt lexical replacement. Innovation in the GE system was not the result of gradual grammaticalisation. Instead when the new variant, and stuff, emerged, it competed with other variants wholesale: its frequency rose but there was no indication of increasing phonetic reduction, decategorialisation or semantic-pragmatic expansion according to speaker age (Tagliamonte and Denis Reference Tagliamonte and Denis2010).

In this chapter, we expand the variationist investigation of discourse-pragmatic change and the role of grammaticalisation and lexical replacement in such change by focusing on what we will refer to as utterance-final tags (UFTs). UFTs are frequent, multifunctional discourse features that typically occur at the end of an utterance and include such forms as eh, you know, right, hey, huh, yeah, innit, okay, you see and others (see, inter alia, Andersen Reference Andersen2001; Columbus Reference Columbus2010; Lam et al. Reference Lam, Thoma and Wiltschko2013; Östman Reference Östman1981; Schiffrin Reference Schiffrin1987; Stenström and Andersen Reference Stenström, Andersen, Percy, Meyer and Lancashire1996; Stenström et al. Reference Stenström, Andersen and Hasund2002; Tagliamonte Reference Tagliamonte2006b). The majority of previous studies have tended to focus on the pragmatic functions of the individual forms categorised here as UFTs. Among others, Östman (Reference Östman1981) and Schiffrin (Reference Schiffrin1987) focus on the pragmatics of you know; and Andersen (Reference Andersen2001) reports on the grammaticalisation of innit in London English. A handful of other studies compare the distribution of multiple forms in corpora. Stenström et al. (Reference Stenström, Andersen and Hasund2002) compare the pragmatic and social distribution of eh, okay, right, yeah and innit in the Bergen Corpus of London Teenage Language, and Columbus (Reference Columbus2010) considers the pragmatic (and syntactic) patterning of fifteen different forms (aacha, ah, ahn, eh, is it, isn’t it, na, no, okay, right, see, yeah, yes, you know, you see) in three varieties of English (British, Indian and New Zealand). These forms can also occur in non-utterance-final position and thus many studies have considered their distribution by syntactic position (see, for example, Lam et al. [Reference Lam, Thoma and Wiltschko2013] on the syntax of eh in Canadian English). Here, we define the variable context structurally and focus on the utterance-final position only, as indicated by the label UFTs (see further Section 4.2.2). In addition, we utilise the following two facts from the literature regarding these forms more generally: (i) they function in relation to the shared knowledge/common ground between speaker and hearer; (ii) they are multifunctional, serving different functions (in relation to shared knowledge/common ground) (Aijmer Reference Aijmer2002: 3; Andersen Reference Andersen2001: 64).

Our approach diverges from this previous work by treating UFTs as a sociolinguistic variable. We focus on the robust variation between the two major variants, you know and right, as in examples (1)–(2), in the Toronto English Archive (TEA), a corpus of contemporary Canadian English that provides an apparent-time perspective of approximately 100 years over the course of the twentieth century (Tagliamonte Reference Tagliamonte2006b).Footnote ¹ The UFT eh, as in (3), is a veritable shibboleth of Canadian identity (Denis Reference Denis and Luo2013; Gold and Tremblay Reference Gold and Tremblay2006); however, its frequency is muted in the TEA. Other variants such as yeah, okay and you see also play only a peripheral role in the variation.Footnote ²

1. (1)

   I was like, ‘Come on, like look at them’ you know?

   (TEA/2j/M/18)Footnote ³

1. (2)

   I thought it would be nice to help her right?

   (TEA/2p/M/14)

1. (3)

   Two kilometers is pretty long eh?

   (TEA/4e/F/18)

Our results reveal a change in progress such that right is innovative in the system, increasing in frequency across apparent time and becoming the most frequent UFT among speakers born after 1970 (under thirty-year-olds). The full trajectory of this variant – from innovation to majority usage – provides a unique opportunity to probe the social and linguistic factors at play during a discourse-pragmatic change (see Denis [Reference Denis2011] on GEs in York and Denis [2015] on GEs in Toronto). We seek to discover who innovates in the selection of a new variant, who leads and what role, if any, grammaticalisation plays. Another goal for this chapter is to highlight the utility of the variationist approach in studying discourse-pragmatic variation and change. We will demonstrate that it is only by examining multiple individual variants in tandem that the driving mechanisms of discourse-pragmatic changes can be elucidated.

The chapter is organised as follows. In the next section, we introduce the TEA, delineate the variable context of UFTs and show how we operationalise discourse context to examine semantic-pragmatic expansion. In Section 4.3, we present the results of a series of statistical tests that suggest that over the twentieth century, right rose in frequency at the expense of you know and that this change was led by male speakers. Section 4.4 closely examines the innovative variant right, in comparison to you know, and we argue that while right rose in frequency, there was no concomitant semantic-pragmatic expansion. In Section 4.5, we discuss the implications of our findings and potential objections. We end the chapter in Section 4.6 by concluding that the rise of right is the result of lexical replacement.

4.2 Data and methodology

4.2.2 The variable context

The circumscription of the variable context is the ‘first and foremost challenge to be confronted’ when embarking upon a variationist study of discourse-pragmatic variation (Tagliamonte Reference Tagliamonte2012: 269). The challenge is to adhere to Labov’s (Reference Labov1972b: 72) ‘principle of accountability’ while recognising that discourse-pragmatic variables do not fit the traditional mould of a linguistic variable, i.e., two or more ways of saying the same thing (Labov Reference Labov1972b: 271). As variationist research evolved over the latter half of the twentieth century, significant methodological advancements ensued, particularly those related to the circumscription of the variable context (see also Waters Chapter 2). In a critique of variationist approaches to discourse-pragmatic variation, Pichler (Reference Pichler2010, Reference Pichler2013) argues that it is often necessary to adhere to a notion of derivational, rather than semantic or functional, equivalence when circumscribing the variable context of a discourse-pragmatic variable. Given that discourse-pragmatic features can express multiple functions simultaneously, an appeal to derivational equivalence allows researchers to observe ‘diachronic meaning changes and synchronic polyvalence’ while anchoring each variant to a shared, delimiting characteristic (Pichler Reference Pichler2010: 590, Reference Pichler2013: 31–2; see also Aaron Reference Aaron2010; Dines Reference Dines1980; Dubois Reference Dubois1992, Sankoff et al. Reference Sankoff, Thibault, Bérubé and Sankoff1978).

Following Tagliamonte and Denis (Reference Tagliamonte and Denis2010), we employ a hybrid approach to circumscribing the variable context by utilising both positional and functional criteria. We included in our envelope of variation any utterance-final discourse-pragmatic feature that primarily communicates to a hearer that the preceding proposition contains shared knowledge (see, for example, Andersen Reference Andersen2001: 69).Footnote ⁵ When right, eh and you know appear independently or in utterance-initial position, they function differently.Footnote ⁶ For example, Andersen (Reference Andersen2001: 73) observes that utterance-initial right in (4) functions to ‘acknowledge mutual contextual assumptions’, i.e., it serves to acknowledge what has been stated in an earlier utterance. Stand-alone eh in (5) functions as a request for repetition. Contrast these with utterance-final you know in (6), which expresses what Andersen (Reference Andersen2001: 73) labels a ‘presumption of assumptions’, i.e., shared knowledge.

1. (4)

   Jane:	Hello Peter! What are you doing here?
   Peter:	Maths course work. Tt.
   Jane:	[Oh oh!]
   Peter:	[Have to] hand it in.
   Jane:	Right fair enough.

   (Andersen Reference Andersen2001: 72, ex. 16)Footnote ⁷

1. (5)

   FP:	Do you know Ann Pearson?
   DM:	No, I don’t.
   FP:	Eh?
   DM:	No, I don’t.
   FP:	Don’t you?
   DM:	No.

   (B75/032/F/96)Footnote ⁸

1. (6)

   Well she don’t like your auntie you know. That’s why she’s crying.

   (Andersen Reference Andersen2001: 73, ex. 17)

In Canadian English, one cannot substitute you know or eh for right in (4) and express the same function nor can one exchange right or you know for eh in (5). However, in (6), eh and right are functionally-equivalent substitutes for you know, as illustrated in (7) and (8), all ‘expressing a presumption of the interlocutors’ common ground’ (Stenström et al. Reference Stenström, Andersen and Hasund2002: 172).

1. (7)

   Sue lives in one of these renovated homes, right. And a lot of them are, you know, people that have roots here.

   (TEA/NX/M/46)

1. (8)

   PT:	Hard on her too eh. She came through it though.
   SM:	Yeah. And the baby’s getting better. Fairly healthy now?

   (TEA/Nh/F/55)

Often in the transcription of fluent discourse, it is difficult to determine whether or not a token modifies the previous utterance (making it utterance-final) or the following utterance (making it utterance-initial) (see also Pichler Chapter 3). For example, you know in (9a) could be either utterance-final (9b) or utterance-initial (9c).

(9)

1. a.

   I see him on occasion and we reminisce about the old days you know so it’s good.

   (TEA/N¥/M/72)

2. b. [[… we reminisce about the old days] you know] [so it’s good]

3. c. [… we reminisce about the olds] [you know [so it’s good]]

However, the intonational phrasing of an utterance correlates with its underlying structure.Footnote ⁹ Whether or not a form is attached to the preceding utterance or the following utterance is perceptually salient and acoustically indicated by short periods of speaker silence. By example, consider the conversation in (10).

1. (10)

   AA:	Well this is this is city homes in here.
   SM:	[Uh-huh.]
   AA:	[You know] it so it’s so it’s a uh city homes and uh the rent is geared towards your income.
   SM:	Mm-hm.
   AA:	Which is great for me [you know]. And it’s convenient
   SM:	[Mm-hm]

   (TEA/N¥/M/72)

The waveform in Figure 4.1a shows that the first instance of you know in (10) is preceded by a short period of silence and is part of the same intonational phrase as the utterance that follows it.Footnote ¹⁰ Tokens like this were considered utterance-initial, as illustrated in (11), and were excluded.

(11)

1. a. _IP(… this is city homes in here)_{IP IP}(you know, it’s city home …)_IP

2. b. [this is city homes in here] [you know [it’s city homes …]]

Figure 4.1a Utterance-initial you know (tokens excluded)

However, any examples like the second instance of you know in (10) were included. Here, we see a period of silence after you know in the waveform in Figure 4.1b, indicating that the form occurs at the end of the first utterance, as illustrated in (12).

(12)

1. a. _IP(… which is great for me you know)_{IP IP}(and it’s convenient)_IP

2. b. [[… which is great for me] you know] [and it’s convenient]

Figure 4.1b Utterance-final you know (tokens included)

In this analysis, we extracted every unambiguously utterance-final discourse-pragmatic feature that could function as an indicator of shared knowledge. This yielded 1938 tokens of the variable for in-depth analysis.

4.3 The trajectory of change in the UFT system

We begin our analysis of UFTs by examining in Table 4.3 the distribution of UFT variants in three broad age ranges: under thirty, thirty to fifty-nine and over sixty. Out of the nearly 2000 tokens of UFTs extracted from the corpus, the variants right and you know are the majority variants, representing 42.2% and 36.4% of the total number of UFTs respectively. Variation between the two variants is strongly correlated with age. Right rises from 2.2% of all UFTs among the oldest speakers in TEA to 60.4% among the youngest speakers. In contrast, you know decreases from 66.2% to 31.2% and 22.5% across age groups. The less frequent variant, yeah, is also in decline, and eh is very infrequent at all ages despite its status as a stereotype of Canadian English (Denis Reference Denis and Luo2013). Other variants, such as you see and okay, also play a peripheral role in this speech community and we will not discuss these further.

Figure 4.2 plots the results of four separate binomial fixed-effects logistic regressions testing the effect of speaker age (on the x-axis) on the probability of right, you know, yeah and eh respectively (on the y-axis). The size of the data points plotted along the top of the figure represents the frequency of occurrence of right, you know, yeah and eh (as indicated by the shape) at that point along the x-axis. The dots along the bottom of the figure represent non-occurrence of the respective variant. Fitted logistic regression lines for speaker age from each of the four models are also plotted. The grey ribbons represent 95% confidence intervals. The patterns in Figure 4.2 confirm the observations made with respect to Table 4.3. For older individuals, right is a marginal variant with a probability hovering under .20. However, the probability of the variant rises steadily among the forty- to sixty-year-olds (born between 1943 and 1963) and continues to rise among young people to a probability of almost .75. In contrast, the probability of you know decreases over this same apparent-time period from just under .75 to around .20 such that right becomes the majority variant among speakers born after approximately 1970. This cross-over pattern is confirmed by independent mixed-effects models that take individual speaker variance into account (not shown). The probabilities of the two marginal variants yeah and eh are consistently low, and speaker age was not significant for either variant. A mixed-effects model with individual speaker taken into account (not shown) confirms that the frequency of these two variants is stable across apparent time.

Figure 4.2 Binominal probability of right, you know, yeah and eh variants by speaker age

Labov’s (Reference Labov2001b) principles of linguistic change (whether from above or below) champion the role of women. Countless studies of linguistic changes in progress have confirmed this observation across many different situations, both cross-dialectally and across languages (see, for example, Dubois and Horvath Reference Dubois and Horvath1999; Haeri Reference Haeri1994; Milroy et al. Reference Milroy, Milroy, Hartley and Walshaw1994; Tagliamonte and D’Arcy Reference Tagliamonte and D’Arcy2009). However, Tagliamonte and Denis (Reference Tagliamonte and Denis2010: 360) report that the recent change in progress in the GE system of Toronto English is led by males. In a multivariate analysis, males significantly favour the innovative variant and stuff while females favour the outgoing variant and things. Further, Denis (Reference Denis2011) observes that the innovators of and stuff in York, England, were both men and women. An added nuance was that the leaders of the change in York were shown to be the most gregarious individuals in the dataset. These findings corroborate the hypothesis that sex asymmetry may not be a necessary concomitant of linguistic change (see Tagliamonte and D’Arcy Reference Tagliamonte and D’Arcy2009: 99).

This brings us to the question of who may be the innovators of right in Toronto English. Table 4.4 shows the results of a fixed-effects binomial logistic regression which tests the effects of age and sex on the occurrence of right.Footnote ¹³ Furthermore, in order to model the non-linear trends over time (note that in Table 4.3, the difference in frequency between the middle-age group and the oldest age group is much greater than the difference between the middle-age group and the youngest age group), we include a quadratic polynomial term for age. (For discussion of modelling age as continuous and with quadratic polynomials, see Tagliamonte and Baayen [Reference Tagliamonte and Baayen2012: 152–3].) The results of the best model are presented in Table 4.4.Footnote ¹⁴

Given the complexity of regression models that include interactions with continuous factors and polynomial terms, we graphically represent these results in Figure 4.3 to facilitate interpretation. (All R code is provided in the Appendix.) Two fitted logistic regression curves are plotted representing the interaction between age and sex against the probability of right. The solid line represents the trajectory of females (the reference level of the model in Table 4.4); the dotted line is the trajectory of males. Light grey ribbons around each line represent 95% confidence intervals. Any region along the horizontal axis (i.e., age range) where the two lines’ confidence intervals do not overlap can be interpreted as a statistically significant difference between the relevant social cohorts in that age range. The points along the top of the figure represent the frequency of occurrence of right by the two sexes, and the points along the bottom of the figure represent the occurrence of any other UFT.

Figure 4.3 A visual summary of the generalised linear model reported in Table 4.4

Figure 4.3 reveals that underlying the general upward trajectory of right observed in Figure 4.2 is a striking effect of sex. Males and females favour (and disfavour) right at different points in apparent time. Males are the first to adopt innovative right beginning with the seventy-year-olds born between 1933 and 1943. However, males under thirty-five have not taken up the new variant to the same degree as their elders. In contrast, there is a sustained linear increase in the use of right among females under fifty years old.Footnote ¹⁵ The result is a cross-over pattern at approximately age twenty.

A so-called ‘retreat’ by men from a variant that is accelerating among females is a well-attested pattern of linguistic change (see, inter alia, Labov Reference Labov2001b: 300; Tagliamonte Reference Tagliamonte2012: 33; Trudgill Reference Trudgill1972). What we observe here seems reminiscent of Tagliamonte and D’Arcy’s (Reference Tagliamonte and D’Arcy2007) observation of males in Toronto retreating from the innovative be like quotative as it rises in frequency among females. In the initial stages of this change, no sex effect was observed but in the most recent developments, the rate of be like among females accelerated, and males not only lagged behind but visibly withdrew from using the new variant. Tagliamonte and Denis (Reference Tagliamonte and Denis2014: 123) observe this same trend (men lagging behind women) with be like in south-eastern Ontario. However, unlike these be like changes (or any change we know of), the increasing use of right has a cross-over pattern. Why is this the case? With both be like and right, the consistency is that young males retreat from whichever variant the young females are using. The nuance with right is that men actually led the change in the first place. But after the females substantially increased their use of this apparently male innovation (among the thirty-year-olds), the rate among younger males significantly lowered, suggesting a withdrawal from its use. A dissociation by men from a change they initially led is a pattern that, to our knowledge, has not been observed before.Footnote ¹⁶

In sum, our statistical analyses up to this point expose a genuine change in progress in Toronto English. Given the patterns in Figures 4.2 and 4.3, right is taking over from you know. The full spectrum of change from zero usage to nearing categorical usage (among young females) provides us with the opportunity to test for gradual semantic-pragmatic expansion of this incoming UFT variant over apparent time. Is the change in frequency visible in Figures 4.2 and 4.3 the result of grammaticalisation, i.e., the gradual expansion of right into semantic-pragmatic contexts formerly reserved for you know, or of lexical replacement, i.e., the complete replacement of one variant for another?

4.4 Semantic-pragmatic context expansion

Bybee (Reference Bybee, Joseph and Janda2003), Heine (Reference Heine, Joseph and Janda2003), Himmelmann (Reference Himmelmann, Bisang, Himmelmann and Wiemer2004) and Traugott (Reference Traugott, Lehmann and Malkiel1982, Reference Traugott1988, Reference Traugott, Stein and Wright1995a,Reference Traugottb) all identify the expansion of semantic-pragmatic contexts as a defining feature of grammaticalisation. Crucially, this functional expansion is theorised to be gradual rather than abrupt (see, for example, Brinton and Traugott Reference Brinton and Traugott2005: 150; Bybee Reference Bybee, Joseph and Janda2003: 612; Traugott and Trousdale Reference Traugott, Trousdale, Traugott and Trousdale2010: 25–6). However, much of the literature that argues for gradual, rather than abrupt, semantic-pragmatic expansion is typically qualitative and considers diachronic correspondences rather than changes currently in progress (see Janda Reference Janda2001: 318). As Janda (Reference Janda2001: 318) observes, ‘[w]hat we are most sorely missing […] are sociolinguistically oriented studies which would compare the ways in which elements apparently undergoing grammaticalisation are used by speakers vis-à-vis those on whom they model their behavior’. That is, it is difficult to assess the gradualness of grammaticalisation without examining changes in progress in consecutive generations (see Denis Reference Denis2015).

If discourse-pragmatic changes such as the rise of right observed earlier are a reflex of grammaticalisation, then we hypothesise that during the increase in frequency of right confirmed in Section 4.3, there should be a concomitant expansion across semantic-pragmatic contexts in apparent time. Upon inception of right as a UFT, the variant is predicted to be contextually restricted, but as it expands in use, it is hypothesised to gradually encroach on more and more contexts. In contrast, if the rise of right is similar to the abrupt lexical replacement of older GEs by and stuff, as documented in Toronto (Tagliamonte and Denis Reference Tagliamonte and Denis2010), then we hypothesise that as soon as right entered the variable UFT system, it would have been immediately available in all of the contexts in which its main competitors were available and thereafter would not necessarily expand further. In this section, we put these two hypotheses to the test. Given the frequency and changing profile of right and you know, we will focus on these two variants in the remainder of the chapter, probing further into the linguistic mechanisms underlying their changing frequencies.

4.4.1 Distribution of discourse contexts in apparent time

This sub-section will serve two purposes. First, we show that both you know and right are used across a range of discourse contexts though the apparent-time trajectories of each are different. Second, we establish that the discourse contexts in which speakers use UFTs are stable across time.

Table 4.5 and Figure 4.4 present the distribution of you know by discourse context across age group.Footnote ¹⁷ We report the frequency with which you know occurred in a particular discourse context, in a particular age group as a proportion of all UFTs in that context in that age group. For example, of the 293 UFTs that occurred in fact-reporting contexts in the youngest age group, 15% were you know. First, note that the distribution of you know across contexts mirrors the downward trajectory observed in Figure 4.2. In all but one discourse context, younger speakers use you know less frequently than older speakers. (The exception is the context ‘accusation’ but this context has particularly low Ns.) Second, although the frequencies have decreased in each context, you know is still used across a wide range of discourse contexts in the youngest age group. This is particularly evident in the six most frequent contexts: fact-reporting, narrative fact, opinion, question, quotative delimiter and strict narrative.

Table 4.5. Distribution of discourse contexts of you know by age group

	>60		30–59		<30
Discourse context	%	N	%	N	%	N
Accusation	–	0	0	2	33	3
Command	–	0	–	0	0	5
Exclamation	67	3	0	4	0	1
Fact-reporting	49	124	33	79	15	293
Narrative fact	82	79	17	12	16	213
Opinion	53	15	39	23	24	46
Question	100	3	0	3	4	25
Quotative delimiter	87	15	26	19	28	68
Strict narrative	67	18	33	12	19	133

Figure 4.4 Proportional frequency of you know by discourse context in three broad age groups

Table 4.5 and Figure 4.4 also suggest that the discourse contexts in which UFTs are used are relatively stable. Fact-reporting, narrative fact, opinion, question, quotative delimiter and strict narrative have non-zero Ns in each generation. However, UFTs only occurred in the command context in the youngest generation and they occurred in the accusation context in the youngest and middle generation. On the basis of the TEA, one could argue that over the twentieth century, UFTs became licit in these contexts and thus what we may have evidence for is a gradual pragmatic expansion of UFTs (in general) to these discourse contexts (though note the very low Ns). However, consider the examples in (16) and (17). Although accusation contexts were rare in the TEA, example (16) from Canadian author H. A. Cody, writing in 1935, suggests that this context was available for UFTs early in the twentieth century, indeed just before the initial upswing in the use of right. Likewise, Avis (Reference Avis1972: 99) lists several early Canadian examples of eh in command contexts, as in (17), suggesting that despite a paucity of examples in our data, this context has allowed UFTs for some time.

1. (16)

   ‘Ho! ho! so you think I am the devil, eh?’

   (H. A. Cody, 1935, The Crimson Sign)

1. (17)

   ‘Listen, Harry, phone me before you go out tonight, eh?’

   (M. Callaghan, 1928, Strange Fugitive)

Table 4.6 and Figure 4.5 present the frequency of right in each discourse context in three age groups as a proportion of all UFTs in that discourse context and age group. From the oldest age group to the youngest age group, the proportional frequency of right in all but one discourse context increases. (The exception is the low-frequency exclamation context [total N = 8].) More importantly, the range of contexts in which right appears increases from two to five to eight. This contrasts with you know which both older and younger speakers use in seven out of nine discourse contexts. A superficial look at these proportions might lead to an argument for gradual semantic-pragmatic context expansion in the use of right in apparent time. However, statistical confirmation is required.

Table 4.6. Distribution of discourse contexts of right by age group

	>60		30–59		<30
Discourse context	%	N	%	N	%	N
Accusation	–	0	0	2	33	3
Command	–	0	–	0	20	5
Exclamation	0	3	25	4	0	1
Fact reporting	2	124	29	79	67	293
Narrative fact	1	79	42	12	73	213
Opinion	0	15	0	23	46	46
Question	0	3	33	3	80	25
Quotative delimiter	0	15	47	19	63	68
Strict narrative	0	18	0	12	12	133

Figure 4.5 Proportional frequency of right by discourse context in three broad age groups

4.5 Assessing grammaticalisation and semantic-pragmatic change

Two objections can be raised regarding this conclusion. First, the time-frame of our data may not have sufficient depth for capturing grammaticalisation in action. Second, using discourse context as an evaluation metric of semantic-pragmatic change may be ill-founded.

Is 100 years in apparent time across the twentieth century sufficient to assess grammatical developments? We have made the argument, following Janda (Reference Janda2001), that only by looking at consecutive generations within the same speech community are we able to assess grammaticalisation. However, in which generations does grammaticalisation occur? Méndez Naya (Reference Méndez Naya2006: 147–8) observed that right has had a long development in English dating back to the Old English adjective riht meaning ‘straight, not bent’; since, the adjective has been variously used as an adjunct, modifier, intensifier and eventually as a discourse-pragmatic feature. Indeed, Méndez Naya argues that as early as the eighteenth century, right functioned as a discourse-pragmatic feature, at that time already the result of a long period of grammaticalisation. Thus, how can developments 200 years later tap discourse-pragmatic grammaticalisation of this variant? Three facts support our approach. First, definitions of grammaticalisation typically emphasise the gradual and continuous development of forms. As Bybee et al. (Reference Bybee, Perkins and Pagliuca1994: 4–5) put it, grammaticalisation involves ‘changes in lexical morphemes by which some few of them become more frequent and general in meaning, gradually shifting to grammatical status, and developing further after grammatical status has been attained’ [emphasis added]. By this definition, the focus of investigation could be any generation, as grammaticalisation is not theorised to end abruptly. Further, even after achieving discourse-pragmatic status, further developments are possible, and indeed likely. In any case, Méndez Naya’s (Reference Méndez Naya2006: 158) eighteenth-century examples are all instances of right ‘used to show that the speaker agrees with a previous statement’, and thus fall outside of our definition of a UFT (see Section 4.2.2). This leaves open the question of when right became a UFT (Méndez Naya Reference Méndez Naya2006: 158). Regardless, linguistic expressions are not independent of the speakers and the speech communities that use them. Even if right may have functioned as a UFT in some community prior to the twentieth century, the fact remains that as a UFT it has only risen in frequency in Toronto English since the 1930s.Footnote ²¹ Once right becomes a variant of the UFT variable, it simply infiltrates all the usual contexts without disrupting the extant system.

Is the expansion of discourse contexts a reasonable proxy for grammaticalisation? Himmelmann (Reference Himmelmann, Bisang, Himmelmann and Wiemer2004), for example, explicitly equates grammaticalisation with context expansion. Others may argue that expansion from one context to another does not equate with pragmatic shift (i.e., propositional to textual to expressive function) often observed in grammaticalisation. The problem is that discourse context and pragmatic function are tightly linked and tricky to disentangle for independent assessment. As Andersen (Reference Andersen2001: 71) observes, ‘it can be difficult to separate attitudinal functions from illocutionary functions, since pragmatic markers such as eh? can indicate both the speaker’s non-committing attitude and directive illocutionary force in the same utterance’. Further, textual and/or expressive functions may be more frequent in certain discourse contexts than in others. For example, Stenström et al.’s (Reference Stenström, Andersen and Hasund2002: 169) epistemic function (i.e., ‘the speaker is uncertain as to the truth of [the proposition] and would like the hearer to verify it’) may be more common with interrogatives. The imagination-appealing function (i.e., the speaker ‘does not assume that [the proposition] is a shared belief, but assumes that it is compatible with the hearer’s background knowledge’) may be more common in narratives. The facilitative function (i.e., ‘the speaker believes that [the proposition] is an opinion or belief shared by the hearer and herself, and that the hearer may wish to corroborate [the proposition]’) may be more common with exclamations. In essence, expansion of discourse context may well correlate with increasingly more non-propositional meanings. In this way, context expansion can be taken to reflect pragmatic shift.Footnote ²²

4.6 Conclusion

In this chapter, we have probed the nature of discourse-pragmatic variation and change by targeting a frequent discourse-pragmatic feature (right), and treating it as a variant within the UFT system. We have employed a representative sociolinguistic dataset, replicable quantitative methods and a variety of statistical tools (both old and new), all of which have offered insights to inform a defensible interpretation of UFT variation and change in Canadian English. These procedures can be applied to the study of any discourse-pragmatic feature once it has been conceptualised (and justified) as a variant in a variable system.

At the outset, a distributional analysis revealed notable generational trends in the data (Table 4.3). Then, using logistic regression models, we first presented statistical confirmation that you know and right are undergoing change in Toronto English. Individuals born in the early decades of the twentieth century up to approximately the 1920s only use you know while those born from the 1930s onwards increasingly use right (Figure 4.2). Further analysis distinguishing males and females (Figure 4.3) exposed that this change is socially complex: males advanced the use of right and females followed suit but much more slowly. The point of male/female convergence at approximately .65 probability is among individuals born in the early 1980s. After this point in the apparent-time array, young females use right even more than their elders but young males’ use wanes. Our results confirm the general pattern of male retreat from features that females use substantially. The new observation here is that males retreat from changes in this way even when they originally led them. This male dissociation leads to a sex cross-over (Figure 4.3).

We also tested for gradual semantic-pragmatic expansion by coding a sub-set of the data for discourse context. To test for expansion in apparent time across contexts, we employed a new statistical technique, namely zero-inflated Poisson regression (Long Reference Long1997). This method disentangles the interaction between: (i) form (in this case right); (ii) its frequency differential across speakers of different ages; and (iii) the associated contexts of its use. Our statistical model suggests that the innovators of right in Toronto could use this new variant wherever UFTs were already used in the community. This result parallels earlier findings for another, utterance-final feature in the same corpus, GEs (Tagliamonte and Denis Reference Tagliamonte and Denis2010). The system of GEs in Toronto underwent a similar change across the same speakers in the same time-frame such that and things was being replaced by and stuff. With this phenomenon as well, there was no evidence for ongoing, gradual grammaticalisation. Instead, the new variant emerged with the full set of characteristics of the variant it was replacing. In the GE and UFT systems, the new variants (and stuff, right) replaced earlier variants (and things, you know). This convergence across two studies of similar discourse-pragmatic features suggests that we may have tapped into a more general mechanism of discourse-pragmatic change: lexical replacement. For both GEs and UFTs, a new variant emerges full-blown with all the functional characteristics of earlier variants in the respective variable system.