1 Introduction
English has three strategies for expressing negation with an indefinite item, which in this article are termed any-negation, no-negation and negative concord, respectively. Any-negation features a negative marker not on the verb (or the enclitic n’t as in (1)), which has scope over an indefinite negative polarity item with the form any(-), such as any, anything, anyone or anybody. No-negation, illustrated in (2), lacks not and instead shows negation on the indefinite item itself, as in no, none, nothing, no one or nobody. Negative concord features both not/n’t on the verb and a no- form, as in (3), but is interpreted as a single instance of negation.Footnote 2
(1) I wasn’t paying any rent here (York, M/26)Footnote 3
(2) There’s nothing you can do about it (Toronto, M/24)
(3) I haven’t got you nothing yet (Tyneside, F/19)
The historical development of negation in English can illuminate how any-negation, no-negation and negative concord have evolved, which in turn will help explain their contemporary distribution. In Old English, the primary negator was ne. When used, ne always appeared immediately before the main verb as a proclitic (Ingham Reference Ingham2013: 123). In addition to ne, negative clauses sometimes featured a negative adverb with forms including nāwiht, nāht, nōht and nōwiht, as in (4) (Nevalainen Reference Nevalainen1998: 267). This became more common in the Middle English period, during which nowiht grammaticalized leading to the development of a compulsory post-verbal form not (van Kemenade Reference van Kemenade2000: 58; Iyeiri Reference Iyeiri2001: 86; Wallage Reference Wallage2012: 722), as shown in (5).Footnote 4 Negative concord with indefinite items, as in (3) above, was common in Middle English (Jack Reference Jack1978: 38), but no-negation could also be used once ‘n-item indefinites became able to introduce negation by themselves’ (Ingham Reference Ingham2013: 144–5). In Early Modern English, the co-occurrence of multiple negative markers declined in frequency, while the occurrence of not with any-items (shown in (6)) became possible in a change reportedly led by the upwardly mobile middle classes, particularly men (Nevalainen Reference Nevalainen1998: 277–8, Reference Nevalainen2006: 580; Nevalainen & Raumolin-Brunberg Reference Nevalainen and Raumolin-Brunberg2006). The choice of any-negation over negative concord at that time has been described as ‘a selective process from above in terms of the speaker-writer’s education and social status’ (Nevalainen Reference Nevalainen2006: 580).
(4) He nōwiht tō gymeleste ne forlet (Bede 206, 17)
‘He didn’t leave no whit (nothing) to neglect’
(5) thou n’art nat put out of it (Chaucer’s Boece, Book I, P5, 9–10) (14th C)
‘you [neg] are not put out of it’
(6) to enjoyne the said Baxter not to prosecute anie accion
‘to order the said Baxter not to prosecute any action’ (Bacon 1590, Privy Council III, 99)
Any-negation had thus become a viable alternative to no-negation in Early Modern English. Did any-negation increase in frequency at the expense of no-negation? Willis et al. (Reference Willis, Lucas and Breitbarth2013: 38) suggest that this has been happening since the Middle English period, but Ingham (Reference Ingham2013: 146) notes that any-negation ‘did not quickly displace no negation’, as no-negation was the favoured variant in sixteenth-century letters. He suggests that increases in the frequency of any-negation over time may be related to the introduction of negative auxiliaries (e.g. don’t) that became used more often from the sixteenth century onwards (Ingham Reference Ingham2013: 146). Alternatively, the purported increase in any-negation ‘may well be an impression due to the disappearance of multiple negation from the standard’ (Mazzon Reference Mazzon2004: 100).
To investigate the historical trajectory of the variation, its contemporary patterning and the constraints on use, Tottie (Reference Tottie1991a, Reference Tottie1991b) undertook an extensive quantitative investigation of any-/no-negation in corpora of Standard British English. Her most contemporary data consisted of two samples from the 1960s – written prose (excluding fiction) from the Lancaster–Oslo/Bergen Corpus of British English (LOB) and spoken spontaneous conversation from the London–Lund Corpus of Spoken English (LLC). In these samples, no-negation was most preferred with existential be constructions, followed by stative have and copula be, whereas lexical verbs tended to occur with not (Tottie Reference Tottie1991a, Reference Tottie1991b). Tottie (Reference Tottie1991a: 440) suggests that these effects reflect lexical diffusion of any-negation over time, namely that ‘the more frequent a construction is, the more likely it is to be retained in its older form for a longer period of time’. When comparing these data with a sample of written texts from the Helsinki Corpus of English Texts (1640–1710), she observes an increase in the frequency of no-negation (for all verb types except copula be) which she suggests ‘could indicate that there has been a development from no-negation to not-negation [any-negation] between the late seventeenth century and the present day’ (Tottie Reference Tottie1991a: 462).
While this trajectory of change seems plausible, the evidence provided in support of this conclusion is not entirely convincing. Firstly, the quantitative trend towards any-negation may have been biased by the inclusion of invariable sentences in the analysis (i.e. tokens where only one variant was possible) alongside variable ones – a decision which was taken ‘because of the problems involved in assessing variability in the historical sample’ (Tottie Reference Tottie1991a: 461). Secondly, the data used to establish verb frequency can likewise be critiqued. Tottie (Reference Tottie1991a, Reference Tottie1991b) refers to do, know, give and make as the lexical verbs that are most frequent with no-negation and links this to their relatively high frequency overall, as indicated by their high ranking among c.6000 items in Francis & Kučera’s (Reference Francis and Kučera1982) Frequency Analysis of English Usage: make (rank 40); know (rank 63); give (rank 72). Although Francis & Kučera (Reference Francis and Kučera1982) also rank do, have and be as more frequent than other (lexical) verbs, which creates a parallel between frequency and rate of no-negation, Tottie (Reference Tottie1991a) points out that this source does not distinguish between main and auxiliary functions of verbs. It is also not clear what the frequency rank was for the other lexical verbs in the sample and whether this correlates with their rate of no-negation. The rates of no-negation for do, make, know and give are calculated based on relatively few tokens and the trends in use fluctuate between speech and writing (Tottie Reference Tottie1991a, Reference Tottie1991b), suggesting that there may not be enough data to interpret the proposed trends unequivocally.
Regardless of whether the frequency-based account is supported, the contemporary variation could indeed reflect ‘a process of transition’ from no-negation to any-negation (Tottie Reference Tottie1991b: 235). However, a more recent corpus-based investigation by Wallage (Reference Wallage2017), who investigated the evidence for change both historically and in modern Standard English, found no evidence of ongoing change. In his comparison of the variation in the Penn–Helsinki Parsed Corpus of Early Modern English (PPCEME) and the British National Corpus (BNC), the frequency of variants was consistent, as were the verb type constraints, leading him to conclude that the results ‘suggest historical persistence of variation rather than ongoing change’ (Wallage Reference Wallage2017: 197). Childs’ (Reference Childs2017) quantitative analysis of this variable in contemporary regional varieties of English spoken in Glasgow (Scotland), Tyneside (North East England) and Salford (Greater Manchester) similarly found no evidence of a change in progress towards no-negation, corroborating Wallage’s findings. As summarized in section 3 of this article, Childs (Reference Childs2017) also highlights some methodological issues concerning Tottie’s (Reference Tottie1991a, Reference Tottie1991b) definition of the variable and the variable context.
Our aim in this article is to investigate the potential competition between any-negation, no-negation and negative concord, to establish whether the variation in contemporary Englishes reflects an ongoing diachronic change from no-negation to any-negation or is instead more indicative of a change which has now become stable. To this end, we examine the phenomena in two sets of vernacular speech corpora, from Northern England and Ontario, Canada, to see whether attending to geographic, linguistic and social factors can offer insights into the current state of the variation.
2 Corpora and samples
The corpora comprise vernacular sociolinguistic interviews with native English speakers, conducted in Northern England and Ontario, Canada. A preliminary study of negation in Ontario was made based on the Canadian corpora analysed here (Harvey Reference Harvey2013). In the present article, we aim to determine whether similar language internal and external tendencies exist in geographically distinct varieties of English, through a comparison of two varieties on either side of the Atlantic: Ontario English and Northern British English. These dialects share historical links in the sense that the vast majority of the early founder populations (Mufwene Reference Mufwene2001: 27–9) to Ontario hailed from the British Isles. Southern Ontario was predominately settled by migrants of British descent from the United States (Loyalists),Footnote 5 whereas more northern climes had significantly more migrants from Northern England, Scotland and Ulster (see Cowan Reference Cowan1961: 288; Elliott Reference Elliott2004: 65; Boberg Reference Boberg2010: 77). We therefore adopt a comparative approach, analysing the distribution of any-negation, no-negation and negative concord across our target locales to (i) establish how variation is conditioned within the grammar(s) of English; (ii) identify how the variation is socially conditioned; and (iii) assess the evidence for linguistic change in progress and the state of the variation in the different varieties (Tagliamonte Reference Tagliamonte2013: 186).
The Canadian recordings (Tagliamonte Reference Tagliamonte2003–Reference Tagliamonte6) are from Toronto, a major urban area with over 5 million inhabitants, and Belleville, a town of approximately 50,000 residents situated two hours east of Toronto. The two locales have distinctive demographic profiles: Toronto is a diverse multicultural urban centre while Belleville is more homogenous with a strong history of Loyalist settlement.
The British recordings are from four Northern English locales, three of which are herein combined as the ‘North East of England’ since the varieties spoken there share similar dialectal features (Beal et al. Reference Beal, Burbano-Elizondo and Llamas2012). These three locales are the urban Tyneside region and two nearby smaller urban areas, Wheatley Hill and Durham, in County Durham. The Tyneside dataset consists of a sample of recordings from the Diachronic Electronic Corpus of Tyneside English (DECTE, Corrigan et al. Reference Corrigan, Buchstaller, Mearns and Moisl2010–12), while the dataset from Wheatley Hill and Durham is from Tagliamonte (Reference Tagliamonte1998, Reference Tagliamonte2003). We compare the results from North East England with those from York (Tagliamonte Reference Tagliamonte1996–8, Reference Tagliamonte1998, Reference Tagliamonte2003), a major city in North Yorkshire, where the native dialect is distinct from that spoken in the North East but they share certain pan-Northern English linguistic features (Trousdale Reference Trousdale2012; Buchstaller & Corrigan Reference Buchstaller and Corrigan2015).
The conversations within these sociolinguistic interviews, which were designed to elicit vernacular speech, are informal. They contain many narratives of personal experience about childhood, hobbies and interests, as well as local history. The interviewees’ speech is representative of the dialect of their local area, having all been born and raised in their respective locales. The corpora have different sizes and the distribution of speakers by age varies across the datasets.Footnote 6 In total, there are 305 speakers in our sample and their ages range from 14 to 80+, as table 1 shows.
Speaker sample
The recordings provide ample tokens of the variable under study, and rich intra-speaker variation, as shown in (7) and (8).Footnote 7 In (7), the clause construction is the same each time, featuring existential there were and the complement jobs, but the speaker alternates between any-negation for the first sentence and no-negation for the second. Similar optionality is shown in (8): negation can be expressed within the verb phrase (8a) or on the indefinite (8b).
(7) (a) There weren’t any jobs.
(b) There were no jobs to be had (Toronto, F/43)
(8) (a) I don’t have any information.
(b) you had no option (Belleville, M/33)
3 The variable context
All instances of any-negation, no-negation and negative concord were extracted from our sample. Any-negation tokens feature the indefinite items any, anybody, anyone, anything, owt (found exclusively in the British data, meaning ‘anything’) or anywhere in the predicate. These are prototypically licensed by a negative marker (not/n’t) on the preceding verb in the clause, which has scope over the indefinite. Within prepositional phrases (henceforth PPs), indefinite any- items are often licensed in this same way (9a), but they can also appear alongside elements such as without (10a). No-indefinites can also occur in these environments with no change in referential meaning, as (9b) and (10b) illustrate.
(9) (a) We’re not under any obligation
(b) We’re under no obligation (Toronto, F/29)
(10) (a) … someone else was appointed without any reference
(b) … someone else was appointed with no reference (York, F/24)
Tokens of no-negation feature the negative counterparts to the indefinite any- items, namely no/none, nobody, no one, nothing, nowt (‘nothing’ – exclusively in the British data) and nowhere. Negative concord tokens were also captured using this latter set of search terms, since they feature these no-forms (in addition to a negatively marked verb). Instances of never and n’t/not…ever were not extracted because never is near-categorically used in this environment, so including those tokens in our analysis would bias the results (Tottie Reference Tottie1991b: 109; Childs Reference Childs2017).
The variable context excludes tokens with a negatively marked verb that has scope over the articles a/an or zero determiner, i.e. sentences of the type in (11a).
(11) (a) well she said # that doesn’t make sense # that’s the cheapest of the lot
(b) well she said # that makes no sense # that’s the cheapest of the lot
(Tottie Reference Tottie1991b: 178, 211)
Although some previous studies of any-negation and no-negation interpret sentences like (11a) and (11b) to be semantically equivalent and include them in their analysis (Tottie Reference Tottie1991a, Reference Tottie1991b), we consider the underlying form of makes no sense in (11b) to be doesn’t make any sense, rather than doesn’t make sense like in (11a). Our rationale is as follows.
Childs (Reference Childs2016, Reference Childs2017) argues that a/an/ø are not equivalent to any because they have distinct semantic and syntactic properties. Firstly, unlike the a/an/ø items, any is a negative polarity item which expresses ‘a kind of extreme non-specificity’ (Lyons Reference Lyons1999: 37) that the former do not – i.e. they are ‘less exception-tolerant’ (Chierchia Reference Chierchia2013: 27).
Secondly, several investigations of negative concord in different varieties of English find that a and an either do not undergo negative concord at all, or do so very rarely (Labov Reference Labov1972a: 806; Cheshire Reference Cheshire1982: 66; Smith Reference Smith2001: 131). Although Howe (Reference Howe2005) finds examples of this kind in African American Vernacular English (AAVE), Labov (Reference Labov1972a: 810–11) had argued (also based on AAVE) that such instances arise because any is inserted prior to negative concord taking place. This can explain why (12b) and (12c) are equivalent in emphatic force while (12a) is much weaker.
(12) (a) He didn’t have a car.
(b) He didn’t have any car.
(c) He didn’t have no car.
Thirdly, Tottie (Reference Tottie1991b: 205) reports that her informant judged he is not/isn’t a moralist as semantically equivalent to he is no moralist, but not semantically equivalent to he is not/isn’t any moralist. This judgement is contrary to the overwhelming consensus that no is equivalent to not any (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 782; Tieken-Boon van Ostade Reference Tieken-Boon van Ostade1997: 188; Anderwald Reference Anderwald2002, Reference Anderwald2005; Peters Reference Peters2008; Peters & Funk Reference Peters and Funk2009; Wallage Reference Wallage2017). Tottie (Reference Tottie1991b: 130) rightly argues elsewhere that Bill is not a doctor and Bill is no doctor do not have the same meaning since the former is a denial while the latter expresses the view that Bill ‘lacks the essential qualities’ to be a good doctor. This same explanation can be extended to he is no moralist, yet that sentence was included and Bill is no doctor was not.
Fourthly, although Tottie includes indefinite articles/zero determiners in her variable context, she acknowledges that instances of indefinites with not-negation that do permit no-negation tend to have any (or potentially allow it), and that no-negation similarly tends to correspond to an underlying any rather than a/an or zero determiners (Tottie Reference Tottie1991b: 263).
We therefore argue based on the discussion in Childs (Reference Childs2016, Reference Childs2017) as summarized above that no-negation is semantically equivalent to n’t/not…any- constructions, rather than n’t/not…a/an/ø. We thus exclude the latter token types and use the term any-negation to refer to the former.
The extracted tokens were rigorously sorted to remove those that fall outside the precise variable context described above. Several other contexts appear to be candidates for any/no-negation and negative concord, but there are cases where either variation is not possible or any-negation is not semantically equivalent to no-negation, which we outline below.
3.1 Indefinites in subject position
As our variable context requires any-/no-forms to be in the predicate, we excluded indefinites in subject position. No is categorical in this context (13a) and alternatives with any-negation as in (13b) did not appear in any of the corpora. Indeed, these are rare even in other varieties such as those spoken in Ireland in which the failure of negative attraction is possible for some speakers (Harris Reference Harris1984: 305; Filppula Reference Filppula.1999: 179–81; Reference Filppula2008: 338). These tokens are therefore not considered further in the investigation.
(13) (a) Nobody would sit in that seat (Toronto, M/36)
(b) *Anybody wouldn’t sit in that seat
3.2 Presence of adverbs
The presence of an adverb in the clause restricts the choice of variant. For example, when actually is in the immediate scope of a negative marker, as in (14a), the sentence is interpreted as ‘a hedged statement’ (Paradis Reference Paradis2003: 202). In contrast, (14b) has ‘the function of emphasizing the subjective judgement of the importance of the situation involved in the proposition in question’ (Paradis Reference Paradis2003: 194). Other adverbs such as absolutely cannot occur after negation (15b), only before it (15a). Therefore, tokens containing adverbs were excluded from the sample given the lack of semantic equivalence between variants.
(14) (a) I didn’t actually need anything (York, F/52)
(b) I actually needed nothing
(15) (a) There’s absolutely no flights out of Victoria (Toronto, M/49)
(b) *There’s not absolutely any flights out of Victoria
3.3 Negative raising or cross-clausal negation
In cross-clausal or negative raising contexts, the movement of the negative marker changes the meaning or force of the sentence, as demonstrated by the subtle differences in (16). Furthermore, certain negative raising constructions such as I don’t think are formulaic and have become grammaticalized (Scheibman Reference Scheibman2000; Pichler Reference Pichler2013), which leads to use of the any-negation variant. Cross-clausal negation was thus also excluded from our sample.
(16) (a) I don’t think I would change anything (Tyneside, F/18)
(b) I think I wouldn’t change anything
(c) I think I would change nothing
3.4 Adjectival complements
As Tottie (Reference Tottie1991b: 131) notes, any- and no-negation often differ in meaning when there is an adjectival complement. Examples like (17a) had already been excluded for having no determiner and it is clear that (17b) has a much stronger, more emphatic reading. Pairs like this are therefore not semantically equivalent and likewise do not form part of the variable context.
(17) (a) It doesn’t look good for a Christian woman (M/SG/121, Tyneside)
(b) It looks no good for a Christian woman
3.5 Unclear tokens
Tokens that were unclear in the audio/transcripts, occurred in unfinished clauses or were ambiguous in any respect were similarly excluded from our sample as in these cases we could not be certain as to their classification.
Observing all of these procedures produced 1,821 tokens where any-negation, no-negation, and negative concord were all viable with semantically equivalent meanings.
4 Coding
We coded for both grammatical and social factors. The grammatical factor is verb/construction type, which has been found to be a major factor governing the variation in previous research (Tottie Reference Tottie1991a, Reference Tottie1991b; Childs Reference Childs2017; Wallage Reference Wallage2017).
Verb/construction type was coded according to the categories in (18). Existential there+be constructions consistently have the highest rates of no-negation, while lexical verbs have the lowest and tend to have any-negation. The percentages of no-negation for be, have and have got differ between previous studies but consistently rank between the existentials and lexical verbs in this regard.
Variation within PPs has drawn considerably less attention in the literature. Although Tottie (Reference Tottie1991b: 265–6) examines PPs and notes that no-negation is infrequent in this environment, her results are based on her overall sample of both invariable and variable structures, in speech and writing. Tottie’s findings for the variable spoken sample, most comparable to the nature of our data (in that they are spoken and comprise only tokens where both any- and no-negation are possible, in contrast to her ‘written’ and ‘invariable’ samples), are inconclusive due to low Ns (N=4). Although PPs do not feature negative marking on a verb, any/no do alternate in this environment, e.g. without any versus with no, as in (18f). We therefore decided to include PPs as part of our investigation as a point of comparison to examine whether they pattern like the examples of verbal negation.
(18) (a) Existentials (there + be) There was no canteen (Belleville, M/52)
(b) be It wasn’t any particular amount (Tyneside, M/25)
(c) have We had nothing (Toronto, F/19)
(d) have got I haven’t got any (Wheatley Hill, F/72)
(e) Lexical verbs He’s not heard anything (York, M/55)
(f) In PP We’ll end up with no Santa’s grotto
(Tyneside, M/21)
Our decision to implement binary social variables in our analysis may mask self-imposed social categories that can be pertinent in the analysis of language variation (Eckert Reference Eckert1989, Reference Eckert2000). However, our data emanate from pre-existing large-scale corpora, which precludes taking a more ethnographic approach. Our aim is to analyse the linguistic variable’s distribution quantitatively in conjunction with the classic sociolinguistic variables of sex (male/female), age (birth year, ranging from 1906 to 1993) and education (with/without post-secondary education), which will allow us to assess the evidence for change in progress.
5 Distributional analysis
5.1 Locality
Figure 1 shows the overall distribution of negative constructions for the four areas under study.
Distribution of any-negation, no-negation and negative concord in each community
Negative concord is virtually absent in Toronto, Belleville and York, and occurs rarely (6.6 per cent of the time) in the North East of England. Because of its low frequency, negative concord is henceforth excluded from our quantitative analysis. In contrast, variation between no- and any-negation is present in all varieties, but the distribution is markedly different for each country. In Canada, the two constructions have near-equal frequency, with a slight preference for no-negation in Toronto. In England, no-negation dominates at 63 per cent in York and 71.9 per cent in the North East. Given that any-negation is the newcomer variant historically, these figures indicate that any-negation has made greater inroads into Canadian English dialects, while in Northern British English varieties the older no-negation variant endures.
5.2 Verb/construction type
Table 2 shows the distribution of no-negation according to verb/construction type in each community, in terms of the categories presented in section 4. The greyed out percentages for have got in Toronto and have got/be in Belleville indicate that there are less than 10 tokens in these cells.
Distribution of no-negation per verb/construction type in each community
Despite the different overall frequencies between the dialects, the patterning of no-negation by verb/construction types is remarkably similar in each community. Existentials (there+be) consistently have the highest frequency of no-negation, with near-categorical rates in England. Constructions with be, have and have got also have high rates of no-negation, ranging from (excluding the two tokens of have got in Belleville) 59 per cent in Belleville for have up to 94 per cent in North East England for be. In contrast, the lexical verbs have a strong tendency to co-occur with any-negation. This pattern is consistent across all four communities, ranging between 7 per cent in Belleville and 36 per cent in North East England. Furthermore, these tendencies are generally in keeping with those observed for this variable in Standard British English (Tottie Reference Tottie1991a, Reference Tottie1991b; Wallage Reference Wallage2017) and varieties of English spoken in Glasgow, Scotland and Salford, Greater Manchester (Childs Reference Childs2017). The consistency in these trends emphasizes the robustness of the verb type constraint on this variation in grammars of English globally.
PPs are positioned between lexical and functional verbs in terms of their propensity to take no-negation, but display different tendencies on each side of the Atlantic: in Ontario, Canada, PPs most often occur with any-negation, while in Northern England they tend to occur with no. The PP environment is therefore one in which the choice of any-negation versus no-negation appears to be subject to cross-dialectal differentiation. The over-arching pattern, however, is a marked division between functional verbs (be, have, have got) versus lexical verbs (see also Wallage Reference Wallage2017).
Bybee & Hopper (Reference Bybee and Hopper2001) argue that constructions such as existentials are highly frequent and therefore are processed and produced as a whole, which could account for their high propensity to occur with no-negation. As discussed in section 1, Tottie (Reference Tottie1991a, Reference Tottie1991b) argues that be and have are also high frequency, making them resistant to change and therefore more likely to retain the variant that is oldest historically, no-negation. In contrast, individual lexical verbs are less frequent, which Tottie (Reference Tottie1991a, Reference Tottie1991b: 232) argues makes them more likely to undergo change, i.e. take any-negation. Tottie (Reference Tottie1991b) notes that verbs which have more tokens within her any-/no-negation variable context do exhibit more no-negation, e.g. existentials were higher frequency (N=38) and had more no-negation than copula be (N=20), but the evidence for this frequency-based account can be questioned, as discussed in section 1.
In his analysis of any-negation and no-negation in Ontario, Canada, Harvey (Reference Harvey2013) appeals to Tottie’s (Reference Tottie1991a, Reference Tottie1991b) frequency account to explain collocational tendencies in his data, but suggests that syntactic factors may also be relevant to better understand the verb type effects in speakers’ choice of variant. Harvey’s (Reference Harvey2013) proposal, based on Smith’s (Reference Smith2000) account of do-absence, appeals to the fact that functional and lexical verbs have different movement properties and positions in the syntactic structure relative to the negative operator. Under this account, since be (obligatorily) and have (optionally) raise for tense and agreement (Pollock Reference Pollock1989) and thus reside in a position that is syntactically close to the negative operator in the functional projection NegP, they are more likely to take no-negation. Lexical verbs, on the other hand, obligatorily remain low in the VP with much greater structural distance between them and the operator, making no-negation more difficult to derive.
Childs (Reference Childs2017) proposes that this effect could be explained in one of two ways: (i) no-forms have an uninterpretable negative feature that must agree with an interpretable negative operator in NegP (Zeijlstra Reference Zeijlstra2004), in which case lexical verbs favour any-negation because they remain in a position between the negative operator and the post-verbal indefinite item and thus can disrupt the Agree relation required for no-negation/negative concord; or (ii) no-negation is distinct from the other two variants in being marked for negation within the post-verbal NP and moving to NegP to receive sentential scope (see Kayne Reference Kayne1998; Svenonius Reference Svenonius2002; Zeijlstra Reference Zeijlstra2011; Tubau Reference Tubau2016), which would be dispreferred with lexical verbs since they constitute additional material that the no-negation must move across. While the former account predicts that both no-negation and negative concord would be dispreferred with lexical verbs, the latter predicts that this is true only of no-negation (Childs Reference Childs2017). Childs (Reference Childs2017) finds in her data from three Northern British communities that the latter account is more strongly supported since any-negation and negative concord behaved in tandem with respect to verb type and overall frequency, while no-negation was distinct.Footnote 8
5.3 Sex
Labov’s (Reference Labov2001) principles of linguistic change emphasize the role of women as leaders of change, whether it is from above (Principle 3: a conscious change whereby women favour a variant with more prestige) or below (Principle 2: an unconscious change in which women use innovative variants more than men). When we consider the historical context of the variation that we are investigating, we see that the inception and rise in the frequency of any-negation over negative concord in Early Modern English has been characterized as a change from above. At that time, any-negation certainly fit the definition of a prestige form that is associated with groups of higher status (e.g. Labov Reference Labov1972b: 138; Van Herk Reference Van Herk2012: 48). It was used by people who were more educated and of a higher social standing (Nevalainen Reference Nevalainen1998: 277–8, Reference Nevalainen2006; Nevalainen & Raumolin-Brunberg Reference Nevalainen and Raumolin-Brunberg2006) and was associated with legal, administrative and professional language (Rissanen Reference Rissanen2000: 125; Nevalainen & Raumolin-Brunberg Reference Nevalainen and Raumolin-Brunberg2006: 150). Contrary to what one would typically expect for a change from above, the increasing use of any-negation as opposed to negative concord in Early Modern English was led by men (Nevalainen Reference Nevalainen1998: 277–8, Reference Nevalainen2006), because at that time women ‘did not promote language changes that emanated from the world of learning and professional use, which lay outside their own spheres of “being”’ (Nevalainen & Raumolin-Brunberg Reference Nevalainen and Raumolin-Brunberg2006: 131). That said, while women ‘did not prove to be the leading influence in this change, … neither did they lag behind in adopting the innovation’ (Nevalainen Reference Nevalainen1998: 284).
In Present-Day English, it is not clear whether any-negation holds any particular prestige over no-negation (which the above accounts did not investigate), though this remains a question for future research. We examine the contemporary frequency of no-negation (versus any-negation) in figure 2. In this distributional analysis, and others pertaining to social variables, we remove existentials given their near-categorical tendency to take no-negation.
Distribution of no-negation in each community according to speaker sex10
In figure 2, we see a reversal of the historical association between any-negation and men. Male speakers now use no-negation more than women in Belleville, North East England and York. In Toronto, on the other hand, there is barely any distinction between the sexes in their use of this variable and this is the only community in which the distribution is not significant.Footnote 9
If these patterns do reflect modern-day competition between variants, the fact that women use any-negation more than men in three out of our four communities may not necessarily represent change from above, but change from below (Labov Reference Labov1966: 207, Reference Labov1972b: 133). This interpretation would capture women’s propensity to lead in unconscious changes towards greater use of an innovative variant without recourse to prestige. While any-negation is not a ‘recent’ innovation, it is historically the newest variant of our negation subtypes. Alternatively, what we may be witnessing here is stable linguistic variation with social patterning between men and women in three of our four communities. Examining how the variation patterns according to speakers’ birth year, as we do in section 5.4, offers us a way of further assessing the evidence for present-day change or stability.Footnote
5.4 Birth year
To explore whether there is evidence for change or stability in British and Canadian vernaculars, we categorized the data according to speakers’ birth year as a proxy for real time. Although there were no speakers born in 1906–20 in the North East England and Belleville samples, and no speakers born in 1981–93 in the York sample, the timespan is nevertheless expansive and allows us to observe diachronic trends in the frequency of no-negation (versus any-negation), as shown in figure 3.
Distribution of no-negation in each community according to speakers’ birth year
The distribution of variants according to birth year is significant in York, but not in any other locale.Footnote 11 In York, there is an upswing among the speakers born in 1971–80 compared to speakers born in the previous few decades. In all communities, even York, there is not a steady increase or decline in the use of no-negation. Therefore, taken as a whole, the results more strongly support the third of our three possible interpretations set out in section 5.3, i.e. that the current variation between any- and no-negation is relatively stable, as opposed to undergoing change from above or below. The nature of these trends is explored further in section 6, where birth year is considered alongside other predictors in a mixed-effects logistic regression analysis to confirm which factors have a significant impact on variant choice while holding the effect of the individual constant.
5.5 Education
The final social factor considered here is education, specifically whether a speaker has completed post-secondary education or not. As D’Arcy & Tagliamonte (Reference D’Arcy and Tagliamonte2010) discovered in their analysis of relative who, linguistic items that were once introduced by change from above can retain their distributional association with higher levels of education and professional status several centuries later in corpus-based analysis. Thus, in this section we investigate the possibility that any-negation, which had prestige and was introduced in a change from above in Early Modern English (Nevalainen Reference Nevalainen1998: 277–8, Reference Nevalainen2006; Nevalainen & Raumolin-Brunberg Reference Nevalainen and Raumolin-Brunberg2006), may be used at higher frequencies amongst more educated speakers.
Figure 4 shows the relationship between the percentage of no-negation and education. Although the distribution is more socially stratified according to education in the two British communities than the two Canadian ones, the effect is not statistically significant in any locale.Footnote 12 The direction of the pattern is the same in both the North East of England and in York: speakers without post-secondary education use no-negation more than those who have been educated beyond secondary school, i.e. those who are more highly educated use any-negation at higher rates. Although no-negation has been considered ‘more literary’ (Biber Reference Biber1988: 245) and is more frequent in writing than speech (Tottie Reference Tottie1991a, Reference Tottie1991b; Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999), our data show that this does not equate to a higher use of no-negation among more educated speakers.
Percentage of no-negation in each community according to speaker education
The distributional results have revealed that both internal and external factors impact upon speakers’ choice between any- and no-negation in British and Canadian English. The following section presents the results of statistical modelling to establish which effects are significant when all are considered simultaneously and to investigate whether they operate consistently on each side of the Atlantic.
6 Statistical modelling
We now undertake mixed-effects logistic regression analysis of the variation using the lme4 package (Bates et al. Reference Bates, Mächler, Bolker and Walker2015) in R (R Core Team 2014), with one model per locale: Toronto, Belleville, North East England and York. The four predictors analysed independently in section 5 were included in the models for each variety: ‘verb/construction type’, ‘sex’, ‘education’ and ‘birth year’ as fixed effects, plus ‘speaker’ as a random effect.Footnote 13 As the results from section 5.2 revealed that be, have and have got all tend to occur with no-negation, these were combined as ‘functional verbs’, as opposed to ‘PPs’ and ‘lexical verbs’. ‘Existentials’ were excluded given their near-categorical tendency to take no-negation. ‘Sex’ was coded as ‘male’ versus ‘female’, and ‘education’ as ‘secondary’ versus ‘post-secondary’. ‘Birth year’ was collapsed from the original eight categories to four larger groups (‘1906–30’, ‘1931–50’, ‘1951–70’, ‘1971–93’), to overcome the fact already mentioned that some of the corpora did not have speakers born in 1906–20 or 1981–93 (see figure 3).
Table 3 shows the results of the regression of the factors affecting the choice of no-negation over any-negation in the four locales.
Mixed-effects logistic regression of factors affecting the choice of no-negation (over any-negation) per locale
The results in table 3 reveal that ‘verb/construction type’ is the major constraint affecting any- and no-negation, with all four locales displaying a statistically significant effect whereby functional verbs (be, have, have got) strongly favour no-negation and lexical verbs disfavour it. The overall frequency of no-negation with PPs was earlier found to differ between Canadian English and British English (see section 5.2), but in table 3, the distinction between functional verbs and PPs is significant only in Toronto. Nevertheless, the same propensity holds across the board: in every community, PPs slightly disfavor no-negation compared to functional verbs. The fact that this effect is significant only in Toronto and not Belleville may simply be because there are fewer tokens of PPs in Belleville (N=13).
Consideration of the social factors shows that men use more no-negation than women across all four communities, at statistically significant levels in the UK locales and in Belleville, which corresponds with the distributional analysis in section 5.3. Education, on the other hand, has no significance in the variation. Birth year meanwhile shows small deviations between the groups but is significant only in York and only between the speakers born earliest (1906–30) and born latest (1971–93).
The statistical analysis therefore confirms that the most significant constraint on the variation between any- and no-negation is linguistic, i.e. verb/construction type. The social effects are secondary: there is an additional association between no-negation and male speakers, but no education-based effects. The evidence for ongoing change in progress is slim overall, as age is not significant in three out of four locales. The only community where a change in progress is plausible is York, given the direction of the effect and the significant distinction between the very oldest and very youngest cohorts.
7 Discussion
Our quantitative comparative sociolinguistic investigation of any-negation and no-negation in Northern England and Ontario, Canada, has demonstrated how the variation is structured, both linguistically and socially. It has situated the variation in these distinctive Englishes in the context of whether there is a continuing longitudinal change from no-negation to any-negation, or relative stability in the modern day.
Our first major finding is that regardless of locality, the underlying linguistic constraints are parallel. The choice of variant is conditioned by the same internal factor, verb/construction type, which operates consistently in all four communities: functional verbs favour no-negation and lexical verbs disfavour this variant. PPs pattern in-between, with the distributions suggesting a Canadian versus UK English distinction, though this is significant only in Toronto. The verb type contrast is the major constraint and corroborates previous findings (Tottie Reference Tottie1991a, Reference Tottie1991b: 232; Childs Reference Childs2017; Wallage Reference Wallage2017). Tottie (Reference Tottie1991a, Reference Tottie1991b) had argued that the high frequency of functional verbs makes them more resistant to change and more likely to retain the older no-negation variant than lexical verbs, which are lower in frequency. However, structural explanations, either with an appeal to the different syntactic positions of functional and lexical verbs (see Childs Reference Childs2017; Harvey Reference Harvey2013) or to a consideration of typologically consistent soft versus hard contrasts (Burnett et al. Reference Burnett, Koopman and Tagliamonte2018), can also account for the same facts. Such analyses are similarly in line with other investigations that have identified the relevance of underlying syntactic mechanisms for other English verb-related phenomena, such as do-absence (Smith Reference Smith2000). This does not preclude the possibility that frequency may still have some role to play in this variation, e.g. in maintaining the use of idiomatic expressions with no outside our variable context, such as no way! (see Peters Reference Peters2008; Peters & Funk Reference Peters and Funk2009). High-frequency verbs can indeed be slower to succumb to a syntactic change (see Lieberman et al. Reference Lieberman, Michel, Jackson, Tang and Nowak2007 on verb regularization; Grieve-Smith Reference Grieve-Smith2009 on negation in French). However, even when studies are designed to replicate each other, there can be conflicting conclusions about the role of frequency. For example, Erker & Guy’s (Reference Erker and Guy2012) investigation of personal pronoun variation in Spanish found no independent frequency effects on the distribution (only interactions), whereas Bayley et al. (Reference Bayley, Greer and Holland2013) did find such effects in their study which was specifically designed to replicate the former. Further research is therefore required into the role of frequency in morphosyntactic variation and change more generally.
Some previous synchronic corpus-based analyses of Standard English had suggested that any-negation is increasing at the expense of no-negation (Tottie Reference Tottie1991a, Reference Tottie1991b), but our data from a range of English vernaculars provide little evidence that such a change is ongoing. The exception to this is in York, where we see a significant difference between the variation for speakers born in 1906–30 versus 1971–93, but the distinction between speakers born in 1906–30 and those born in the intermediate decades (1931–50, 1950–71) is not significant. This could therefore reflect slow change in this community which is only observable after several decades. The apparent lack of change in progress in Toronto, Belleville and North East England is consistent with conclusions drawn from other recent investigations of this variation in dialects of English spoken in Glasgow and Salford in the UK (Childs Reference Childs2017) and in the comparison of the variation between PPCEME and the BNC (Wallage Reference Wallage2017).
The distinction between York and North East England with respect to change in progress for this variable could reflect the latter’s more conservative profile. The stronger persistence of no-negation in the North East may be a reflex of local societal norms – it is a region which has not been subject to much socio-demographic change in its recent history, largely on account of its disadvantaged status relative to the rest of the UK (Robinson Reference Robinson2002: 322). In York, a city that has, in contrast, undergone substantial social reorganization over the last 50 years (Huby et al. Reference Huby, Bradshaw and Corden1999), we see some indication of movement towards any-negation, though further research with a longer diachronic time-depth would allow us to investigate this trajectory.
The variation remains significantly affected by sex in all communities except Toronto, though the trend is the same: men use no-negation more than women. The prestige once associated with any-negation in Early Modern English (Nevalainen Reference Nevalainen1998: 277–8, Reference Nevalainen2006: 580; Nevalainen & Raumolin-Brunberg Reference Nevalainen and Raumolin-Brunberg2006) may therefore manifest itself in its modern-day distribution where it is favoured by women. However, as noted previously, we did not observe any effect of education on the variation. Whether speakers actually perceive any-negation as more prestigious than no-negation remains a question for further investigation, since both variants are Standard English alternatives. Based on our results, we would not expect any corresponding prestige-based stylistic variation between the variants. However, the variants have become specialized to achieve different discoursal effects, with no-negation favoured when introducing discourse-new information and any-negation more likely to relate back to discourse-old propositions (Wallage Reference Wallage2015, Reference Wallage2017; Childs Reference Childs2017; see also Tottie Reference Tottie1991b on discourse effects on the variation).
We therefore conclude that the correlates between no-negation and male speech are characteristic of the difference between a conservative variant and a historically newer variant (Labov Reference Labov1966et seq.). The fact that significant social effects are not found in Toronto whereas the other communities show at least some social stratification may reflect differences in the socio-historical context, beginning with the divergent input of the founders to different parts of Ontario (who hailed from diverse dialect regions of the British Isles, the United States and elsewhere) as well as the social conditions that led to the subsequent levelling of their varieties (Chambers Reference Chambers1991; Dollinger Reference Dollinger2006: 10; Tagliamonte & Denis Reference Tagliamonte and Denis2014).
Our investigation demonstrates the advantages of cross-varietal comparison in the analysis of morphosyntactic variation, both for establishing the robustness of linguistic constraints on usage and identifying how the variation is embedded socially in different communities. The social patterns in the variation reflect a distinction between an older variant (no-negation) and a newer variant (any-negation), but in largely stable variation. Even in diverse global spaces such as Ontario and Northern England, we see the consistency and primacy of internal constraints on morphosyntactic variation.
