variable deletion of french ne: a cross-stylistic perspective

Variable deletion of French ne: a cross-stylisticperspective

Nigel Armstrong *

Department of French, University of Leeds, Leeds LS2 9JT, UK

Abstract

This paper considers the socio-stylistic distribution of the French variable morpho-syntacticparticle ne. The interspeaker axes of variation in ne are summarised first, prior to a con-

sideration of some intraspeaker data deriving from a corpus of spoken French. As well aspresenting group scores, we examine intraspeaker variation in ne by focusing on use of thevariable by a single speaker in both speech styles recorded. One principal finding is that theformal speech situation was such as to work against the strong constraint that induces French

speakers to produce clitic+verb sequence as perhaps the default form; a constraint thatstrongly favours ne deletion. In informal style, several examples of retention of ne seem toshow that the tone of a stretch of discourse can vary independently of the informality of the

speech style. Several of the stylistic effects produced by the informants in conversation stylethrough their use of ne show that the micro-style variation which is present in both of thebroad styles is reflected in the use of ne through a reduction in degrees of style shift, since the

‘formal’ episodes in informal style call for the use of ne frequently relative to formal style —albeit within the context of very little use of ne overall. # 2001 Published by Elsevier ScienceLtd. All rights reserved.

Keywords: French; Linguistic variation; Sociolinguistics

1. Introduction

In the paradigm of variationist sociolinguistics, the phonological level of analysishas presented and continues to present fruitful areas of enquiry, while in contrast,variation in morpho-syntax has proved more recalcitrant to Labovian-type analysis.Specifically, the analysis of variation on the morpho-syntactic levels raises not onlyserious methodological difficulties, but also some quite fundamental issues of socio-linguistic definition, which we discuss here. These latter issues are in addition to

Language Sciences 24 (2002) 153–173

www.elsevier.com/locate/langsci

0388-0001/01/$ - see front matter # 2001 Published by Elsevier Science Ltd. All rights reserved.

PI I : S0388-0001(01 )00015 -8

* Tel.: +44-113-233-3490; fax: +44-113-233-3477.

E-mail address: [email protected] (N. Armstrong).

those concerning the equivalence between grammatical variants on the semantic,pragmatic and discourse levels which have been much discussed in the literature(Sankoff, 1973; Lavandera, 1978; Labov, 1978; Romaine, 1981).Grammatical variation is of course not monolithic, as implied by a conventional

acceptation of the term ‘grammar’ that comprises morphology and syntax. Thusvariation in French grammar ranges from relatively low-level morpho-syntacticvariables such as deletion of ne, and some variable liaison that has a morphologicalinput, to more complex, meaning-bearing areas of variation such as subject- andobject-NP (noun phrase) doubling, interrogative sequences, the synthetic and peri-phrastic future tense forms, variation between auxiliary verbs (avoir and etre) andpronouns with definite and indefinite reference (nous, on, tu and vous). Broadly,variationist methods have been most successful, in the sense of producing robustresults, when applied to the former type of variable. Although quantitative patternshave emerged from studies of variables of the latter type (e.g. Sankoff and Thibault,1980), small token numbers have in general precluded the formulation of strongconclusions based on statistically robust results.In this article I consider the socio-stylistic distribution of the French variable

morpho-syntactic particle ne. A principal aim here is to evaluate the extent to whichpragmatic factors need to be taken into account in an analysis of stylistic or intras-peaker variation in ne (intraspeaker henceforward). The pragmatic dimension ofanalysis has not yet been examined to any depth in relation to ne deletion, at least so faras intraspeaker variation is concerned. I conduct this analysis through a considera-tion of some intraspeaker data on ne deriving from the ‘Dieuze corpus’ (Armstrong1993, 1996). I consider first briefly the interspeaker axes of variation in ne.

2. Interspeaker variation in French ne

As Milroy points out (1987, p. 143–144), the salient difference between phonolo-gical and grammatical systems from the viewpoint of data collection and analysis isthat any phonemic inventory, even though it will inevitably be subject to some ‘leak’as a result of variation and change working within it, is a more or less closed system.Consequently one can expect speakers to employ phonological variables of interestrelatively frequently, even if, as Milroy remarks, the full range of variants of a non-binary variable may be difficult to obtain within the limits of a standard sociolinguisticinterview.In contrast, one result of the open-ended nature of grammatical systems relative to

what obtains in phonology, is that in many cases speakers can exercise far greaterchoice in their use of grammatical structures to express the meaning they wish toconvey. As stated above, however, exceptions to this generalisation are provided bylow-level morpho-syntactic variables such as French ne, an ‘omissible item’ in Cov-eney’s (1996, p. 30) typology of grammatical variables. As Coveney remarks (1996,p. 55), the presence or absence of ne is ‘possibly the best known sociolinguisticvariable in contemporary French’. Several scholars Ashby (1981, 1991) and Coveney(1996, 1998) have reported on deletion of ne in the French of France; more recent

154 N. Armstrong / Language Sciences 24 (2002) 153–173

reports still, on Canadian French, are by Tennant (1995) and Sandy (1997). Giventhe relatively well-known character of ne deletion, it seems sufficient to state brieflyhere that standard French can express negation through the ‘embracing’ structurethat most commonly comprises pre-verbal ne, verb form, plus one of several post-verbal items: pas, jamais, plus, rien, and personne are the most frequent. The issuehere is of course that everyday spoken French very largely has no ne, relying insteadon the negative and restrictive items listed above, which are more salient than ne byvirtue of being capable of receiving phrase-final stress. In addition to these negativeand restrictive items, the restrictive adverb que is also frequent; this lexical item isnot found phrase-finally, but can receive stress in sequences such as: ‘ceux qui (ne)font que francais’.The negative particle ne can therefore be compared to a phonological variable

such as schwa or a liquid consonant in that broadly, it is variably present or ‘varieswith zero’, as well as in its frequency of occurrence. Ashby (1981) and Coveney(1996) have discussed in considerable detail the linguistic constraints bearing uponthe deletion of ne; these are relatively complex, but no more so than the arrays thatconstrain some phonological variables. Those weighing upon the variable omissionof ne are connected with phonology as well as syntax: in oversimplified terms, ne ineveryday spoken French is overwhelmingly deleted after clitic pronouns, with somevariation due to phonological factors, but tends to be retained after full nounphrases and other complex syntactic constructions. We discuss more fully in a sub-sequent section some linguistic constraints bearing upon the variable omission of ne,principally in relation to its value as a marker of style.Clearly, what distinguishes ne from phonological variables is its involvement in

syntax. Despite this fundamental difference, one might wish to argue that the simi-larity between ne and certain phonological variables is reflected in the sociolinguisticdistribution shown by ne, of which a typical and recent example is reported byCoveney (1996, p. 86), who recorded a corpus of spoken French in the 1980s inchildren’s summer camps, chiefly in Picardy. The informants were for the most partmonitors in the camps, and speech was recorded in an informal interview style. Oneor two classic patterns are observable in Coveney’s results: less retention of ne byyounger and working-class speakers; against this, the degree of sex differentiationpresent is very small. However, the principal feature of these results for the purposesof the present discussion is the way in which the rather low level of ne retention isdistributed across all social groups; all speakers, whether differentiated by age, sexor social class, are involved in this area of sociolinguistic variation. The variablepresence of ne therefore fulfils one of the criteria proposed by Coveney (1996, p. 52–54), as being applicable in the identification of grammatical sociolinguistic variables.We need not list all of the criteria here; the requirement that is relevant to the pre-sent discussion (p. 53) is that the area of variation in question should be widely dis-tributed across the community: ‘the variants should be socially differentiated withinthe speech community’; and further ‘the variants must occur in the vernacular styleof at least some speakers in the speech community’. We can suggest that these cri-teria stem from the concern of the variationist programme with the diffusion ofvariation and change through linguistic systems, which by and large are assumed to

N. Armstrong / Language Sciences 24 (2002) 153–173 155

be the property of all of their speakers; items diffusing through the system generallyoriginate in vernacular speech varieties. Exceptions to this pattern of community-wide distribution are provided by ‘dialect divergent’ situations, reported for exampleby Johnston (1983) and Newbrook (1986). In these situations, considerable struc-tural differences between the standard and vernacular varieties produce patterns ofsocial-stylistic variation running contrary to the situation described above. Furtherexceptions are ‘embryonic’ and ‘vestigial’ variables (Trudgill, 1999), i.e. those in theprocess of entering or leaving a speech community; these also show a distributionthat is restricted to certain speaker groups or even individual speakers, and henceare difficult to study using variationist methods.In summary therefore, the patterns observable in the literature suggest that along

the social-group or interspeaker dimension of linguistic variation, ne is a syntacticvariable of the type that can be thought of as being capable of responding to a var-iationist analysis; that is, the variable occurs frequently enough for speakers to beable to employ percentage-wise the (non)standard variant to signal various aspectsof their social identity: their age, gender, social class, etc. A further qualification is ofcourse that the standard variant is used with greater frequency by more linguisticallyconservative speakers (broadly older, some female, middle-class) and this reflects theprestige that accrues to the presence of ne, perhaps principally by virtue of itsinvariable use in conservative written varieties of French. Finally, in the vastmajority of cases no propositional meaning attaches to the presence or absence ofne, although the particle may be inserted to repair a miscomprehension, as reportedby Coveney (1996, p. 56) and discussed in the Dieuze data below. So just as noconceptual meaning is involved in the omission or retention of French schwa in asequence such as pendant la s(e)maine [padalas( e)men], so speakers seem able, inmany linguistic contexts, to exercise a degree of choice in the omission or retentionof ne that is not present, or in a lesser measure, in the case of other syntactic items orconstructions; French interrogatives are perhaps the most notable instance in thisregard (Coveney, 1996; Gadet, 1998). In other words, semantic and pragmatic fac-tors appear to constrain the choice of speakers to a much lesser extent in their vari-able treatment of ne than is the case with many if not most morpho-syntacticvariables, such that ne seems to resemble an item of phonology rather than ofgrammar from the point of view of its interspeaker value and distribution. We turnnow to a consideration of these factors as they influence the intraspeaker axis ofvariation in ne.

3. Intraspeaker variation in ne

Evidence of intraspeaker variation in French ne is largely anecdotal. Coveney(1996, p. 88–89) suggests that hyperstyle grammatical variation in French maycharacterise the behaviour of ne; he reported a massive style shift in his corpus by a 35-year-old intermediate-class male informant, the director of one of the holiday campswhere the language data was being recorded. This informant was interviewed byCoveney on two occasions clearly distinguished by their formality, the subject-matter


discussed, and separated in time: on the first occasion, recorded in his office, theinformant was explaining the way in which the camp was organised; the followingmorning, when the second part of the interview was recorded, the informant andfieldworker were strolling in the grounds of the camp and chatting less formally.Between a 50% ne-retention rate in the first part of the interview and 11.4% in thesecond, the shift is quite considerable, and exceeds the largest degree of groupinterspeaker variation apparent in Coveney’s results, 20.4% between the oldest andyoungest speaker groups. Coveney suggests (1996, p. 89–90) that ‘although weclearly need more information about the social and stylistic differentiation of ne invarious communities, it seems possible that it is [. . .] a ‘‘hyper-style’’ variable’. Cov-eney (ibid.) goes on to draw a parallel between hyperstyle variation in French (i.e.where intraspeaker variation exceeds interspeaker, as in the above result) and thatreported in Tehrani Persian by Jahangiri and Hudson (1982), and noted by Bell(1984) as the only reliable example he found in his comprehensive survey of thesociolinguistic literature. Coveney (ibid.) proposes a comparison in the followingterms:

Contemporary France and Iran might be thought to have little in common, butperhaps one significant sociolinguistic similarity helps to account for the factthat hyper-style variables are found in both societies: formal styles, whichreflect the conservative written language, seem to involve, in certain respects atleast, a quite different type of linguistic behaviour than informal styles in bothFrench and Persian. Bell reports that Persian variables are losing their inter-speaker variation (1984, p. 156). As the loss of the French negative particlenears completion, we might expect that the same process would take place [. . .].

Evidence supporting this proposition of Coveney’s regarding the impending lossof interspeaker variation in the use of ne, and its corresponding acquisition ofhyperstyle value, is provided by the fact that the use of ne now appears to be ‘nolonger part of the vernacular linguistic system which is transmitted from generationto generation by the normal processes of acquisition’ (Coveney 1996, p. 90). This isendorsed by the fact that pre-school French children show near-zero use of ne,acquiring the negative particle only when they start attending school and comeincreasingly into contact with written French. Below we present some results deriv-ing from a quantitative study of intraspeaker variation in ne, with the aim ofthrowing new light on the issue of the stylistic value of the negative particle.

3.1. The Dieuze corpus

Quantitative evidence of the variable treatment ne is presented in Table 1, whichshows patterns of ne-retention in a corpus of spoken French recorded in Dieuze, atown of about 5000 inhabitants in the Moselle departement in the Lorraine region ofnorth-eastern France. The recordings were made in the town’s college et lycee or 11–18 secondary school, in which all the informants were pupils, and where the


researcher had some years previously been English language assistant. The corpus ismore fully described in Armstrong (1993, 1996), but for convenience we mention itspertinent features here. In Dieuze the informants were recorded in two styles,designated henceforward ‘interview’ and ‘conversation’. In interview style, infor-mants were recorded one-to-one with the researcher. Conversation style was elicitedby the use of ‘peer conferences’, i.e. the recording of two or three informants of thesame age and sex, in the absence of the researcher. Interview style was assumed to bethe more formal of the two. Two age groups were recorded: 11–12 and 16–19 years,as well as two gender groups. The corpus was recorded for the purposes of a vari-able phonological analysis, a fact whose implications we discuss below.The sub-sample of 16 informants whose linguistic behaviour is discussed here is

differentiated by the social variables of age and gender. Social class was not included

Table 1

Observed frequencies (n) and percentages of ne retention (%) in the Dieuze speaker sample

Gender/age Interview Conversation

% n % n

Males 16–19

JM 0.9 1/109 1.4 1/72

BR 0.9 1/116 1.4 2/144

BB 0 0/70 0 0/41

ES 5.0 2/40 6.6 3/45

Group average 1.2 4/335 1.9 6/302

Females 16–19

VN 0 0/60 0 0/126

SH 4.6 3/65 1.2 1/78

EM 4.6 4/86 0.9 1/103

CN 7.7 10/129 1.2 2/164

Group average 5.0 17/340 0.8 4/471

Males 11–12

CF 3.3 2/60 0 0/97

AR 0 0/29 0 0/112

JB 0 0/45 0 0/55

BW 0 0/50 0 0/122

Group average 1.1 2/184 0 0/386

Females 11–12

SP 0 0/46 1.3 1/73

SM 0 0/41 0 0/71

NG 3.2 2/62 4.6 3/65

CY 8.3 6/72 1.9 1/53

Group average 3.6 8/221 1.9 5/262

All groups 2.9 31/1080 1.1 15/1421


as a variable, since no consistent patterning of the data with this variable wasobserved. Most informants were in any event drawn from social classes intermediatein the continuum, and in view of their age the indirect relation between the infor-mants’ social class, determined by reference to that of their father or mother, andtheir linguistic behaviour is perhaps unsurprising, given the often close relationbetween occupation and social class.Before discussing results, we examine briefly the issue of tokens excluded because

of invariance and other factors.

3.2. Identification difficulties

It is worth remarking firstly, in the context of the light that methodological issuescan throw upon changes proceeding in colloquial French, that the figures in the ‘n’columns in Table 1 refer, not to all tokens of omissible ne, but to those which werecapable of being unambiguously identified as occurring at loci of variation. Oneprincipal context which presents identification difficulties is that where ne potentiallyfollows a lexical item having a terminal /n/ and precedes a vowel-initial word. This isof course because ne reduces to /n/ standardly before a vowel, as in: on (n’) a pas deresponsabilite. At the same time liaison is near-invariable in on, so that the /n/ isalmost always pronounced. The preceding example (from the Dieuze data) reduced,so far as it is possible to determine without recourse to instruments, to the followingform: on a pas de responsabilite. However, following the practice of Coveney (1996,p. 66), tokens were excluded in this context in view of the difficulty of determiningwith certainty the presence or absence of ne. As Coveney remarks, speakers mayoccasionally signal the presence of ne in these contexts in a relatively easily percep-tible way, perhaps most often through a lengthened /n/, but no such instances werenoted in the Dieuze corpus. Two further ambiguous instances are as follows: jamaison (n’) irait taper sur quelqu’un; on (n’) est pas dans un lycee technique. It was theauthor’s impression that a lengthened /n/ was present in these utterances, but thetokens were nevertheless excluded in the absence, as noted above, of a reliablemeans of identification. A similar difficulty of identification can occur after other/n/-final negative items such as rien, aucun and personne, as well of course as any /n/-final word preceding ne+vowel-initial word. This problem is worthy of mentionbecause of its sociolinguistic interest; as a result of the pronoun’s frequency, pre-ceding on is by far the commonest locus in the Dieuze corpus that presents thisidentification difficulty. This of course reflects the virtual absence of the expressionof 4th-person subject reference through the use of nous+4th-person verb form ineveryday spoken French (cf. Coveney, 2000).We may mention further that several tokens were excluded because, again in the

absence of instruments, it proved impossible to distinguish between /l/ and /n/ insequences such as: mon pere i(l) (n’) a pas voulu qu je l fasse (from the Dieuze data).Coveney (1996, p. 81) reports a 2.5% retention rate in this context, so that it is verylikely that ne was deleted in these instances in the Dieuze corpus. The proximity ofthe place of articulation of /l/ and /n/, as well as acoustic similarities between the


sounds, account perhaps for this identification difficulty. The presence of back-ground noise was also a factor in these cases.

3.3. Other excluded tokens

Aside from identification problems, the further issue influencing the decision toexclude loci is that of invariance. Pre-formed sequences such as n’importe and neserait-ce, where the speaker clearly has no choice regarding the retention of ne, wererare in the corpus; the issue of invariance presented itself rather in connexion withinvariable absence. Clear examples are those such as ca va pas! or ca va pas, non?perhaps suitably translated into English as ‘Forget it!’, ‘No way!’ or ‘You’re joking!’.These sequences appear not to offer speakers choice over the presence or absence ofne, and were therefore excluded. More numerous examples concern reflexes of je saispas, which reduce to [

Ðepa] or [

Ðepa] or even [

Ð:pa]. These forms were used by

several speakers, seemingly without literal force, for various pragmatic purposes: asa filler, to mitigate, and sometimes as a particle expressing indifference, with thepragmatic force of peut-etre, as in: tu peux me l donner demain [

Ðepa] moi. Non-

standard phonological reductions are given henceforward in ordinary spelling, as inthe previous example: il > i; me > m; vous > v; tu > t; je > j; etc. Given the easeof identifying in context the difference between tokens of [

Ðepa] used as a negative

response to a direct question and those used epistemically, these latter were excludedfrom the quantification given above. It seems defensible to exclude these for thefurther reason that their use is idiolectal rather than sociolinguistic, some speakersusing them considerably more than others. This is pertinent here because groupresults are discussed below. It should be borne in mind that the exclusion of tokensof [

Ðepa] contributes in some measure to the quite considerable variation in tokens

numbers between individual speakers, an issue we discuss in detail below.It may be pointed out finally in this connexion that several speakers tended to

alternate between a radically reduced form of je ne sais pas with literal force, such as[Ðepa] in conversation style, and the fuller reflex je sais pas in interview, thereby

displaying style shift between the forms that appeared to be more readily availableto them than the seemingly now rather formal je ne sais pas.A further potentially categorical locus concerns the sequence pas mal, literally of

course ‘not bad’ but now perhaps lexicalised to the equivalent of bien or beaucoupfor some speakers. Following Coveney (1996, p. 69), instances of pas mal wereincluded as potential loci for use of ne where the literal force was clearly translatableas ‘not bad’ (as in il est pas mal), rather than as equivalent to beaucoup, as in on avaitpas mal discute. Elliptical sequences with no verb (pas mal) were of course excluded,as containing no locus for variation.We turn now to a discussion of the results for ne in the Dieuze corpus.

4. Patterns of variation in ne in the Dieuze corpus: linguistic factors

Table 1 shows percentages of ne retention in a sample of 16 Dieuze informants.Most immediately observable in Table 1 are the near-zero levels of retention for all


speaker groups, broadly irrespective of speech style. Nevertheless, rather modestdegrees of style shift in the expected direction by the female groups stand in fairlysharp contrast to the treatment of ne by both male groups, who show negligible orindeed negative style shift. These results are in conformity with many others in theliterature that show especially female speakers’ linguistic behaviour approximatingmore closely to the conservative standard in more formal styles. Corresponding tothis is the fact of the younger males’ producing just two tokens of ne out of apotential 570. To this extent the results are in line with expectations: nevertheless,where they do occur, degrees of use of ne and of style shift are very modest, and thiscan be explained in various related ways. One explanation refers to the developmentof communicative competence in the Dieuze speaker sample, who can plausibly beassumed to have had no substantial experience of performing in highly formal,principally public situations of the type that have been reported as triggering veryhigh rates of ne: sermons and lectures, for instance (Pohl, 1975, p. 21). Another wayof explaining these low degrees of style shift is to suggest that the majority of thespeakers did not feel that the two speech styles were sufficiently differentiated informality for high rates of ne to be appropriate in interview style. Against this it canbe pointed out that large degrees of style shift have been found in the corpus forvariation in lexis (Armstrong, 1998a) and some phonological contexts (Armstrong,1996). The obvious conclusion is that stylistic competence emerges at different rateson the different levels of linguistic analysis, and hence that this speaker sample doesnot yet use ne to any substantial extent. This suggestion receives some support fromthe near-zero rates of use of and style shift in variable liaison observed in the corpus(Armstrong, 1993, p. 250). Variable liaison is comparable to ne in being an area ofvariation in French that is linked to literacy and associated with highly formalspeech styles. A further, related line of explanation refers to the pragmatic level ofanalysis; we develop this below.Several speakers did feel impelled to employ ne somewhat more frequently in

interview than conversation (although style shift is nowhere dramatic) and the phe-nomena of interest underlying these results are both linguistic and social. Turningfirst to the linguistic factors associated with style shift, several researchers havenoted (Moreau, 1986; Coveney, 1996) that deletion of ne when it follows a subjectNP, as it overwhelmingly does in the Dieuze corpus, is constrained linguistically bythe nature of the NP; broadly, a preceding subject pronoun, either conjunctive ordisjunctive, favours deletion, while a more complex lexical NP subject promotesretention of ne. Thus of the 46 instances of ne retention shown in Table 1 in the ‘Allgroups’ row, 34 were preceded by a non-pronoun subject such as a lexical NP withor without a following relative, a demonstrative such as celles followed by a relativepronoun, or a disjunctive pronoun such as eux. Of the remaining 12 tokens of ne, 11were preceded by a clitic pronoun, and the remaining one was associated with animperative sentence.We shall discuss the interrelation of linguistic and social factors by concentrating

chiefly on the use of ne by one speaker, the older female informant CN. This infor-mant is suitable because she was talkative, and hence produced a copious volume ofspeech in both styles, with several interesting examples. She also produced style shift


in the expected direction. The instances of insertion of ne that she produced are alsorepresentative of those produced by other members of the speaker sample. Of the 10tokens of ne produced by CN in interview style, three followed a clitic, while sevenfollowed a non-pronoun subject: either a lexical NP:

(1) les profs n’ont pas la loi(2) les parents ne savent pas(3) parce que [l]es sujets ne nous plaisaient pas trop quoi(4) les cours n’etaient pas comme ici

demonstrative plus relative:

(5) mais celles qui n’ont pas l permis i (sic) doivent s debrouiller toutes seules

disjunctive pronoun:

(6) et eux n’acceptent pas les Continentaux

or inverted negative item:

(7) personne ne l’a encore fait

The negative particle was clearly perceptible in (7) as [n e], providing an exceptionto the type of identification difficulty discussed in Section 3.2.In addition to the seven non-pronoun NP loci that triggered ne, only one other

associated with ne was observed in CN’s speech in the interview; this was:

(8) une prof qui faisait rien

Clearly there is an association here between a tendency to use lexical or other non-pronoun NPs and the nature of the speech situation; if CN’s production is repre-sentative, non-pronoun NPs are rare in speech that is not highly formal. They arecertainly rare in the two styles recorded in Dieuze. A figure of eight non-pronounsubjects out of 129 potential loci for insertion of ne is rather spectacularly low(6.2%), but a comparison with CN’s speech in conversation style is even morerevealing: out of 164 potential loci, only three (1.8%) can be unambiguouslyinterpreted as potential lexical NPs, and none of these triggered ne. These were asfollows:

(9) celui d’avant i marchait pas(10) je faisais celle qui comprenait rien quoi(11) je dis: ‘Madame, les X marchent pas’

Regarding the first example cited immediately above, we can remark that thesyntax is in line with the tendency in French, frequently noted (e.g. Harris, 1988, p.231–232), for clitic subject pronouns to occur even where there is already an NPsubject leftwards in the string. This process is said to have come about as a result ofthe non-contrastive character of most French present-tense verb paradigms, owingto the fact that five out of the seven verb forms are homophonous. Harris suggests


that this homophony has resulted in the cliticisation of the conjunctive personalpronouns, at least in non-formal speech, so as to ensure contrast of person. Hestates that:[. . .] we may regard French ils aiment [izem] ‘they love’ as one polymorphemic

word (subject-prefix+stem) in exactly the same way as one regards Latin AMANT,or Old French aiment, as one polymorphemic word (stem+subject suffix).Against Harris’s view, it should be pointed out that an interpretation of ils aiment

as ‘one polymorphemic word’ is contradicted by the fact that the sequence can beinterrupted, although only by other pronouns and of course ne. With regard to useof ne, the result of this very frequent retention of the clitic subject pronoun beforethe verb is perhaps that irrespective of the presence of a preceding NP subject,speakers are in the great majority of cases operating with the quasi-fused cli-tic+verb sequence, and that this is a constraint that greatly disfavours the use of ne.It is clear from CN’s use in interview style of full NPs without pronoun doublingthat this constraint can be overridden where greater degrees of self-monitoring arerequired, and where the form which is perhaps the default ([NP]+clitic+verb) is feltto be inappropriate. Similarly, (7) above could well in more informal speech havebeen organised using clitic+verb: il [n’] y a personne qui l’a encore fait, i.e. as asequence beginning in principle with a subject clitic, even if in everyday speech thepronoun in il y a is rarely realised. From a sociolinguistic viewpoint, it appears likelythat forms of the NP+clitic+verb type, because of their surface pleonastic struc-ture, are the object of disapproval by teachers and others concerned with exertingnormative pressure.Sequence (10) above is quite closely comparable to (5), celles qui n’ont pas l permis,

the example produced by CN in interview style. The pertinent difference is of coursethat the second utterance has the sequence qui+ont, which Coveney (1996, p. 80)reports as a context favouring ne retention in his corpus as a result of the juxta-position of vowels consequent on deletion; the syntax of the two sequences isotherwise similar. We may suggest therefore that these two sequences are sufficientlycomparable to illustrate the operation across the two speech styles of ‘audiencedesign’ in a straightforward sense; the adaptation of an utterance in response to thestatus of the addressee.Regarding the factors influencing use of ne after a clitic without preceding NP, of

the three cases produced by CN in interview style, one was a case of ne inserted torepair a miscomprehension, as follows, where FW=the fieldworker and CN=theinformant:

(13) CN: je m foule pasFW: comment?CN: je n me foule pas.

The negative particle was inserted here as an aid to comprehension (the phrase sefouler has the sense, in colloquial European French, of ‘to work hard’).A further example appears to have been influenced by the ‘serial effect’ that has

been reported as tending to produce a string of standard or non-standard variants ina particular stretch of discourse (Cheshire, 1997, p. 6):


(14) et eux n’acceptent pas les Continentaux comme i disent/i ne veulent que desCorses

It is possible here that the relative proximity of the first occurrence of ne hasinfluenced the second. A further factor may be the higher probability of ne retentionassociated with que; Coveney (1996, p. 76) reports 34.9% retention before que,compared with 16.6% before pas, while Ashby (1981, p. 678) has 59% compared to33%, a roughly similar ratio.The third instance of ne occurring after a clitic is as follows:

(15) et l’annee derniere i n’avaient pas euh (pause)

This instance seems recalcitrant to explanation in terms of the surrounding dis-course; we mentioned in Section 3.1 above the identification difficulty concerningsequences such as i(l) (n’) a pas voulu, where some speakers appeared to insert [n]rather than the [l] of the pronoun. CN’s utterance was very clear and showed anunambiguous token of ne. The explanation advanced by Coveney (1996, p. 81) maybe applicable here; he noted a higher retention rate (30%) than the average (18.8%)in his Somme corpus in the sequence ils avaient, and suggested that such sequences‘involve a more or less obligatory liaison if ne is omitted, and it may be that speakersfeel that this would highlight the absence of the negative particle, which is still, ofcourse, considered a non-standard feature’.In the following sections we examine in more detail the embedding of ne in

discourse, from the viewpoint of microstyle variation and audience design.

4.1. Issues connected with speaker choice: the importance of token numbers

It will be seen in Table 1 that numbers of tokens of ne are rather modest for mostspeakers, at least compared to those reported by other researchers (Coveney, 1996,p. 86 has an average of 105 tokens per speaker). The mean number of tokens perspeaker in both styles in Table 1 is 2501�32 or some 78. It can be seen further thatthe value for N varies across speaker groups, individual speakers and styles. Thisvariation is in line with that observed for the other levels of linguistic analysis stu-died in the Dieuze data (phonology and lexis), and is probably caused in part by theelicitation methods used. Especially younger children (and perhaps above allyounger boys) do not in their spare time stand about chatting sedately, respectingthe rules of turn-taking and politeness as observed by adults; the younger boysespecially often talked over one another, making comprehension difficult. Several ofthe younger informants became restless during the peer conferences and rangedaround the room, interfered with the recording equipment, or engaged in activitiesnear the equipment that caused a good deal of background noise, thereby makingsections of the recordings very indistinct or even inaudible. So in spite of their valuein eliciting spontaneous and often very informal data, the peer-conference record-ings were no doubt less effective than more sophisticated methods (perhaps mostnotably a radio microphone) in capturing considerable volumes of clear speech.Informants were also advised to turn off the tape-recorder in the peer conferenceswhen periods of silences intervened. Furthermore, the first 10 minutes of the infor-


mal recordings were excluded from study, on the assumption that informants wouldhave become less constrained by the presence of the recording equipment after thislapse of time. All of these factors contributed to the reduction of the volume ofspeech recorded in conversation style. For the analysis of variable phonology, thiswas not crucially important, but it is evident that the result is rather modest tokensnumbers for ne for many informants, and especially for the younger groups.Nevertheless it can be seen that the mean number of tokens per speaker in con-versation style is appreciably higher than in interview (89 against 68) and this sug-gests that speakers were more relaxed in the conversations, at least judging by thevolume of talk produced; that is, they experienced the conversations as being lessformal speech events.Turning to the interviews, each lasted between 30 and 45 minutes, and it can be

seen that tokens of ne analysed for each speaker group vary quite considerably.Although individual responses to the interviewer’s questions varied a good deal, oneat least of the older females (CN) produced notably more tokens than many other ofthe older speakers, and to some extent this reflected the older females’ style ofinteraction in interview (and indeed conversation) style, which was in general morecooperative than the other groups’; the older and younger males were often contentwith informative but brief answers to the interviewer’s questions and comments,while some of the younger females appeared to be slightly intimidated by the inter-view conditions, and hence often responded rather minimally. These differences indiscourse style across the speaker groups are pertinent to a consideration of thediscourse patterns influencing the rates of ne deletion indicated in the figures shownin Table 1; we discuss these patterns in the following section. Firstly however weconsider a theoretical issue underlying the relative paucity of tokens of ne availablein the Dieuze data.We mentioned in Section 2 that ne comparable to omissible phonological variables

in its frequency and lack of involvement in propositional meaning. Once again, whatobviously distinguishes ne is its situation in variation at the syntactic level, andthrough syntax to discourse; a low-level example has already been discussed above,that of [

Ðepa] used as a discourse particle. This instance raises the question of the

level of analysis at which speaker choice in the use of ne is operating. A comparisonwith variable phonology is again illuminating. A basic axiom of the variationistapproach is of course that speakers ‘locate themselves socially in a multi-dimen-sional space, as an ‘‘act of identity’’’ (Hudson, 1996, p. 207) through the probabil-istic use of phonological variants, among other means. The substantial amount ofevidence amassed so far indicates that speakers do this in a regular and predictableway, so that only a relatively restricted number of tokens of any phonological vari-able needs to be quantified for a reliable analysis to be performed for a given speechstyle. Milroy (1987, p. 134) suggests a target of some 30 tokens per speaker, pervariable in the linguistic environment of interest. This figure is arrived at on the basisof ‘[. . .] general statistical laws; if the number of tokens is lower than 10, there is astrong likelihood of random fluctuation, while a figure higher than 10 movestowards 90 per cent conformity with the predicted norm, rising to 100 per centwith 35 tokens.’ (Milroy, ibid.). This statistical view suggests that sociolinguistic


behaviour on the variable phonological level proceeds largely outside a speaker’sconscious control, and can thus be compared with other types of quasi-consciousbehaviour that are generally amenable to relatively straightforward quantificationand have a distribution in the relevant population that can be modelled statistically.This proposition appears not to be tenable in regard to variation in ne, or at least

to a lesser extent. This is because variation in the use of ne is closely tied to factorsthat are influential at the discourse level of analysis, perhaps most notably to topicand tone; by this latter term is meant here the attitude adopted by a speaker towardsthe topic under discussion. The underlying assumption here is that tone is mostsaliently manipulated at the levels of grammar and lexis, but it is worth noting thatthe use of a vowel or consonant in an unexpected context, or the use of a phonolo-gical stereotype (Chambers and Trudgill, 1998, p. 75), can also serve to subvert orreinforce the tone of an interchange in a way that is closely comparable to whatoccurs on the more salient linguistic levels.The issues of topic and tone have been analysed in connexion with lexical varia-

tion in the Dieuze data (Armstrong, 1998a); clearly, a topic that is relatively seriouswill tend to call for more bookish vocabulary, as well as a tone that is towards theconservative end of the conservative-subversive continuum (Armstrong, 1998b, p.142). In regard to conversational tone, Holmes (1997, p. 207) points out the need tobear in mind that the analysis of any stretch of linguistic behaviour is a snapshot,and that speakers are continually engaged in a process of construction of theiridentity that leads them to present different aspects of this identity in response tovarious motivating factors: ‘gender [to which we may add all social] identity is con-stantly being constructed, and people may reinforce norms at one point, but chal-lenge and contest them at others’. This quotation is designed to illustrate what ismeant above by the conservative–subversive continuum of tone or orientation,which speakers can be assumed to modulate continually in the course of an inter-action.It is true furthermore that the various factors influencing speech production do

not necessarily coincide: the topic under discussion may be serious, but the tonesubversive and the speech situation informal. Leaving aside the question of con-versational tone, these factors raised serious problems of comparability in theDieuze data on the level of variable lexis; for instance, one discussion in conversa-tion style between 16-year-old girls about boys and dating called for much use of theinformal term mec; unsurprisingly, no such discussion occurred interview style. Theclose involvement of lexis in variation in choice of topic puts serious difficulties inthe way of any attempt to set up Labovian-type lexical variables having two or morevariants, as in mec�garcon. Intraspeaker variation in ne is also mediated throughthe factors of topic and tone, and in ways that appear to be just as complex as whatobtains in lexis.We examine in Section 5, from the point of view of its influence upon ne deletion

rates, how the tone of a narrative can work against the topic and style. We note herehowever that the issue of token numbers, referred to initially in this discussion, isrelatively unimportant for the reasons connected with topic and tone; clearly, areasonably adequate number is required for a reliable quantification of ne, but in


view of the unpredictable nature of the variation in topic and tone found in theDieuze corpus it is unclear whether differentiation in these factors across a very largecorpus would ever even out sufficiently to give a fully representative picture of theuse of ne as it interacts with discourse through its situation in syntax.Several researchers have remarked the influence of topic upon rates of ne deletion;

Sankoff and Vincent (1980, p. 302–303) noted that the rare cases of use of ne bytheir Montreal informants tended to occur when the topic was a serious one, such asreligion or education. As pointed out previously (Armstrong, 1998b), a principaldifference between the two speech styles in the Dieuze data is that while both inter-view and conversation styles were broadly devoted to discussion and narratives,conversation style was often characterised by an emphasis on the affective aspect ofthe topic under discussion, with interviews tending to be more narrowly focused onthe informational dimension of variation. In interview style, informants typicallytalked about school, school subjects, their family, their home town or village etc., inresponse to the interviewer’s questions. The speech thus elicited can therefore bedefined as being broadly ‘message-oriented’, using Brown’s (1982, p. 77) definition,where the aim of the interchanges tends to be the ‘communication of a propositionalor cognitive (information-bearing) message to the listener’. Brown distinguishesbetween this type of speech and ‘listener-oriented’ speech, where, as seemed often tobe the case in the Dieuze conversations, the goal is often to maintain friendly rela-tions, and where ‘it is often the case that speakers [. . .] don’t seem to be talkingabout anything very much’. The distinction between interview and conversationstyles in the Dieuze corpus is probably less sharp than that expressed by Brown, andthere will of course be a continuum between these two poles. In any event it is diffi-cult to imagine an extended interchange that is wholly non-informational. The dis-tinction between the two styles in the Dieuze data might be more aptly expressed interms of the pragmatic use which the speakers make of the subjects being discussed.Thus the speech elicited in interview style may be broadly described as message-oriented; by contrast, speakers in conversation style, although they discuss specificsubjects, often appear to do so in order to express or produce an emotional effect. Amore suitable broad distinction between interview and conversation styles mighttherefore be expressed respectively in terms of an ‘ideational’ or ‘propositional’orientation on the one hand, and on the other an ‘affective’ approach.This ideational–affective polarity raises the question whether the variation shown

in Table 1, as opposed to being ‘sociolinguistic’ in the sense used here so far, isindicative rather of variation caused by a somewhat different type of audience designthat reflects an assumption on the informants’ part of less shared knowledge, bothlinguistic and non-linguistic, on the part of the fieldworker, as a result of his non-native-speaker status. This view would still assume Bell’s analysis of self-monitoringas a mechanism whose function is to ensure the accuracy of audience design; but theDieuze situation recalls also a Bernsteinian duality of ‘restricted’ and ‘elaborated’language production, rather than the formal–informal polarity postulated by Bellwhich refers essentially to the stimulation of language production in response to thedegree of intimacy subsisting between speakers, and perhaps more importantly thepower relations that obtain in any given interaction. In the following sub-section we


summarise briefly Bell’s audience design hypothesis, relating it subsequently tointraspeaker variation in ne.

4.2. Intraspeaker value of ne and audience design

Bell’s audience design principle (1984) broadly states that speakers adapt theirspeech to the perceived social status of their interlocutors, in order to gain approvalor acceptance through accommodation. Formal speech situations call for speechthat derives from the prestigious end of the social-class continuum; in this perspec-tive Bell’s principle seems to apply most closely to a language such as English,because speakers employ the mechanism of self-monitoring to select (among others)certain arbitrary, highly prized phonological variables that they feel to be appro-priate for the purposes of accommodation to their interlocutor’s social status.Arbitrary variables are those whose use or avoidance cannot be explained by refer-ence to ease-of-articulation considerations. Self-monitoring is involved because ofthe awareness on the part of speakers of the disapproval to which certain social-regional variables are subject. Speakers therefore avoid these variables in more for-mal styles: that is, when speaking to interlocutors they do not know well, and whosesocial-linguistic attitudes they therefore can only gauge indirectly; or to speakerswhose approval they wish to canvass, perhaps because an asymmetrical power rela-tion obtains. Interspeaker variation exceeds intraspeaker, firstly because no onespeaker or speaker group controls the entire range of socially–conditioned languageconfidently; a further, perhaps more important factor is that short-term linguisticaccommodation is a complex process that often does not imply mere imitation on aspeaker’s part of an interlocutor’s language production, but rather a degree ofcompromise, although close imitation can happen in service encounters where thedistribution of power between interactants is highly asymmetrical, as illustrated byCoupland’s (1980) well-known travel-agency study.At first sight, the foregoing account of the psychosocial motivations that propel

style shift would appear to apply to ne in quite a satisfactory way. As stated above,formal situations calls for speech that derives from the ‘standard’ or prestige end ofthe social-class continuum of language varieties. Linguistic items in these varietiesderive prestige, either because their use is associated with social groups whosebehaviour is highly prized, or because they are associated with the standard writtenlanguage; these factors will often coincide, of course. As noted by Labov (1994, p.347), an English example of the former type of variable is the pair of vowels dis-tributed in lexical splits exemplified by pairs such as ‘putt’ and ‘put’, which are dif-ferentiated in southern/standard English by two distinct back vowels, respectively[�] and [

�

], and whose distribution is not reliably marked in the orthography. Thus,the guidance provided by the spelling of ‘putt’ [p�t] and ‘put’ [p

�

t] is contradictedby ‘puss’ (standard realisation [p

�

s]) and ‘pus’ ([p�s]). The French variable ne isclearly of the latter type, whose clear representation in writing can be presumed toenhance its socio-stylistic value. As stated previously, Coveney has suggested (1996,p. 89) that ‘formal styles [in France and Iran], which reflect the conservative writtenlanguage, seem to involve, in certain respects at least, a quite different type of


linguistic behaviour than informal styles’. In face-to-face, unscripted interactions, itseems likely that the influence of writing can influence style shift in a more formaldirection because an association exists in speakers’ minds between formal, non-inti-mate speech situations and the type of linguistic production expected in highly nor-mative situations such as the classroom.A related explanation refers to the influence of the interviewer; in contrast to

many sociolinguistic studies that aimed to study intraspeaker variation, the Dieuzesurvey compared variation across two unscripted speech styles differentiated by theinteractants involved, rather than across scripted and unscripted styles. Regardingthe latter type of variation, we may mention Milroy’s (1987, p. 173) formulation that‘the contrast between spontaneous speech and reading styles is a simulation, forexperimental purposes, of differences in the amount of attention paid to speechwhich crop up in natural interaction’. Regarding the elicitation of unscripted styles,Giles (1973) has the interesting criticism that researchers interviewing informants inthe context of a sociolinguistic interview may be responsible for inducing style shiftby themselves unconsciously accommodating their speech in the direction of theproduction they expect to elicit. This is of course a reflex of the observer’s paradox,expressed in somewhat subtler terms than is customary, and can plausibly be arguedto be a natural tendency that one would expect to occur in non-experimental con-ditions; in any event it is difficult to see how it could be resisted in experimentalconditions, given the unplanned nature of most unscripted speech production. Iftrue, however, one might presume that the tendency that Giles postulates wouldhave the effect of compressing shift between unscripted styles to degrees smaller thanmight have been expected otherwise. This is because a tape-recorded interviewbetween interviewer and informant might be expected to produce, following theconcern of the fieldworker to elicit an adequate volume of language data, moreinformal speech than is implied by the label ‘formal style’ used by Trudgill, forexample (1974, p. 46–47), following the model of Labov (1966). In contrast, casualstyle in the interview is defined as occurring in response to cues such as interactionswith speakers other than the interviewer, and narratives that induce the informant toreduce self-monitoring, such as the celebrated danger-of death stimulus.Whether or not Giles’s objection is generally valid (for instance, Trudgill,

1986, p. 5–11 devotes considerable space to rebutting it with regard to hisown Norwich results), the salient difference between the elicitation methodsdiscussed above and those used in the Dieuze survey is clearly that in Dieuze theinformal language data were collected in peer dyads or triads and in the researcher’sabsence, while the formal data were elicited through an approximation to thestandard sociolinguistic interview, with informants interviewed one-to-one withthe researcher. The significance of the fieldworker’s non-native-speaker status forthe purposes of the present discussion is that this status may well have broughtabout more formal interview conditions than are common, as a result of the field-worker’s relatively limited capacity to operate the type of accommodation postu-lated by Giles, at least compared to that of a native speaker. It seems plausibletherefore that the fieldworker’s speech may have contributed substantially, in theminds of a majority of the informants, to the creation of a rather highly formal


speech situation, analogous perhaps to a highly ‘tense’ situation such as a jobinterview.However this may be, the result of the Dieuze elicitation methods was, as previously

noted, rather high rates of style shift at the levels of some variable phonology and oflexis. We have seen that degrees of style shift for ne are very modest, but the typicalcase of CN discussed in Section 4 demonstrates that shift across the two stylesreflects in part, through a somewhat more frequent use of NP subjects in interviewthan in conversation, a concern on the informant’s part to issue decontextualisedinformation to an interlocutor who did not share the same knowledge base. Corre-spondingly, the more frequent use of clitic subjects in the conversation stylereflects the more extended nature of the discussions and narratives that took place;for example, at one point in a conversation CN and her interlocutors engaged ina comparison of their experiences when taking driving lessons that lasted some15 min. Clearly, an extended discussion implies more frequent use of pronounsubjects to the extent that the participants will tend to be aware of the identity of thereferents. By contrast, discussions in the interviews generally covered more topics inless depth.As stated previously, we examine below some examples of insertion of ne that

illustrate how the tone of a discussion can work against its topic and style.

5. Relation between discourse and intraspeaker variation in ne

Following on from the issues discussed above is the indirect relation, mediatedthrough discourse structure, between rates of ne deletion and audience design. InSection 4 we discussed the cases of the use of ne by CN in interview style, suggestingthat the nature of the speech situation was such as to work against the strong con-straint that induces speakers to produce the clitic+verb sequence as perhaps thedefault form. In the previous section we suggested, more specifically, that the per-ceived lack of shared knowledge between fieldworker and informants was respon-sible, through the more frequent use of NP subjects, for higher rates of ne. We turnnow to use of ne in the conversation style, again referring principally to instancesproduced by CN.Two examples illustrate the ‘metaphorical’ use of ne in formal style quite clearly.

The first concerns CN, who in the course of discussing a fellow pupil who was quiteseriously overweight remarked: elle ne peut pas bouger. The use of ne here appears toreflect a concern on CN’s part to adopt a serious, compassionate tone, and this wasendorsed by the intonation and ‘warm’ voice quality employed. An even morestriking example was produced by EM, another older female. She was describing inconversation style her close relationship with a female cousin who lived in anothervillage some distance away, and with whom she regularly exchanged very long let-ters. At one point she remarked on the intimacy of the letters; she customarily dis-cussed with her cousin any topic, however personal: comme si les kilometres ne nousseparaient pas. Clearly, we cannot state with certainty which factors account for theuse of ne in a given sequence, but we may note here the conversational tone adopted


by the speaker, who clearly is taking the subject very seriously, and the use of newhere its deletion would jar with the almost solemn effect aimed at.We can draw a parallel here with the concept of ‘metaphorical code-switching’

formulated by Blom and Gumperz (1972) in order to account for switching betweencodes practised by bilinguals, of the type which cannot be explained by a change insituational factors but which is prompted rather by the desire to ‘convey specialcommunicative intent’ (Li, 1998, p. 156). Clearly, any comparison between aninstance of intra-language style-shift and what is happening in cross-language code-switch is problematic, perhaps most notably in regard to the level of salience thatobtains in each case. A related question is whether the formal and formal varieties ofFrench are separate for their speakers in a way that is true of clearly delimited lan-guages. But as Milroy (1987, p. 171) suggests, ‘although bilingual or bidialectalswitching is a more clearly visible process than monolingual style-shifting, the psy-chosocial dynamics underlying these different kinds of intra-speaker variation aresimilar’. It is possible to quarrel with this proposition in its application to everyrespect in which style-shifting and code-switching are comparable, but the quotationemphasises the aspect of intra-sentential style-shifting and code-switching that is ofinterest here: the psychosocial motivation responsible for the selection of a givenlanguage, dialect or variety fragment. It appears plausible that the motivationspropelling instances of style-shift referred to here as ‘metaphorical’ are similar tothose associated with metaphorical code-switching. Certainly the formulation citedabove, referring to the intention to desire to ‘convey special communicative intent’,captures very well the motivation underlying insertion of ne in the instancesdiscussed immediately above.

6. Summary and conclusion

The results, data and argumentation presented above can be summarised as fol-lows: a Labovian analysis of intraspeaker variation in ne needs to be supplementedby qualitative, discourse-analytic methods. This is because of the situation of ne ingrammar, which in turn is indissociable from the discourse level. Several of the sty-listic effects produced by the Dieuze informants in conversation style through theiruse of ne (those given above are representative) contradict a simplex formal–infor-mal analysis of style variation. Micro-style variation, present of course in both of thebroad styles discussed here, is reflected in the use of ne through a reduction in thedegrees of style shift, since the ‘formal’ episodes in conversation style call for the useof ne quite frequently relative to interview style.In the context of the relation between variation in grammar and its role in dis-

course structure, we may note here the suggestion of Cheshire (1997, p. 5) to theeffect that English non-standard morpho-syntactic phenomena, such as subject-verbconcord, are more frequent in cognitively prominent sentence types such as inter-rogatives and negatives, which have ‘a more direct connection to the context of face-to-face interaction than others’ by virtue of their involving the hearer more closelythan less prominent structures. Thus variation in English grammar appears also be


more tightly constrained than in phonology, by reason of its localisation in promi-nent, high-involvement speech acts. If true, this reinforces our suggestion of a rathercomplex relation between the intraspeaker and interspeaker dimensions wheregrammatical variation in French ne is concerned. From the point of view of varia-tion in its relation to change, Cheshire argues that low-involvement sentence typesare more susceptible to penetration by standard grammatical forms than lessprominent constructions.From this latter point of view, diachronic deletion of ne in French is of course a

process that is largely complete; or perhaps still in progress and proceeding ratherslowly. One consequence is that the interspeaker value of ne is rather limited.Nevertheless we can draw a parallel with Cheshire’s hypothesis by pointing out thatthe insertion of ne in everyday French is associated with constructions such as lexicalNPs, which can be assumed to be cognitively prominent by virtue of their relativecomplexity and infrequency. Speakers can often if they wish avoid such construc-tions through NP doubling, as we have seen above. But the ‘productive use of ne’(Sankoff and Vincent, 1980) remains available as a stylistic resource. What we haveattempted to present here in a synthetic way is the rather complex nature of thevarious components of style, as well as of their interactions with the syntacticconstraints influencing the choices speakers exercise in their use of ne.

References

Armstrong, N., 1993. A study of phonological variation in French secondary school pupils. PhD thesis,

University of Newcastle upon Tyne.

Armstrong, N., 1996. Deletion of French /l/: linguistic, social and stylistic factors. Journal of French

Language Studies 6 (1), 1–21.

Armstrong, N., 1998a. La variation sociolinguistique dans le lexique francais. Zeitschrift fur romanische.

Philologie 114 (3), 462–495.

Armstrong, N., 1998b. The sociolinguistic gender pattern in French: a comparison of two linguistic levels.

Journal of French Language Studies 8 (2), 139–158.

Ashby, W., 1981. The loss of the negative particle ne in French: a syntactic change in progress. Language

57, 674–687.

Ashby, W., 1991. When does variation indicate linguistic change in progress? Journal of French Language

Studies 1 (1), 1–19.

Bell, A., 1984. Language style as audience design. Language in Society 13 (2), 145–204.

Blom, J.-P., Gumperz, J., 1972. Social meaning in linguistic structure: code-switching in Norway. In:

Gumperz, J., Hymes, D. (Eds.), Directions in Sociolinguistics. Holt, Rhinehart and Wilson, New York,

pp. 407–434.

Brown, G., 1982. The spoken language. In: Carter, R. (Ed.), Linguistics and the Teacher. Routledge and

Kegan Paul, London, pp. 75–87.

Chambers, J.K., Trudgill, P., 1998. Dialectology, 2nd Edition. Cambridge University Press, Cambridge.

Cheshire, J., 1997. Syntactic variation and the concept of prominence. In: Klemola, J., Kyto, M., Rissa-

nen, M. (Eds.), Speech Past and Present. Studies in English Dialectology in Memory of Ossi Ihalainen.

Peter Lang, Bern, pp. 1–17.

Coupland, N., 1980. Style-shifting in a Cardiff work-setting. Language in Society 9, 1–12.

Coveney, A.B., 1996. Variability in Spoken French. A Sociolinguistic Study of Interrogation and

Negation. Elm Bank, Exeter.


Coveney, A.B., 1998. Awareness of linguistic constraints on variable ne omission. Journal of French

Language Studies 8 (2), 159–187.

Coveney, A. B. (2000). Vestiges of NOUS and 1st person plural verbs in informal spoken French.

Language Sciences 22(3), 447–481.

Gadet, F., 1998. La variation, plus qu’une ecume. Langue Francaise 115, 5–18.

Giles, H., 1973. Accent mobility: a model and some data. Anthropological Linguistics 15, 87–105.

Harris, M., 1988. French. In: Harris, M., Vincent, N. (Eds.), The Romance Languages. Croom Helm,

London, pp. 209–245.

Holmes, J., 1997. Women, language and identity. Journal of Sociolinguistics 2 (1), 195–223.

Hudson, R.A., 1996. Sociolinguistics, 2nd Edition. Cambridge University Press, Cambridge.

Jahangiri, N., Hudson, R.A., 1982. Patterns of variation in Tehrani Persian. In: Romaine, S. (Ed.),

Sociolinguistic Variation in Speech Communities. Edward Arnold, London, pp. 49–63.

Johnston, P., 1983. Irregular style variation patterns in Edinburgh speech. Scottish Language 2, 1–19.

Labov, W., 1966. The Social Stratification of English in New York City. Center for Applied Linguistics,

Washington, DC.

Labov, W., 1978., Where does the sociolinguistic variable stop? A response to Beatriz Lavandera. Work-

ing Papers in Sociolinguistics 44. South-West Educational Development Laboratory, University of

Texas, Austin, Texas.

Labov, W., 1994. Principles of Linguistic Change. Oxford University Press, Oxford.

Lavandera, B., 1978. Where does the sociolinguistic variable stop? Language in Society 7, 171–182.

Li, W., 1998. The ‘‘why’’ and ‘‘how’’ questions in the analysis of conversational code-switching. In: Auer,

P. (Ed.), Code-switching in Conversation, Language, Interaction and Identity. Routledge, London, pp.

156–176.

Milroy, L., 1987. Observing and Analysing Natural Language. Blackwell, Oxford.

Moreau, M.-L., 1986. Les sequences preformees: entre les combinaisons libres et les idiomatismes. Le cas

de la negation avec et sans ne. Le Francais Moderne 54, 137–160.

Newbrook, M., 1986. Sociolinguistic Reflexes of Dialect Interference in West Wirral. Peter Lang, Bern.

Pohl, J., 1975. L’omission de NE dans le francais parle contemporain. Le Francais Dans le Monde 111,

17–23.

Romaine, S., 1981. On the problem of syntactic variation: a reply to B. Lavandera and W. Labov.

Working Papers in Sociolinguistics 82, South-West Educational Development Laboratory, University

of Texas, Austin, Texas.

Sandy, S., 1997. L’emploi variable de la particule negative ne dans le parler des Franco-Ontariens

adolescents. MA dissertation, York University.

Sankoff, G., 1973. Above and beyond phonology in variable rules. In: Bailey, C.-J.N., Shuy, R. (Eds.),

New Ways of Analyzing Variation in English. Georgetown University Press, Washington, pp. 81–93.

Sankoff, G., Thibault, P., 1980. The alternation between the auxiliaries avoir and etre in Montreal French.

In: Sankoff, G. (Ed.), The Social Life of Language. University of Pennsylvania Press, Philadelphia, pp.

311–345.

Sankoff, G., Vincent, D., 1980. The productive use in ne in spoken Montreal French. In: Sankoff, G.

(Ed.), The Social Life of Language. University of Pennsylvania Press, Philadelphia, pp. 295–310.

Tennant, J., 1995. Variation morphologique dans les francais parle des adolescents de North Bay

(Ontario). PhD thesis, University of Toronto.

Trudgill, P., 1974. The Social Differentiation of English in Norwich. Cambridge University Press,

Cambridge.

Trudgill, P., 1986. Dialects in Contact. Blackwell, Oxford.

Trudgill, P., 1999. New-dialect formation and dedialectalization: embryonic and vestigial variants. Jour-

nal of English Linguistics 27 (4), 319–327.


variable deletion of french ne: a cross-stylistic perspective

Documents