Sociolinguistic Variation in Contemporary French [1 ed.] 9789027288998, 9789027218650

Citation preview

Sociolinguistic Variation in Contemporary French

IMPACT: Studies in Language and Society IMPACT publishes monographs, collective volumes, and text books on topics in sociolinguistics. The scope of the series is broad, with special emphasis on areas such as language planning and language policies; language conflict and language death; language standards and language change; dialectology; diglossia; discourse studies; language and social identity (gender, ethnicity, class, ideology); and history and methods of sociolinguistics.

General Editor Ana Deumert

University of Cape Town

Advisory Board Peter Auer

Marlis Hellinger

Jan Blommaert

Elizabeth Lanza

Annick De Houwer

William Labov

J. Joseph Errington

Peter L. Patrick

Anna Maria Escobar

Jeanine Treffers-Daller

Guus Extra

Victor Webb

University of Freiburg Ghent University University of Erfurt Yale University

University of Illinois at Urbana Tilburg University

University of Frankfurt am Main University of Oslo University of Pennsylvania University of Essex University of the West of England University of Pretoria

Volume 26 Sociolinguistic Variation in Contemporary French. Edited by Kate Beeching, Nigel Armstrong and FranÇoise Gadet

Sociolinguistic Variation in Contemporary French Edited by

Kate Beeching University of the West of England, Bristol

Nigel Armstrong University of Leeds

FranÇoise Gadet Université de Paris Ouest Nanterre la Défence

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data Sociolinguistic variation in contemporary French / edited by Kate Beeching, Nigel Armstrong and Françoise Gadet. p. cm. (IMPACT: Studies in Language and Society, issn 1385-7908 ; v. 26) Includes bibliographical references and index. 1. French language--Social aspects--France. 2. French language--Social aspects--Foreign countries. 3. French language--Variation. 4. Sociolinguistics. I. Beeching, Kate. II. Armstrong, Nigel. III. Gadet, Françoise. PC2074.75.S63   2009 447--dc22 isbn 978 90 272 1865 0 (hb; alk. paper) isbn 978 90 272 8899 8 (eb)

2009023931

© 2009 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

Table of contents Introduction Kate Beeching

vii

Section I.  Phonological variation and leveling Introduction Nigel Armstrong

3

Perception and production in French dialect leveling Nigel Armstrong and Zoë Boughton

9

The sociolinguistic relevance of regional categories: Some evidence from word-final consonant devoicing in French spoken in Belgium Philippe Hambye

25

Prosodic style-shifting as audience design: Real-time monitoring of pitch range and contour types in Swiss French Jessica Sertling Miller

43

The immigrant factor in phonological leveling Tim Pooley

63

A prototype-theoretic model of Southern French Elissa Pustka

77

The law of position revisited: The case of mid-vowels in Briançon French Anne Violin-Wigent

95

Section II.  Stylistic and syntactic variation Introduction Françoise Gadet

115



Sociolinguistic Variation in Contemporary French

Variation in first and second language French: The case of parce que Mireille Bilger and Henry Tyne French preadolescents’ perceptions of stylistic variation: A contrastive sociolinguistic study Laurence Buson

121

141

Sociolinguistic variation in African French: The Ivorian relative clause Anne Moseng Knutsen

159

Register variation in the non-standard use of non-finite forms Nathalie Rossi-Gensane

177

Section III.  Lexical variation and semantic change Introduction Kate Beeching and Richard Waltereit

195

Discourse markers and regional variation in French: A lexico-semantic approach Gaétane Dostie

201

Sociolinguistic factors and the pragmaticalization of bon in contemporary spoken French Kate Beeching

215

From ‘luck’ to ‘wealth’: The stylistic (re)distribution of fortuné in Modern French Bruno Courbon

231

Index

253

Introduction Kate Beeching

University of the West of England, Bristol

It may seem surprising, in the 21st. century, to edit a book whose title contains the words ‘Sociolinguistic Variation’. After all, a succession of recent works in the constructionist model (Le Page & Tabouret-Keller, 1985; Butler, 1990; Rampton, 1995, 2006, Schiffrin, 1996; Mills, 2003 and Coupland, 2007) have roundly criticized quantitative sociolinguistics and the classic Labovian variationist paradigm, suggesting that these approaches project a static conceptualization of the relationship between language and social identity. Viewed through the constructionist lens, variationism and social stratification studies set the identification of social groups with linguistic features in stone. Social identities, it is argued, are, on the contrary, fluid and dynamic, locally situated and constructed. Not only that but speakers consciously ‘style’ their words and use genres strategically in order to create specific meanings and effects. In defense of quantitative approaches Some defense of the notion of ‘variation’, of social stratification studies and even of quantitative methods appears to be required before introducing the chapters in the volume. Firstly, then, in its defense, it must be pointed out that variationist sociolinguistics, since its earliest days in the 1960s, has been concerned with variation and change – there is an explicit recognition that the identification of social groups with language features shifts and that this factor is, indeed, instrumental in language change. The association of variationism with a static conceptualization of the relationship between language forms and social groups appears then to be a fundamental misapprehension of the sociolinguistic enterprise as it was originally conceived. Secondly, when speakers use styles strategically to create specific meanings, such ‘styles’ evoke either stereotyped or commonly recognized fashions of speaking or salient features of these. If, as Schiffrin (1996: 199) remarks, “identity is neither categorical nor is it fixed: we may act more or less middle-class, more or

 Kate Beeching

less female, and so on, depending on what we are doing and with whom”, individuals must have some image of what it is to “act middle-class, female and so on”, in other words, they must be tapping into generally recognized features which sum up these categories. On the one hand, speakers may draw on a stereotypical or unreal image of what it is to speak like someone from the Midi, from the cités or like a radio or TV presenter. On the other, the features calqued or parodied in styling may be reasonably close to a real representation of the community of speakers thus styled. It seems, then, a reasonable project to find out what linguistic features are indeed used by particular communities of French speakers in different contexts at particular points in time – an area which has not necessarily been comprehensively explored. The problem of how ‘style’ might be defined is taken up by Coupland (2007, Chapter 1) and a distinction has to be made between the classic Labovian model (e.g. Labov, 2006: 167) and the interactional view which examines the way in which the linguistic variables associated with a particular context are put to work creatively. In classic variationist sociolinguistics, style is strictly defined by the assumed decreasing levels of formality reflected in individuals’ pronunciation of key variables when reading a word-list, reading a text, in a formal interview situation and in informal conversation. The proportion of occurrences of informal forms is, in turn, mirrored in terms of social background – in early studies, both Labov and Trudgill found a correlation between the forms used by the least educated and those used in less formal situations (such as everyday conversation). In other words, in the early variationist studies, the most educated speakers used fewer ‘informal’ variants overall but they did use informal variants in a graduated manner across the various styles. Interactionists argue that speakers can adjust their language creatively to suggest particular contexts or levels of (in)formality and can do so for expressive effect. These might be termed ‘styles of speaking’. If it is the case, however, that speakers adopt ‘styles of speaking’, how can the quantitative researcher collect data which is socially or situationally constituted (rather than exhibiting conscious features of style)? Rickford (2001: 16) “cautions against rejecting the more predictable, often automatic, aspects of stylistic variation that are the focus of quantitative studies of variation”. It seems likely that ‘styling’ is restricted to particular types of public performance, a point raised by Rickford (2001: 230) when he says: some verbal (and non-verbal) performances, especially those that involve radio broadcasts, large audiences and public occasions are more stylized than others… people in such situations are trying more consciously than most of us may do in everyday life, to project personas of various types…. There are undoubtedly parallels to this kind of stylization in one-to-one conversation, but the opportunities and possibilities for it seem to increase as audience size grows.



Introduction 

The data studied in the current volume were, in the vast majority of cases, collected as part of one-to-one conversations and, it might thus be argued, constitute a fruitful hunting-ground for socially constituted as opposed to ‘styled’ modes of speech. Coupland (2007: 5) concedes, moreover, that, though they have not encouraged us to understand what people meaningfully achieve through linguistic variation, “the survey designs of variationist research … have been remarkably successful in revealing broad patterns of linguistic diversity and change”. Tagliamonte (2001: 730) concurs with this when she points out that “when the goal of research is to gauge and model the individual and combinatory effects of multidimensional internal linguistic factors alongside broadly defined external factors, a quantitative approach is particularly useful”. Without over-simplifying what are extremely complex relations and inter-relations between form, context and the individual, the chapters in this volume hope to make a real contribution to discerning the broad patterns of linguistic patterning in contemporary French, looking more closely at what linguistic variation might mean and at how it is acquired in learners of French as a second language. Overview of the volume The volume is divided into three main sections addressing issues in variation and change in contemporary French at the phonological, syntactic and lexico-semantic/pragmatic levels of linguistic analysis, respectively. Written entirely in English it will appeal to (socio-)linguists with an interest in contemporary French as well as to post-graduate students of French and specialists in the field. In introductions to the three main sections, the co-editors of the volume contribute overviews of their respective areas of expertise, namely phonological variation and leveling in hexagonal1 French and elsewhere, syntactic and stylistic variation and lexical variation and semantic change. These set the scene for the more focused studies presented in the individual chapters. A unifying theme of the volume is the role of external sociolinguistic factors in both variation and change. From a diatopic perspective, varieties of French in France, Belgium, Switzerland, Africa and Canada are considered, mainly with respect to phonological variation (Section I) but also with respect to syntactic and lexical features (Knutsen in Section II on French in the Ivory Coast; Dostie in Section III on Canadian French). 1. The term “L’Hexagone” is often used to denote France, because of its geographical shape; hexagonal French refers then to the variety spoken in France (rather than that of Belgium, Switzerland and so on).



Kate Beeching

The intersection between social factors and phonological change is explored in Armstrong and Boughton’s chapter on dialect leveling in hexagonal French. Hambye’s paper on word final consonant devoicing in Belgium is original in highlighting diastratic as well as regional factors, while variation with age is the focus of both Violin-Wigent’s and Pustka’s papers on the phonological features of Southern France. Miller’s study of style-shifting in Swiss French is unique in the volume in focusing on prosody and finally, a contrast between English and French is drawn in Pooley’s review of the impact of an influx of immigrants on local varieties. Many of these studies paint a similar overall picture: that European French is undergoing standardization or is tending towards a supra-local norm, a norm which is not necessarily that of classical standard French but has currency over the region as a whole and is not restricted to one locality. Sociostylistic and syntactic features are the focus of a number of the papers in Section II. Bilger and Tyne’s study of syntax across speech styles in UK anglophone learners focuses on the use of parce que while Rossi-Gensane compares the use of non-standard non-finite forms in the spoken language, in newspapers and in novels. Knutsen examines relative clauses in the French spoken in the Ivory Coast, while Buson traces the perceptions of pre-teenagers with respect to stylistic variation and relates these to social class. Finally, in Section III, we look at the impact of social and stylistic factors in lexical variation and semantic change. Dostie looks at diatopic variation, suggesting that the degree of pragmaticalization of discourse markers in Quebec French may vary from one region to another. Beeching, on the other hand, examines the extent to which sociosituational variation has contributed to the rise in distributional frequency and pragmaticalization of discourse-marking bon in modern hexagonal French. The social and economic factors behind the shift in meaning of fortuné from ‘luck’ to ‘wealth’, along with the resistance to change evident in metalinguistic works across the centuries, form the subject of Courbon’s highly detailed study of this lexical item which concludes the volume. In a volume of this scope one cannot hope to give comprehensive coverage of the array of phonological, syntactic and lexical features which underpin sociolinguistic variation in contemporary French. We hope, however, to have given some flavor of the richness afforded by the language in all its variety and to have pinpointed areas of particular contemporary significance. My heartfelt thanks go to the many people who have made this volume possible, the authors themselves, the co-editors, who wrote the Introductions to each sub-section and who provided much useful comment on the shape of the volume, and on the contributions themselves, to Richard Waltereit for accepting the invitation to co-author the introduction to the section on Lexical Variation and Semantic Change, to the anonymous reviewers of each chapter and of the volume as a whole.



Introduction 

I acknowledge my indebtedness, too, to the Research Committee of the School of Humanities, Languages and Social Sciences of the University of the West of England, Bristol, for granting me the Research Leave which permitted me to honor my editorial duties and to complete my own contribution to this collection. References Butler, J. 1990. Gender Trouble. London: Routledge. Coupland, N. 2007. Style. Language and Variation. Cambridge: CUP. Labov, W. 2006. Social Stratification in New York City, 2nd edn. Cambridge: CUP. Le Page, R.B. & Tabouret-Keller, A. 1985. Acts of Identity: Creole-based Approaches to Language and Ethnicity. Cambridge: CUP. Mills, S. 2003. Gender and Politeness. Cambridge: CUP. Rampton, B. 1995. Crossing. London: Longman. Rampton, B. 2006. Language in Late Modernity: Interaction in an Urban School. Cambridge: CUP. Rickford, J. 2001. Style and stylizing from the perspective of a non-autonomous sociolinguistics. In Style and Sociolinguistic Variation, P. Eckert & J.R. Rickford (eds), 220-234. Cambridge: CUP. Schiffrin, D. 1996. Narrative as self-portrait: sociolinguistic constructions of identity. Language in Society 25: 167-203. Tagliamonte, S. 2001. Comparative sociolinguistics. In The Handbook of Language Variation and Change, J.K. Chambers, P. Trudgill & N. Schilling-Estes (eds), 729-763. Oxford: Blackwell.

section i

Phonological variation and leveling

Introduction Nigel Armstrong University of Leeds

The chapters in the phonology section of this volume range over a number of the varieties of French, in���������������������������������������������������������� cluding ‘French’ as commonly understood, that is the dominant variety of France that is frequently referred to as ‘supralocal French’, but also varieties spoken in Belgium and Switzerland, as well as Lille and midi French. Several areas of enquiry within the sub-discipline are represented too: variationist or quantitative sociolinguistics, as well as perceptual dialectology, historical perspectives on the language, and prosody. A concept that links every chapter in this section, and one that is currently of interest in sociolinguistics, is that of ‘phonological leveling’. Leveling is a term used to describe the spread in the UK and elsewhere of non-standard and non-localized features. Speakers retain some local features, generally distributed over a wider area, but these are overlaid by non-prestige supralocal forms. This development appears to echo other societal shifts like the ubiquitous wearing of informal items of clothing such as jeans. The question arises whether a similar social and linguistic evolution is occurring in France – and other French-speaking regions. If it is accepted that social change drives linguistic change (an axiom of sociolinguistics) then it must be that large-scale processes of language change like leveling are similarly motivated, as Meillet (1921: 17–18) suggested in general terms some time ago: The only variable element to which one can turn to explain linguistic change is social change […] and it is changes in the structure of society which alone can modify the conditions of existence of a language. We need to determine which social structure corresponds to a given linguistic structure and how, in general, changes in social structure translate into changes in linguistic structure.

Determining these things is a formidable task indeed. Sociolinguists commonly attempt to correlate elements of social structure like class and gender with linguistic features undergoing variation and possible change, with a view to identifying regularities or even universals of variation and change. A further way of theorizing



Nigel Armstrong

social structure, from the viewpoint of the individual, is to look at variable language in relation to ‘social identity’. It is after all individuals who change languages, as they interact with other individuals, as representatives of social groups. Coveney (2001: 1) has pointed out the close link between pronunciation and social identity: “Pronunciation is particularly closely linked to identity, more so on the whole than lexis and grammar. […]”, on account of its concrete character, which is allied to other physical aspects of behavior like gesture and dress. Further, (ibid.): “The connection [between pronunciation and identity] is also due to the fact that speakers use relatively minor differences in pronunciation to signal aspects of their identity, while generally maintaining intelligibility with speakers of the same language”. All spoken language is of course concrete in having physical realization, and the peculiarity of pronunciation is perhaps its capacity to convey regional features. From this point of view Coveney’s remark is more closely applicable to languages in which regional origin has the potential to form an important element of a speaker’s social identity, if only because regional accent is, or can be, closely linked to social status. The concept of social identity, central to the study of the relations between society and language as well as many other social practices, is useful in attempting to explore the relationships between opposites like group and individual, status and solidarity, in-group and out-group. Individual social identity, a psychosocial continuity which is at the same time susceptible to development, is difficult to theorize or even describe with much rigor, perhaps most obviously because it is experienced subjectively, is multi-faceted and dynamic, is perhaps rarely the object of conscious reflection, and is composed of elements which are in any event recalcitrant to precise measurement or even definition: class, gender, age, region and ethnicity are the most frequently evoked. Even if some of these categories are more discrete, or less abstract, than others, they are all ‘moving targets’ in the sense of being capable of ongoing development or construction, both in the short and long term. Region or locality is perhaps one of the least abstract of these, although even the ongoing construction of one’s regional identity is influenced by a process that is also complex and dynamic, because subject to continuing socio-geographical developments as well as personal trajectory. As Trudgill points out (1990:  1), a positive sense of regional identity is nevertheless likely to be imprinted in an individual at an early and hence impressionable age, and so continue to make its force felt. This is simply because many if not most people have early positive associations with a distinctive locality or region. Thus, to take the example of the broad north–south linguistic split in the UK, northern speakers may feel quite strongly that their social identity is bound up in an intimate way with their pronunciation of the (a) vowel in the lexical set that comprises words like glass and bath. To pronounce bath as [b[]θ], with a south-



Introduction: Phonological variation and leveling

ern back /a/, can be construed as showing disloyalty to that element of one’s social identity which might be labelled ‘Northerner’. Conversely, so to reconstruct that component of one’s accent as to abandon northern front /a/ may be to acknowledge diminished allegiance to one’s Northerner status. These remarks are of course applicable mutatis mutandis to France, as well as to other countries having a marked two-way national divide. It is unclear why (a) has this totemic and nearinvariable character for speakers of northern English, although Trudgill (1987: 11– 21) suggests that speakers may have a particular resistance to the importation into their speech of variables that are salient or already have phonemic status in their native variety. Salience is of course a problematic term in its definition – social as well as linguistic salience can be at issue, nor are these two aspects clearly separable – and in its introduction as an explanatory factor the concept can itself be in need of explanation (cf. Kerswill and Williams 2002). In this connexion it has been shown that some speakers of the southern varieties of French are converging in certain aspects of their pronunciation to the dominant northern standard variety. Stereotypical features of the southern French varieties, as is well known, include oral vowel plus consonantal appendage where the standard variety has nasal vowel, more frequent realization of schwa in all contexts, and a distribution of the mid-vowels that is close to being complementary in open and closed syllables – the so-called loi de position. Some evidence points to convergence to supralocal French in the treatment of schwa and the nasal vowels (Durand et al. 1987, Armstrong and Unsworth 1999, Pickles 2001). Further evidence is presented in this section. In relation to the link between pronunciation and social identity illustrated above by the quotation from Coveney, one implication for language variation and change is that the high processing cost of sometimes minute phonological adjustments is offset by the social advantage gained by alignment to the desired ‘reference group’. Many if not most speakers of a language are capable of imitating other varieties of the language. Imitation or mimicry, which by definition take place over the short term, are however very different from long-tern accommodation to a reference group, with all the psychosocial investment that implies. The reference group was defined by Merton (1968: 287) as: “any of the groups of which one is a member, and these are comparatively few, as well as groups of which one is not a member, and these are, of course, legion, [which] can become reference points for shaping one’s attitudes, evaluations and behaviour”. The position then is essentially that speakers may adopt new linguistic forms, the property of a given reference group, because they seek thereby to gain social advantage. As Labov points out (2001: 191), this view is similar to that articulated by Le Page and TabouretKeller (1985), who expressed their perspective on the adoption of linguistic forms with this purpose in the well-known phrase ‘acts of identity’.





Nigel Armstrong

The situation of supralocal French seems unusual because the adjustments speakers require to accommodate to any reference group are largely unavailable in pronunciation and are made in grammar and lexis, the levels of linguistic analysis having items that are discrete and open-ended compared to pronunciation and, following Coveney’s suggestion, perhaps less closely linked to identity. The major exception in the French situation overall is the convergence which seems to be taking place in southern French towards northern supralocal French, one example of which is discussed in this section. There is of course no language which could be used as a benchmark in the study of leveling, and Coveney’s suggested link between pronunciation and identity, cited above and seemingly applied to French, applies perhaps more closely to countries where regional and social features are closely associated, as we suggested above. Coveney remarks further (2001: 1) that: “producing […] sounds involves a series of physical acts which to an extent affect also our very appearance (notably the shape of the lips)”. These remarks can perhaps be applied to French national identity, as French has a proportionally large number of highly rounded vowels (six of the twelve oral vowels). In contrast, English speech sounds, because they often betray regional and thus social identity, can more plausibly be thought of as bracketed off from variable grammar and lexis, in the way that Coveney suggests. In the absence of a high degree of regional marking in French, there seems therefore to be no reason why variation on the three linguistic levels should not function in essentially similar ways, even if for English speakers, this situation seems idiosyncratic because they perceive grammar as more ‘cognitive’ than pronunciation – perhaps because of the strong standardizing pressures that seek to exclude, on grounds of specious rationality, features like multiple negation from educated discourse. The chapters collected in this section are eclectic, but all can be read as having similarities in bearing upon the issues touched on above. Armstrong and Boughton consider how the dominant variety of French is undergoing phonological leveling, a topic that is currently receiving a good deal of attention in languages other than French. They suggest that the supralocal French situation is distinctive on account of its relative lack of regional differentiation in conjunction with a variable pronunciation system that continues to reflect the social hierarchy, a suggestion that seems to throw new light on the general leveling debate. The chapters on what one might call peripheral French have distinctly different perspectives, while remaining linked to this section’s broad subject. Hambye’s chapter on word-final devoicing in Belgian French combines a rigorous linguistic analysis with a discussion that suggests continuing differentiation in Walloon Belgium, both within the region and as regards the standard French language, always a looming presence south of the border. Again in the context of the leveling debate, Pooley considers how Lille French speakers of immigrant origin behave in their treatment of pronunciation



Introduction: Phonological variation and leveling

features that, taken as elements in a structured system, serve as highly distinctive regional markers in this variety that diverges from supralocal French in several important respects. Miller shows that for one variety of Swiss French at least, prosody is an important element in stylistic variation, a finding that calls for extrapolation to other varieties of French in view of what has been suggested above concerning the relative lack of regional variation in the French segmental system. Violin-Wigent reports evidence of convergence to supralocal French in the treatment of the loi de position among speakers in Briançon, while Pustka proposes a prototypical model of the southern French varieties that attempts to establish a hierarchy of their features from central to peripheral. It is evident that the work of these authors brings fresh evidence and theory to the problems of social identity in language sketched above, as well as to the leveling debate. The situation bears watching, and more information is required. It is to be hoped that the chapters in this section will give rise to further reflection and research on the subject. References Armstrong, N. & Unsworth, S. 1999. Sociolinguistic variation in southern French schwa. Linguistics 37(1): 127–156. Coveney, A. 2001. The Sounds of Contemporary French: Articulation and Diversity. Exeter: Elm Bank. Durand, J., Slater, C. & Wise, H. 1987. Observations on schwa in southern French. Linguistics 25(5): 983–1004. Kerswill, P. & Williams, A. 2002. “Salience” as an explanatory factor in language change: Evidence from dialect levelling in urban England. In Language Change. The Interplay of Internal, External and Extra-linguistic Factors, M. C. Jones & E. Esch (eds), 81–110. Berlin: Mouton de Gruyter. Labov, W. 2001. Principles of Linguistic Change. Social Factors. Oxford: Blackwell. Le Page, R.B & A. Tabouret-Keller. 1985. Acts of Identity: Creole-based Approaches to Language and Ethnicity. Cambridge: CUP. Meillet, A. 1921. Linguistique historique et linguistique générale. Paris: Champion. Merton, R.K. 1968. Social Theory and Social Structure. Glencoe IL: Free Press. Pickles, M. 2001. La mère de mon père est née à Grenade: Some phonological features of the French of teenagers in Perpignan. In French Accents: Phonological, Sociolinguistic and Learning Perspectives, M.-A. Hintze, A Judge & T. Pooley (eds), 128–47. London: AFLS/ CILT. Trudgill, P. 1987. Dialects in Contact. Oxford: Blackwell. Trudgill, P. 1990. The Dialects of England. Oxford: Blackwell.



Perception and production in French dialect leveling Nigel Armstrong and Zoë Boughton University of Leeds, University of Exeter

In the first part of this chapter we provide a rigorous definition of what might be meant by the proposition that the supralocal or dominant variety of the French of France is ‘leveled’. We consider firstly some evidence that supports the notion of leveling as the diminution of regional features, looking secondly at the definition of dialect leveling that has to do with ‘social leveling’. Comparing UK English and French, we then discuss some perceptual evidence suggesting that supralocal French is now regionally but not socially leveled. In a subsequent section we consider behavioral data, characterizing the current French leveled phonological system in terms of what structural adjustments speakers have made and need to make in order to converge to this system.

Introduction: definition of leveling Variation in language is an intuitive concept, since social difference finds expression in linguistic difference, even if the mechanisms and motivations underlying variation, latent in the patterns that quantification shows, remain recalcitrant to ready definition. While variation is at least superficially obvious, what appears to be its contrary, leveling, is harder to define in a way that applies across languages, or even within one language. In a more or less everyday phrase like ‘social leveling’, or more explicitly ‘leveling down’, the implication or outright sense is of convergence towards a majority social practice, and moreover one that enjoyed little prestige, or perhaps ‘covert prestige’ in the sociolinguistic jargon, before the mutation occurred. Non-linguistic signs of leveling in this sense are plentiful, and are perhaps easier to conceptualize than linguistic change. Clothing is an obvious example; as Adonis and Pollard (1997: 242–43) point out, items of work and sports wear like t-shirts, jeans and trainers are now increasingly acceptable in more formal contexts. The authors speculate, following Perkin (2002:  431–32), whether such trends reverse:



Nigel Armstrong and Zoë Boughton

the sociologists’ ‘principle of stratified diffusion’ (the theory that trends in dress, music, entertainment, and lifestyle always begin at the top and work their way down through society). If it had ever been wholly true — and the past history of the upward trajectory of trousers, the lounge suit, and casual wear generally […] suggest otherwise — the 1960s turned it upside down, with the young of the upper and middle classes emulating the denizens of Liverpool and the East End of London.

Language and clothing are of course both ‘social practices’. The ‘language of clothing’ is a complex subject, but a glance at photographs of earlier periods suggests at once greater uniformity and hierarchy (upper- and middle-class clothing imitated by the working classes) in contrast to the present fragmented, non-hierarchical situation, and this change seems to have a wider application to social practices generally, including variable language. Against this, ‘standardization’ refers to the (development and) maintenance of an overtly prestigious set of norms. Leveling is sometimes opposed to standardization, yet it seems plausible that both processes will have the effect (or at least the intention, in the latter case) of reducing linguistic variation. The term leveling implies a degree of homogenization, or a reduction in the number of linguistic features that differentiate the speaker groups that sociology usually distinguishes (class, age, gender, ethnicity, etc.), as well as across regions. The French case is curious from this point of view because ‘standard’ French appears already to be fairly highly leveled; there is some evidence for a quite striking degree of uniformity in pronunciation in the dominant northern urban “oïl” region. We discuss some of this evidence below. Dialect leveling taken in a rather different sense is currently a prominent focus of interest in sociolinguistics, understandably so in view of the processes of social and geographical mobility currently weakening a social fabric that has hitherto been fairly cohesive. Foulkes and Docherty (1999: 13) have the following definition that is widely cited: Leveling differs from standardization (or dedialectalization) in that speakers do not automatically abandon their local forms in preference for the standard. Rather, there appears to be a tension between speakers’ desire to continue signaling loyalty to their local community by using local speech norms, and a concurrent urge to appear outward-looking or more cosmopolitan.

This definition, while it might correspond to the UK situation and others, contrasts sharply with what was suggested above, since the recession of localized forms in favour of more widely distributed (but not necessarily overtly prestigious) features implies not linguistic homogenization, but rather greater diversity, at least in an intermediate perspective.



Perception and production in French dialect leveling

A further contrast is that the creation of wider regional forms at the expense of the standard suggests leveling in a different sense: a process of social rather than, or as well as, linguistic leveling, which would see the standard language transformed from its status as hegemonic variety to one among others. This seems intuitive if we consider the rapid and thorough-going changes that have occurred in all Western societies recently, taking perhaps the 1960s as a definitive turning point, as suggested in the discussion of ‘stratified diffusion’ cited above. In this view, the admittedly rather vague notion of ‘social mobility’ can be evoked, for example in the UK, to point out that several hitherto elite professions are now open to non-RP speakers. Other notions, again intuitive but difficult to define with any rigour, let alone correlate with language change, such as ‘social leveling’ referred to above or a general ‘informalization’ that erodes social hierarchies, can also be cited as factors that may have linguistic effects. In summary, dialect leveling can be understood in two senses that are relevant here. The first sense is linguistic, and can be interpreted broadly as a reduction in the number of variable linguistic features distinguishing the various social categories that make up a speech community. The second is social, and becomes clearer if we distinguish between the ‘horizontal’ or regional dimension of leveling, and the ‘vertical’ or social-class dimension. In the latter perspective, the development of supra-regional forms suggests leveling as a process of symbolic social leveling rather than (or as well as) linguistic leveling, whereby the standard would be transformed from its status as superordinate variety to one among others. This therefore is leveling in a sense opposed to standardization. The term ‘process’ must be emphasized; it is hardly justifiable to suggest that the standard is no longer superordinate. The second part of the definition cited above summarizes one aspect of this process quite neatly: even if ‘cosmopolitan’ seems to imply preference for a standard language variety, the more neutral term ‘outward-looking’ certainly seems apposite since the abandonment of local forms is not, at least in the UK, necessarily reinforcing the standard. If the foregoing is applied to French, two questions may be posed: (i) to what extent is the dominant or supralocal French variety leveled in the sense of betraying relatively few regional features? And inseparable from this, (ii) to what extent is French leveled in the sense of a ‘social narrowing’ working through the adoption by middleclass speakers of working-class forms, as appears to be happening elsewhere? Perceptions of leveling The results of a perceptual study by Armstrong and Boughton (1998) suggested that a sample of urban oïl French speaker-listeners from Rennes were able with





Nigel Armstrong and Zoë Boughton

quite a high degree of success to identify the social class of a further sample of speakers from Nancy on the basis of their speech, but were much less successful in identifying accurately any regional element. Forty Rennes informants were asked to listen to eight randomly-ordered one-minute extracts of spontaneous speech from interviews recorded in Nancy, each taken from a different cell of the overall speaker sample which was stratified by age, class and sex, thus one younger working-class male (YWCM), one older middle-class female (OMCF), etc. After each extract, the Rennes respondents answered the following deliberately broadly framed questions: A. In your opinion, does the person speaking have an accent? B. Is the person more likely to be working class or middle class? C. Can you identify the person’s region of origin? D. Do you find the person’s accent pleasant? The results for social class against regional perceptions (questions B and C) are shown below in Table 1. From the sociolinguistic literature one would expect WC males to rank highest on this test, and MC females lowest. Table 1 shows this broadly to be the case, with the notable exception of the older middle-class female, who was of working-class origin and had retained some localized linguistic features. This OMCF speaker’s regional origin was also correctly identified by eight of the 40 informants, although it should be noted that the answers counted as accurate were subject to a liberal interpretation: ‘correct’ answers were those that mentioned eastern France, even if other (incorrect) guesses were made at the same Table 1.  Rank order of Nancy speakers by Rennes informants’ perception as working class against regional identifications by the same listener-judges (Armstrong and Boughton 1998) Nancy speaker

YWCM OWCM OWCF OMCF YWCF YMCM YMCF OMCM

Perception as WC by Rennes informants N / 40 37 35 28 23 14 11 2 1

‘Correct’ identification by Rennes informants of Nancy speakers’ regional origin N / 40 1 0 5 8 1 0 0 0



Perception and production in French dialect leveling

time. If only those responses were shown which mentioned Nancy or Lorraine, the numbers of correct responses would be negligible. Significantly, the OMCF speaker’s regional origin was identified by several Rennes informants through a feature of intonation rather than of accent understood in the segmental sense. Table 1 shows very clearly that even for those Nancy speakers who were almost universally perceived to be working class, rather few Rennes respondents were able to say where they might be from with any precision. The YWCM speaker in the Nancy sample illustrates clearly the disjunction between the Rennes informants’ perception of social class as against regional origin. 92.5% (i.e. 37 out of the 40) of the Rennes informants perceived him to be working class while he rated only 2.5% of rather vaguely ‘correct’ responses (i.e. one informant judgement out of 40) in relation to his regional origin; 22.5% (9 out of 40) of the Rennes informants hazarded the guess that, like them, he came from Rennes or Brittany and the majority (57.5% or 23 out of 40) were not able even to hazard a guess. As Rennes and Nancy are about 450 miles apart, this result is striking from the viewpoint of a UK speaker-listener (the distance is comparable to that between Exeter and Edinburgh). A tentative conclusion emerging from this is that one variety of French is perceived by a sample of 40 French informants as ‘leveled’, if by this is meant that regional features appear to have been largely lost from the variety (question (i) above). In general, it was not the case that the respondents perceived a regional accent but were unable to place it; rather, they did not, on the whole, perceive any divergence in pronunciation that could be characterized as regional. The second conclusion, relating to social-class identification, may result from a limitation of the study, as indeed could the first. The limitations that are relevant here are: (i) given that we are concentrating on phonological leveling, differences of subject matter, lexis, grammar, etc., in the non-scripted Nancy extracts may well have influenced the judgements of the Rennes listeners and distracted them from pronunciation features alone; and (ii) questions B and C appear unevenly framed: B in particular, ‘is the person more likely to be working-class or middle-class?’, offers a 50% chance of correct social-class identification, while C, ‘can you identify the person’s region of origin?’, is more open and gives the respondent no guidance as to the correct answer. Bearing these limitations in mind, Table 1 does suggest that working-class features, most likely non-localized ones, are present in the speech of the Nancy sample and were largely correctly identified by the Rennes panel. The pattern in the left-hand column of Table 1 suggests a sharply gradated perception of socialclass accents. Given that the speech on the stimulus tape was not uniform with regard to subject matter, it seems likely that a combination of features in pronunciation, lexis and grammar as well as topic produced an effect in each case. The results in the right-hand column of Table 1 show that perception of social-class





Nigel Armstrong and Zoë Boughton

accent did not depend on the recognition of local features. Table 2 below suggests that at least one largely non-localized linguistic feature may have provided a cue to aid social-class identification by the Rennes panel. Table 2 shows a fairly close correlation between the Rennes panel’s perceptions of the Nancy speakers as working class, and the latter group’s rates of deletion of /l/ and /r/ word-finally following an obstruent (tab(le), quat(re), etc.). While none of the Rennes respondents commented on this feature explicitly during the perceptual test, several of them did make impressionistic observations regarding the working-class Nancy speakers (e.g., elle mange des lettres; il finit pas ses mots ‘she eats letters’; ‘he does not finish his words’) which suggest that the respondents were referring to relatively high rates of deletion of various segments in the stream of speech. The deletion rates are ranked very much in line with social-class expectations, and the quite large gap that separates the WC from the MC speakers is particularly noticeable, as is the OMCF speaker’s production. It will be recalled that her regional provenance was correctly identified by several Rennes informants; in contrast with this, from a social-class point of view, her behavior with regard to liquid deletion is situated broadly where one would expect to find it. On the basis of this admittedly slender and indirect data-base, social leveling as defined above does not therefore appear to be taking place in French pronunciation, neither in perception nor production. This is in contrast to the informalization noticeable elsewhere. The general lack of perception among the Rennes respondents of distinctive localized pronunciation traits is unsurprising if the most marked regional or local features are in fact disappearing, following the overarching processes of standardization and leveling referred to previously. An alternative Table 2.  Correlation between rank order of Nancy speakers by word-final post-obstruent liquid deletion rates and perception as working class (Boughton 2003) Nancy speaker

YWCM OWCF OWCM YWCF OMCF YMCM YMCF OMCM

% liquid deletion

Perception as WC by Rennes informants (N / 40)

83.7 69.1 67.5 54.8 44.4 39.6 28.6 25.0

37 28 35 14 23 11  2  1



Perception and production in French dialect leveling

explanation is that regional features in the speech of the Nancy informants, while present, were not recognized by the Rennes informants, for whatever reason (for instance, in the UK regional accents are widely diffused on television). However this may be, the lack of recognition appears at first puzzling when compared to results obtained by Kuiper (1999) in a ‘classical’ perceptual dialectology study which employed the more usual mapping and rating methods (as opposed to direct linguistic stimulus) to tap into the perceptions of regional French of a panel of Parisian respondents. Kuiper’s 76 informants were given a map of France and asked to “circle and identify in writing any regions ‘where people have a particular way of speaking’” (Kuiper 1999: 244). A composite map produced from these responses shows that Alsace–Lorraine was the second most frequently designated area (55/76 or 72% of responses); in this it was preceded only by Provence (63/76), and followed by the Nord / Lille (44/76). The informants were then given a list of 24 regional varieties of French, and asked to rate them according to degree of difference from the norm (that is, the respondents’ own variety), correctness and pleasantness. Lorraine French rated very highly (or badly) on all three rankings: 20th out of 24 for degree of difference (where 24th was maximally different, i.e. perceived to be furthest from the norm), 21st for correctness and 22nd for pleasantness. For these Parisians therefore, Lorraine French sits very near the bottom of the perceptual heap, as they believe it to be strongly divergent from the norm, incorrect and unpleasant. But what the Rennes test results show is that when a different method is adopted, namely when listenerjudges are presented with authentic samples of Lorraine French (albeit urban, Romance-substrate Lorraine French), they perceive very little regional divergence, and are largely unable to link it to a particular geographical area when they do. There seems then to be a mismatch between these two different kinds of perceptions, elicited using different experimental tools, and corresponding to the perceptual in the imagination, prompted by the ‘draw-a-map’ and ranking methods used by Kuiper, and that which responds to actual language data, as in the Nancy– Rennes study. We can suggest that the conflation by Kuiper’s respondents of Lorraine with Alsace shows a link between the two that has more to do with outdated historical-political factors than linguistic considerations, since while many Alsace speakers continue to have a marked accent influenced by the Germanic substrate dialect still spoken there, all evidence points to a more regionally neutral Lorraine accent on the other side of the Romance–Germanic boundary. As regards the perceptions of Kuiper’s informants of correctness and pleasantness, they are still vitiated by the Alsace–Lorraine conflation but do not go against our suggestion given above concerning the lack of social leveling in France, for they clearly imply a continuing hierarchy in speakers’ minds that ranks the standard accent most





Nigel Armstrong and Zoë Boughton

highly, to the extent that the middle-class Parisian variety is the standard, as was assumed by Kuiper and his informants. We may suggest therefore that French is more regionally than socially leveled. To the extent that cross-national comparisons are useful, this is in stark contrast to situations where highly marked regional pronunciation features seem to be involved indissociably in the reduction of social differentiation. The lack of copious results on French precludes any confident statement, but it appears that the sociolinguistic situation in non-southern France has changed a good deal since Walter (1982: 52) felt able to state that: Ce qu’il faut reconnaître, c’est que les différences sur le plan géographique l’emportent pour le moment, dans nos régions, sur les différences sociales. (‘We must recognize that at the moment, geographical differences are more important in our regions than social differences.’) We turn now to a consideration of what is involved in the process of leveling in French from the point of view of linguistic production or behavior. Dialect leveling in the production of French When considering the phonology of oïl or supralocal French, which excludes the southern varieties and those found on the peripheries of the dominant northern region, we concentrate on the vocalic system, where most variation is to be found. We alluded above to liquid deletion, but most evidence points to the stability and largely non-localized character of this feature (cf. the results of Boughton’s (2003: 135) quantitative variationist analysis however: a considerable difference of over 20% was observed between the average deletion rates of identically stratified speaker samples in Rennes (67.3% deletion) compared with Nancy (46.7%)). Within the two-way split between supralocal French and the other varieties, we examine leveling within the oïl region on the one hand and leveling towards it on the other, concentrating in the interests of concision on the southern varieties. Leveling within supralocal French The ‘minimal’ analysis of French as having a phonemic system of seven oral vowels (Landick 2004: 12; Gadet 1989: 89–94) is as follows: i y

u

e œ

o

a



Perception and production in French dialect leveling 

Within this minimal system there is of course a good deal of allophonic and in principle phonemic variation, although it is unclear how much of this has sociostylistic value. The canonical vocalic system is usually represented as comprising twelve oral vowels and four degrees of aperture: i y

u

e ø

o

ε œ

f

ә

a



No variation has been reported in supralocal French in the close vowels [i y u]. We now discuss the other elements in the system. The mid-vowels The mid-vowels shown in an archiphonemic representation in the minimal chart above are conventionally represented in the phonemic distribution shown in Figure 1.  

Open Syllable

Closed Syllable

e ε

aller; fée; j’irai; poignée allais; fait; j’irais; poignet

belle; père

Ø

feu; deux; malheureux

jeûne (noun); veule (adj.); –euse (suffix)

œ o f

jeune (adj.); veulent (verb form) peau; dos philosophie

saule; rose sol; dot; port

Figure 1.  Distribution of mid-vowels in open and closed syllables (cf. Valdman 1976: 57)

Figure 1 shows a patchy phonemic distribution: the vowels /e ε/ contrast in open syllables only and /ø œ/, as shown, contrast in closed syllables in just two minimal pairs. While /o f/ also contrast in closed syllables in a few minimal pairs, /f/ is found in open syllables in many polysyllabic words. As Gadet remarks (1989: 93), since almost all the members of the minimal pairs concerned belong to different word classes, disambiguation proceeds through syntax and the functional yield of the mid-vowels is therefore low. Any variation between the mid-vowel pairs in Figure 1 can be thought of as exemplifying mergers, in both the phonemic and phonetic senses. The front



Nigel Armstrong and Zoë Boughton

unrounded mid-vowels /e/ and /ε/ are both found in open syllables, as in the examples cited above, and it has been pointed out (cf. Armstrong 1932: 85) that an intermediate variant in an unstressed syllable is common in this context. Gadet’s (1989: 92–94) analysis, cited above, suggests a simplification of the system from four degrees of aperture to three, since the four-aperture system is vulnerable in view of the low functional yield of the mid-vowels. The front unrounded midvowels are most productive phonemically in their distribution in inflectional -er verb suffixes (infinitive, future, conditional, present, perfect and imperfect forms all have mid-vowel realizations), but here again Gadet points out that their role in conveying the relevant semantic feature is subsidiary to that played by syntax. The other two mid-vowel sets, /ø/ ~ /œ/ and /o/ ~ /f/, are much less productive phonemically; their distribution in open and closed syllables is near-complementary. Recent variationist studies (Landick 1995; 2004) concerning male Parisian MC and WC speakers (1995) and MC speakers (2004) show no sharp differentiating patterns or apparent change in progress in the mid-vowels. The difference between the front unrounded mid-vowels and the other two pairs is perhaps that the former pair is the object of overt pedagogical comment and drill in exercises like dictation. Mute-e or schwa Coveney (2001: 186) suggests that “the elision of schwa is one of the most important phonological variables in contemporary French”, because its elision reduces the number of syllables in colloquial speech and increases the number of closed syllables (going against the canonical pattern), thereby bringing consonants into contact. This latter effect causes in turn the assimilation processes that are common in colloquial supralocal French, resulting in sequences that are quite strikingly different from the standard. For example, following elision of mute-e, the sequence là-dedans is capable of being realized as [lann~] as a result of manner-of~] as the most formal realization. Evidence is articulation assimilation, with [ladәd lacking regarding the socio-stylistic value of processes of this kind, but they are certainly one element that contributes to the considerable gulf that divides colloquial supralocal French from more formal speech styles. As Coveney points out (ibid.) they also differentiate supralocal French from the southern varieties, by reason of the lower rates of mute-e elision in the latter. At the same time it is rather hard to characterize mute-e as a sociolinguistic variable in supralocal French. It is in fact hard to characterize it as schwa, since the



Perception and production in French dialect leveling 

vowel is realized as one of the front rounded mid-vowels by most speakers. Martinet and Walter (1973: 24) remark of the phonetic realization of mute-e that: For some speakers, this vowel still has a distinct quality, that of the central vowel transcribed as [ә]. For others, now more numerous, it is realized as [œ] or [ø].

If French mute-e is not schwa, then it is hard to justify its inclusion in the category of ease-of-articulation features, and indeed Gadet (1989) does not, devoting a separate chapter to it. It seems stable in most contexts, and is perhaps better thought of as an insertion feature, like variable liaison, that responds above all to scripted styles. Hansen’s (2000) results for variable schwa between two consonants bear this out, showing a steep hyperstyle ratio (a predominance of intraspeaker variation over interspeaker) between spontaneous and scripted speech styles. This predominance affecting mute-e in this context seems connected with the explicit teaching in schools of a recitation style that includes the vowel. Mute-e before a pause The sociolinguistic behavior of mute-e varies greatly according to its linguistic environment, to the extent that it is hard to theorize as a unitary variable. In almost all contexts it appears stable, but its distribution after a single consonant + pause seems to be changing in supralocal French. The relative recency of this shift means that very reliable data are lacking, but Table 3 shows a small increase between 1989 and 1993 in what Hansen (2003) calls “prepausal schwa”, in two corpora recorded in Paris. As Hansen points out, realization rates across the two corpora are by no means dramatically different, and individual realization rates vary considerably despite the small overall increase. It is perhaps noteworthy that insertion rates are higher in a less informal style in the later corpus. Certainly the short period between the two corpora rules out any confident claim regarding a change in progress. The author suggests, supporting her argument by analyzing several stretches of conversation in her corpora, that realization of the vowel has generally a pragmatic function, that of highlighting an important element in the discourse. The author traces the development of the phenomenon from being at the outset a naturally motivated post-consonantal ‘release phenomenon’ – a schwa is not uncommonly heard after a prepausal consonantal release, perhaps most noticeably in children’s speech – to acquiring a sociolinguistic distribution. Younger urban female speakers, as so often, appear to be leading this change, if that is indeed what it is. At present prepausal schwa is associated with a rise-fall intonation pattern. In this it differs from the prepausal schwa commonly heard in southern French.



Nigel Armstrong and Zoë Boughton

Table 3.  Rate of realization of prepausal schwa after a single consonant in Hansen’s 1989 and 1993 data (Hansen 2003), as in: c’est Pierre-euh, bonjour-euh 1989, Conversation + Interview

1993, Interview

Overall realization rates

% 18

% 25

Range of individual rates

6%–55%

N 2291

N 359

8%–37%

The current variation and possible change in prepausal schwa shows also that realization of the vowel in this position accords it a salience that it lacks in other contexts in supralocal French. Hansen suggests that the ‘warmth’ that prepausal schwa conveys might in part explain its growing use. All authors having reported its use agree that it conveys a speaker attitude, perhaps simply the desire to transmit emphasis, but it is tempting to speculate that mute-e in this position evokes the warmth associated with the generally positively regarded southern French pronunciation, in line with Kuiper’s (1999) perceptual study discussed previously. Variation in the A vowel Variation between the variants of A, to reprise the notation employed above, is also essentially concerned with a simplification of the system. The back variant appears to be non-standard or regional in words like pas pronounced as [p], [p] or [pf], with the ‘standard’ realization as [pa]. Many young speakers frequently have the back vowel in sequences like je sais pas, j’y vais pas, etc., realized as [∫εp], [ʒivep]. This has a curious socio-stylistic effect as it seems, at least to the present authors, to be used with special communicative intent, perhaps condescension or apathy on the speaker’s part. This is however only one aspect of the story; some older, more conservative speakers retain [] as a phoneme in minimal pairs like patte ~ pâte, and as an allophone in the suffix -ation, as do some rural speakers. ~], [] has a limited lexical distribution, and as Like the recessive nasal vowel [œ Coveney points out (2001: 188), “the very low frequency in the lexis of the back vowel // has led to its demise among most supralocal French speakers”. If the vowel is in recession, its social distribution remains complex, seemingly being found in northern speech and in upper- and lower-class supralocal French speech, but less frequently in the middle of the social distribution (Gadet 1989: 94). In other words, it is a minority form that is peripheral both geographically and socially.



Perception and production in French dialect leveling 

The fronting of /f/ A linguistic change which appears to be in progress in supralocal French is the fronting or centralization of the back mid-vowel [f], to [œ] as the frontmost realization, as in joli ‘pretty’ realized as [ʒf¨li] or [ʒœli] where the canonical realization is [ʒfli] (Landick 1995). Landick’s results show rather little use of the non-standard variant, and little social differentiation. Martinet (1969) has analyzed this change in functional terms, arguing that a system of back vowels with four degrees of aperture, like that of standard supralocal French — /u o f / — will be susceptible to fronting, owing to the relative compression of the back system compared to a front inventory of four degrees, where the vocalic space is greater. Evidence of /f/-fronting exists from the 17th century, in the form of disapproving comments by prescriptivists, and it may be that Martinet’s functional argument is valid in historical terms, while more recent change has come to be “ideologised”, in Milroy’s phrase (2003). Recent research in the Franco-Provençal area (Low 2006) suggests a generational shift that sees female speakers having more /f/-fronting, and this can be analyzed as a shift towards the supralocal French norm. This change is perhaps more plausibly interpreted as being socially motivated, because the speaker sample in question shows very little use of back //. Carton (2001: 9) suggests that while /f/-fronting was criticized as a working-class feature in the 19th century, it is currently ‘snobbish’ or ‘trendy’. In Table 4 below, the proportions of fronted, centralized and back /f/ are shown for individual speakers and groups in the sample analyzed by Low (2006) in Roanne, a small town in the Loire département some 50 miles north-west of Lyon. These figures are based on an instrumental analysis of each token of /f/. The gender difference is quite clear, especially for fully fronted tokens of /f/ — those with a value of 2 on a three-way fronting scale where 0 indicates back /f/, 1 a centralized variant and 2 fronted /f/. Observed frequencies have been suppressed so as not to overload the table with information (N = 372). The notation referring to individual speakers gives gender and year of birth. The information given in the rightmost column of Table 4, showing use of the fully fronted /f/ variant, in other words the variant that is the most ‘advanced’ both phonetically and socially, shows the clear patterns of age and (especially) gender of interest here. Although the male informants are not behaving in a homogeneous fashion, the females are, especially the younger group, and one can suggest very tentatively, in the absence of collateral studies, that any change in progress is being led by them. These results could have arisen by chance, but the regularity in the patterns of female behavior remains quite striking.



Nigel Armstrong and Zoë Boughton

Table 4.  Degrees of /f/-fronting in Roanne based on formant frequency analysis (Low 2006) 0

Percentage fronting 1

2

m. 1942 m. 1956 Older Males

30.00 43.37 38.35

34.00 27.71 30.08

36.00 22.89 27.82

f. 1954 f. 1955 Older Females

19.39 25.86 21.79

31.63 34.48 32.69

44.90 39.66 42.95

m. 1982 m. 1981 Younger Males

21.82 43.90 31.25

32.73 29.27 31.25

39.09 19.51 30.73

f. 1981 f. 1981 Younger Females

26.19 35.96 31.82

27.38 20.18 23.23

46.43 43.86 44.95

Informant

Summary of leveling within supralocal French In summary, it appears that functional arguments are applicable to most of the variable features discussed above, using ‘functional’ in a broad sense that includes the possibility both of deleting and merging segments where propositional meaning remains intact. The claim that the low functional yield of the mid-vowels strongly favors their simplification to a three-way opposition between e, œ and o seems plausible and can also be applied to the A vowel. Very little direct evidence indicating pronunciation changes is available for supralocal French, but if any are proceeding they seem to be unlike those in UK English discussed above, where it was suggested that certain marked, non-standard, widely distributed social–regional features are being assimilated by the standard language as well as by the (supra-) regional varieties. What is striking in the case of supralocal French is that most of these mutations are system-internal. We can contrast the UK situation with the French system-internal changes using the example of th-fronting in English. Obviously, the shift from /θ/ and /ð/ to /f/ and /v/ across two varieties is system-internal in introducing no new pronunciation feature into the innovating variety, and indeed true innovation in the sense of the introduction of a completely new feature must be uncommon: even labial /r/ has a long attestation in English. Equally obviously, in the th-fronting example the allophonic structure of the innovating variety mutates, and the innovating feature



Perception and production in French dialect leveling 

remains socially marked and stigmatized in its new distribution so long as change is incomplete. Any changes taking place in supralocal French are harder to think of in these terms; the notable exception is /f/-fronting, originally a working-class feature but now seemingly middle-class and a feature which is the object of convergence. The closest parallel to the UK situation is perhaps the variation reported by Jamin (2003) in banlieue French; but the author reports no indications of innovations into supralocal French from this variety. Leveling towards supralocal French If one aspect of what seems to distinguish France from elsewhere in its sociolinguistic situation is the fact that supralocal French is already quite highly leveled, then the correlate to this is that recent leveling tendencies observed in the nondominant French varieties (i.e southern and peripheral) take place towards the everyday French spoken norm; what we have called supralocal French. The most important changes in demographic terms concern the southern French varieties. Lack of space precludes any detailed discussion of these, which are in any event examined by other authors in this volume. Thus the very salient southern French feature, the so-called loi de position that sees open mid-vowels only in closed syllables and vice versa, and results in the distinctive pronunciation of words like rose and chose as [rfz] and [∫fz] where the dominant variety goes against the broad rule in having [roz] and [∫oz], is argued by Violin-Wigent (this volume) to be showing signs of convergence towards supralocal French among younger speaker groups. If this is the case, then it appears that southern speakers are adding elements of complexity to their phonological system. This is also true of the difference between the northern and southern systems that regulate the treatment of schwa. It is therefore perhaps unsuitable to view such examples as instances of ‘leveling’. Summary and concluding remarks We have suggested that phonological leveling in the supralocal French vowel system broadly involves variable simplification to a seven-vowel, three-aperture inventory. This needs a good deal of hedging, especially in view of the lack of variationist data on most variables. The exception is schwa, which obeys complex insertion rules and appears to respond above all to scripted speech styles. As we said immediately above, southern speakers adopting the northern system need to internalize more complexity than less, an observation that complicates the leveling debate still further.



Nigel Armstrong and Zoë Boughton

References Adonis, A. & Pollard, S. 1997. A Class Act: The Myth of Britain’s Classless Society. London: Penguin. Armstrong, L E. 1932. The Phonetics of French. London: Bell & Hyman. Armstrong, N. & Boughton, Z. 1998. Identification and evaluation responses to a French accent: Some results and issues of methodology. Revue PArole 5/6: 27–60. Boughton, Z. 2003. Phonological Variation in Contemporary Standard French: A Tale of Two Cities. PhD dissertation, University of Newcastle upon Tyne. Boughton, Z. 2007. Ce que prononcer veut dire: The social value of variable phonology in French. Nottingham French Studies 46(2): 7–22. Carton, F. 2001. Quelques évolutions récentes dans la prononciation du français. In French Accents: Phonological and Sociolinguistic Perspectives, M.-A. Hintze, T. Pooley & A. Judge (eds), 7–23. London: AFLS/CILT. Coveney, A. 2001. The Sounds of Contemporary French: Articulation and Diversity. Exeter: Elm Bank. Foulkes, P. & Docherty, G. J. 1999. Urban voices — an overview. In Urban Voices: Accent Studies in the British Isles, P. Foulkes & G.J. Docherty (eds), 1–24. London: Arnold. Gadet, F. 1989. Le français ordinaire. Paris: Armand Colin. Hansen, A.B. 2000. Le E caduc interconsonantique en tant que variable sociolinguistique. Une étude en région parisienne. LINX [Revue des Linguistes de l’Université Paris-X Nanterre] 42(1): 45–58. Hansen, A.B. 2003. Le [ә] prépausal et l’interaction. In Structures linguistiques et interactionnelles dans le français parlé, A.B. Hansen & M.-B. Mosegaard Hansen (eds), 89–109. Copenhagen: Museum Tusculanum Press. Jamin, M. 2003. Sociolinguistic variation in the Paris suburbs. Camling Proceedings 1: 100–106. Kuiper, L. 1999. Variation and the norm: Parisian perceptions of regional French. In Handbook of Perceptual Dialectology, Vol. 1, D. Preston (ed.), 243–262. Amsterdam: John Benjamins. Landick, M. 1995. The mid-vowels in figures: Hard facts. French Review 69(1): 88–103. Landick, M. 2004. Enquête sur la prononciation du français de référence. Paris: L’Harmattan. Low, J. 2006. /o/-fronting in Roanne French. MA dissertation, University of Leeds. Martinet, A. 1969. C’est jeuli, le Mareuc! In Le français sans fard, A. Martinet, 191–208. Paris: Presses Universitaires de France. Martinet, A. & Walter, H. 1973. Dictionnaire de la prononciation française dans son usage réel. Paris: France Expansion. Milroy, L. 2003. Social and linguistic dimensions of phonological change: Fitting the pieces of the puzzle together. In Social Dialectology: In Honour of Peter Trudgill [IMPACT: Studies in Language and Society 16], D. Britain & J. Cheshire (eds), 155–72. Amsterdam: John Benjamins. Perkin, H. 2002. The Rise of Professional Society: England since 1880. London: Routledge. Valdman, A. 1976. Introduction to French Phonology and Morphology. Rowley MA: Newbury House. Walter, H. 1982. Enquête phonologique et variétés régionales du français. Paris: Presses Universitaires de France.

The sociolinguistic relevance of regional categories Some evidence from word-final consonant devoicing in French spoken in Belgium Philippe Hambye

Centre de recherches VALIBEL, UCL, Belgium Several authors have claimed that regional variation in language is no longer very relevant in the French context. In this paper we aim to investigate whether this claim is also valid in Belgium. Although linguistic practices in Wallonia have shown a strong tendency towards homogenization (a term we define below), they are still perceived as significantly different from the standard norm prevalent in France and also as showing internal diversity. Through the analysis of a phonological variable (word-final consonant devoicing), we try to assess to what extent linguistic practices in Wallonia differ from standard French and whether they also manifest significant internal diversity. Our results show that these differences are substantial and reveal structurally distinct ways of speaking French in the areas under scrutiny.

Introduction Sociolinguists generally agree that there are two reasons why ‘diatopic’ (regional or geographical) variation in pronunciation has been a relatively neglected research area in France. One is related to the way French sociolinguistics has developed as a discipline (see Gadet 1996, 2003; Pooley 2000; Hornsby and Pooley 2001). These authors suggest that sociolinguistics in France developed in different ways from those characterizing British and US sociolinguistics, namely through a much greater focus on linguistic attitudes and rather less on strictly variationist work. The other reason relates to the fact that regional accent differentiation is no longer very salient in French (Gadet 1996; Armstrong 2002). There is, however, another view, that put forward by Walter (1982, esp. p. 52) some time ago, that geographical differences are more significant than social differences in French, and this view



Philippe Hambye

has also been commonly accepted and taken up by several other authors (see Battye and Hintze 1992: 306 ff.; Durand 1993: 258; Blanchet 2001: 59). Some work has shown that the standardization process has led to the emergence of new regional varieties of French, especially in urban regions at the periphery of European Francophonie (the Nord-Pas-de-Calais, see Pooley 1996 and Hornsby 2006; southern France, see Taylor 1996; Belgium, see Hambye 2005). The aim of this chapter is to contribute to the evaluation of linguistic standardization in French-speaking Belgium in general and in Wallonia in particular.1 Although linguistic practices in Wallonia have shown a strong tendency towards homogenization (using this term to cover both standardization and leveling) since the end of the First World War (Hambye 2008) – notably through the decline or at least weakening of traditional regional languages (mainly wallon and picard) – these practices are still perceived as significantly different from the standard norm prevalent in France, and also as showing internal diversity. Indeed, descriptions of the particular characteristics of a supposed ‘Belgian French’ have shed light both on some widespread linguistic features in Belgium which are unknown south of the Belgian-French border (the famous statalismes of Pohl 1983, 1985) and on more localized characteristics varying from one region to another (see Hambye 2005: 69–96 for a survey). Furthermore, speakers from Brussels and Wallonia often stress the diversity of the French spoken in Belgium (associating different ways of speaking with the major cities or sub-regions), when they are asked to comment on the distinguishing features of French spoken in Belgium (Lafontaine 1991: 33– 34; Francard 1993; Moreau et al. 1999: 27). As Francard points out, this internal differentiation makes a major distinction between the west of Wallonia – very close to France geographically and in terms of its social practices and economic activities – and the eastern part, generally assumed to be more ready to display its local identity, partly through language (Francard 2001b: 256). However, to the extent that these descriptions rely mainly on disparate observations and subjective beliefs and not on systematically gathered empirical data, it is still difficult to assess the distinctiveness and variability of linguistic practices within Wallonia. Differences with respect to the French of France are often said to be fewer than speakers imagine (Francard 2001a), but they can be sharp, salient and qualitatively or symbolically important. Through the analysis of a particular phonological phenomenon (word-final consonant devoicing)2 in three different towns of Wallonia, we will try to shed light on several questions which still need 1. Wallonia is the southern part of Belgium where most of the French-speaking Belgians live (more than 3 million), the remainder living mainly in the officially bilingual Brussels Region. 2. Word-final consonant devoicing is the process exemplified by bouge ‘move’ pronounced [bu˜∫]. As is shown by this phonetic transcription, in Belgium at least, accented vowels preceding



Word-final consonant devoicing in Belgium 

answers: To what extent are linguistic practices in Wallonia divergent from the standard Parisian norm, especially among young people? To what extent is there a significant difference between ways of speaking in different sub-regions of Frenchspeaking Belgium? How can we account for these differences, if indeed they exist, knowing that the ‘Belgian accent’ in French has always been stigmatized and that the French of France still seems to be held up as the standard variety for many speakers in Wallonia? And what role could these differences play in support of social categorization and differentiation? We first present the context of our study by reporting previous research on the standardization process in Wallonia and on French-speaking Belgians’ stance regarding linguistic norms. We focus secondly on the phenomenon under scrutiny, with special attention paid to former descriptions of consonant devoicing in Belgium. We follow this with the presentation of our methodology and results, which are discussed in the final section of the chapter. Regional French in Belgium and standardization The traits characteristic of French spoken in Belgium were for a long time the subject of criticism and strong stigmatization. In the traditional view shared by authors like Remacle (1969) or Warnant (1997), regional variation in French is the product of the ‘bastardizing’ influence of formerly spoken regional languages (the so-called patois or dialectes) among less educated speakers. Within this framework, regional variation is a contingent language contact phenomenon, which, it is assumed, will disappear progressively as the diffusion of ‘correct’ French and instruction in the standard language increase. Convergence towards the standard language, combined with the decline of regional languages, was viewed therefore as the only possible outcome of linguistic practices in modern states like France and Belgium. Purist discourse was so strong at that time that one could indeed imagine that these processes would lead to a complete and definitive rejection of any distinctive feature in the way Belgians spoke French: all the research on language attitudes in Francophone Belgium showed that speakers from Brussels and Wallonia expressed negative attitudes towards their own vernacular and tended to consider that the legitimate form of French was that spoken in France and in Paris in particular (Garsou 1991; Lafontaine 1991; Francard 1993). Recent studies have shown, however, that Francophone Belgians are ready to evaluate their ways of speaking more positively, as demonstrated for example by voiced plosive and fricative consonants are usually lengthened and are also pronounced long when the consonant is devoiced.



Philippe Hambye

Moreau through different matched-guise-type studies (Moreau et al 1999; Bouchard et al. 2004). Different methods of investigation (interviews, questionnaires, etc.) sometimes produced contrasting results where speakers may at the same time despise Parisian French as ‘snob’, express their preference for their own way of speaking as the one they ‘like’ the most, and perhaps continue to consider the French of France as the most ‘proper‘ way of speaking (Francard and Franke 2005; see also Klinkenberg 2000a: 707). As we have argued elsewhere (Hambye 2007, 2008), Belgian Francophones’ attitudes towards their language have always been and still are ambivalent (partly because evaluating a variety positively and attributing it legitimacy are two different attitudes), so much so that there still seems to be a gap between Francophone Belgians’ norm of reference3 (which is still located in France) and how they speak (retaining some regional features, even for example in the formal speech of a university professor). Stigmatized features supposedly inherited from contact with the traditional regional languages may now carry a particular social meaning and value, playing for example the role of positive markers of regional and social identity. In other words, even if standardization is still in progress, the existence of regionally-specific language features in the French spoken in Belgium is not, or not only, a historical residue and a transitional step towards the homogenization of linguistic practices, as it results from current social divisions which are reflected and reproduced through the differentiation of linguistic norms and habitus. Word-final consonant devoicing: sociolinguistic and phonetic perspectives Word final consonant devoicing (WFCD) has for some time been described as one of the main features characterizing French spoken in Belgium. Indeed, several authors such as Grégoire (1956: 78), Remacle (1969: 121–123), Pohl (1983: 30), Piron (1985:  374), or more recently Klinkenberg (2000b: 345) and Francard (2001a: 252–255) have mentioned this phenomenon in their descriptions of the differences between the pronunciation of French in Belgium and in France. Moreover, written reproductions of devoicing have been attested several times, showing that the phenomenon is sometimes consciously perceived by speakers.4

3. We use these terms to designate the set of forms considered as more correct or legitimate in themselves. 4. Internet instances of written forms reproducing WFCD are readily available, in words like belge (written belche, ‘Belgian’), Liège (Lièche, the city), les Rouges (les Rouches, ‘The Reds’) and les Mauves (« Mauffes », ‘The Purples’), two Belgian football teams.



Word-final consonant devoicing in Belgium 

In line with the normative perspective of many descriptive works on French spoken in Belgium, most of the writings on WFCD in Belgium consider this phenomenon as a pronunciation defect which is due to the influence of substrate regional languages and of neighbouring Germanic languages in the Brussels area (Flemish and standard Dutch), and found mainly among working-class speakers. Indeed, Wallonia’s main endogenous languages, wallon and picard (see Francard and Morin 1986; Pooley 1994), as well as Dutch (see Baetens-Beardsmore 1971), present word-final consonant devoicing as a phonological rule. Since they hold that WFCD was simply transposed from these contact languages into Belgian French, some authors view it as a categorical phenomenon in French too, which is related to a phonological rule of neutralization of the opposition between voiced and voiceless consonants (see Walter 1982: 113; Pohl 1983: 30; Reuse 1987: 110; Warnant 1997: 174). Recent quantitative sociolinguistic studies by Bauvois (2000, 2002) in the city of Mons (in the picard area) are important in helping us gain an idea of the frequency of WFCD in the French spoken in Belgium. An initial study by Moreau and Bauvois (1998) showed that among young informants engaged in post-secondary education, a WFCD frequency of 22.5%5 was observed during reading and sentence-filling tasks. This percentage rose to 43.3% in another study (Bauvois 2000) among informants from contrasting educational backgrounds (vocational colleges and university). Concerning older informants (between 32 and 43 years old), mainly of a middle-class background, Bauvois (2002) observed a rate of 38% during a conversation designed to be informal and 20% in a reading task. In all the cases, WFCD is more frequent among less educated speakers. Compared to results obtained by Pooley (1994) among speakers from the Nord region of France, these figures seem to indicate that WFCD is particularly frequent in Belgium. It is important for our purpose to distinguish between WFCD and other phonetic processes having similar effects, such as regressive assimilation which involves the devoicing/voicing of a word-final (or syllable-final) obstruent when it is followed by a voiceless/voiced obstruent.6 Regressive assimilation is a very general phonetic phenomenon in French, observable in the speech of all speakers: contrary to WFCD, many scholars consider assimilation as almost unavoidable, at least at a syllable boundary if not always at word boundary (cf. Snoeren and Segui 2003; Calliope 1989: 10). A major difference with WFCD is that the latter may appear before a pause, whereas assimilation occurs only before a consonant. In this 5. Percentages are not always given by the authors. We have calculated them on the basis of raw figures given in the papers. 6. For example, une robe sale ‘a dirty dress’ is pronounced [yntfpsal] when regressive assimilation occurs.



Philippe Hambye

sense, regressive assimilation is a phonetic phenomenon caused by articulatory constraints (Smith 1997: 474, 494–496; Snoeren and Segui 2003: 2328); WFCD is a phonological phenomenon reflecting the adoption by the speaker of a particular set of phonological rules and linguistic norms. The two phenomena may nevertheless overlap and be indistinguishable before a voiceless obstruent. Data collection and analysis The data analyzed below were collected in the framework of the author’s participation in the research project Phonologie du français contemporain (PFC).7 One of the core objectives of this project is to collect recorded data of spoken French everywhere in the Francophone world, using the same methods of speaker sampling, elicitation and transcription (see Durand and Lyche 2003). The recordings used in this study were made during two kinds of interaction with the informants: one was a formal interview (in the traditional Labovian style), while the other was close to an informal conversation involving the informant and one of his or her acquaintances, who also helped to introduce the interviewer for the first interaction. Surveys of this type were carried out in several places in Francophone Belgium. We use here the data gathered in three cities, Gembloux, Liège and Tournai. These three sites were chosen because they allowed us to capture the sociolinguistic diversity of Wallonia. The Liège conurbation is the biggest in Wallonia and is located in the east of the country. As Francard’s quotation above stated, Liégeois are known within Francophone Belgium for being proud of their city and for having a very recognisable accent. Gembloux and Tournai are smaller towns, situated in more semi-urban areas, respectively in the centre and the west of Belgium. Tournai is very close to the French border and its inhabitants have regular contacts with French people. These particular links with France can be presumed to have consequences for language attitudes and practices among speakers from Tournai. Against this, we might expect speakers in Tournai and Gembloux to be less ready to follow linguistic changes as they live in less urban areas, and these are known to be more conservative. In each city, 12 people were recorded: 6 men and 6 women belonging to 3 age groups (AGE1: over 55, AGE2: 35 – 50 and AGE3: under 30) and 3 social groups distinguished by their educational background (EB1: those who did not follow post-secondary education, EB2: those who studied or were studying in a postsecondary institution which is not a university, and EB3: those having followed or

7.

See http://www.projet-pfc.net/



Word-final consonant devoicing in Belgium

following a university degree course). These rather small informant numbers oblige us to be prudent in the interpretations that follow. For each informant, we took into account, in each type of recorded conversation, the first 15 occurrences of a word-final obstruent which was voiced in the standard and written form of the word. We assessed whether the consonant was voiced or not, based on our own perception and on the Praat pitch analysis tool. The Praat analysis was not always very reliable, however, owing to the recording conditions which did not allow a very precise measure of segment voicing. Furthermore, we know that voicing is not the only factor determining the voiced or voiceless nature of a consonant (Smith 1997: 473; Snoeren and Segui 2003). This is why our analysis relied mainly on the perception of the consonants once isolated within the sound signal. Cases where the analysis was impossible to carry out (because of the presence of noise, other speaker overlapping, or when the consonant was simply deleted) were excluded. The following characteristics were noted for each token of the variable: nature of the word-final consonant, right context (a voiceless/voiced obstruent, a sonorant, a vowel, a schwa or other hesitation vowel or a pause), the left context (vowel or consonant) and the presence/absence of a phrase-final prosodic boundary (where right context was not a pause). Results As was expected, different variation patterns were observed between the cities, and in particular a major difference between Tournai on the one hand, and Gembloux and Liège on the other. This difference is shown through the overall N of devoicing among speaker samples in each city (see Appendix, Table 1), which shows statistically significant differences between Tournai and Gembloux and Liège considered together (chi-square = p < 0.001). This opposition is clearer when one looks at the way each sub-group differentiates in each city. Figures 1 and 28 show that WFCD is a phenomenon characteristic of older and less educated speakers in Liège and Gembloux, whereas it appears not to be affected by external factors in Tournai (where none of the differences between sub-groups are statistically significant).9 8. See Appendix, Tables 2 and 3 for detailed results. 9. The score difference between AGE 1 group and AGE2/3 groups is significant both in Gembloux and Liège (p < 0.001). Concerning groups differentiated by educational background, the score difference is significant between EB1 and EB2 (p < 0.01) and between EB1 and EB3 (p < 0.001) in Gembloux; between EB1 and EB3 (p < 0.001) in Liège. Note that age and educational background overlap in our sample, as highly-educated people are over-represented in AGE1 group, while less educated informants are mostly in AGE3 group.





Philippe Hambye 70% 60% 50%

40%

Gembloux Liège

30%

Tournai

20%

10% 0% AGE1

AGE2

AGE3

Figure 1.  Devoicing against age in Gembloux, Liège and Tournai Legend. AGE1: over 55; AGE2: 35 – 50; AGE3: under 30 70%

60%

50%

40%

Gembloux Liège Tournai

30%

20%

10%

0% EB1

EB2

EB3

Figure 2.  Devoicing against educational background in Gembloux, Liège and Tournai

Legend. EB1: secondary education; EB2: vocational post-secondary education; EB3: university education



Word-final consonant devoicing in Belgium  100% 90% 80% 70% 60%

Gembloux Liège

50%

Tournai

40% 30% 20% 10% 0%

Voiced obs

Vowel

Schwa / HV

Sonorant

Pause

Voiceless obs

Figure 3.  Devoicing and the following context of the consonant in Gembloux, Liège and Tournai

70% 60% 50% 40%

Gembloux Liège Tournai

30% 20% 10% 0% /b/

/ɡ /

/d/

/v/

/z/

/ʒ/

Figure 4.  Nature of consonant and devoicing in Gembloux, Liège and Tournai



Philippe Hambye

An analysis of the role of internal factors10 gives a better understanding of how speakers from Tournai are distinguished from those from the other cities. Figure 3 shows that in Tournai, WFCD appears almost exclusively before a voiceless obstruent (VO) or before a pause (as in les choses que j’ai ‘the things that I have’ and dis-le à Claude ‘say it to Claude’), i.e. in contexts where devoicing of the word-final segment is favoured by phonetic constraints (see above). At the same time, there is still a noticeable difference in Tournai between the percentages of devoicing before a VO and a pause (p < 0.001). Similarly, we see in Figure 4 that among speakers from Tournai, fricatives are more often devoiced than plosives (p < 0.01): Tournaisiens tend therefore to devoice consonants which are phonetically most subject to devoicing, as can be seen from our data and from previous studies on consonant devoicing (see Bauvois 2000:40, Smith 1997: 488–491)). We can see on the other hand that speakers from Gembloux have high percentages of devoicing even before a vowel or a hesitation schwa, such that there is no significant difference in the scores of WFCD before VO and before schwa or sonorant (while the same difference is significant in Tournai). Moreover, in Gembloux the difference between the results before VO and before pause is not statistically significant. Results for Liège reveal an intermediate position. As was seen in Tournai, WFCD is clearly more frequent in contexts favouring devoicing (before VO and before a pause; p < 0.001). However, in contexts which do not favour devoicing, WFCD is significantly more frequent among Liégeois than among Tournaisiens (differences are significant for all contexts except before a sonorant). Similarly, we can see in Figure 4 that the influence of the nature of the wordfinal consonant is not obvious in Gembloux where WFCD seems much more generalized than in Tournai. In Gembloux, there is no significant difference between the results for the various types of consonants involved; the same is true for Liège, with the exception of the difference between the results for /b/ and those for /v/, /z/ and /ʒ/. The only convergent result among the three groups concerns the percentage of devoicing before a voiced obstruent: it is very low in each city, even if it is only in Tournai that it equals zero, as might be expected in a context where regressive assimilation may block any devoicing. Finally, we can note that there was no gender difference in our sample and that a difference in score between the two speech styles is only noticeable in Gembloux (devoicing is 40.1% in the interviews and 52.1% in the conversations; the difference between these results is significant at p < 0.05). Looking at the distribution of devoiced tokens among the individual speakers for three contrasting phonological contexts helps us to focus our description and 10. See Appendix, Tables 4 and 5 for detailed results.



Word-final consonant devoicing in Belgium 

allows us to exclude the possibility that our results were due to some exceptional individual behavior. Indeed, they confirm the pattern discussed above and the differences between the three cities. We see for example that in Tournai, all the occurrences of devoicing before a vowel are due to a single speaker (and always in the same word, week-end). This speaker is also one of the three who are the only ones in Tournai who present devoicing before a pause (incidentally, this group includes the oldest speaker in the Tournai sample, an 82-year-old woman). When the wordfinal consonant precedes a VO, devoicing appears in general with similar rates among each speaker in the Tournai sample, showing that it constitutes a regular phenomenon in this case. In Liège, devoicing before VO and before pause is fairly evenly distributed among the speakers, though in the latter context, older speakers score higher than younger. Before a vowel, however, results contrast more sharply: indeed, the four older speakers are responsible for more than 80% of the devoiced tokens. In Gembloux, devoicing is frequent across all members of the speaker sample, even in contexts unfavourable to devoicing. The lower devoicing rate before VO in Gembloux compared to the other cities is also noticeable for every Gembloutois – and is still difficult to explain. It is in Gembloux however that we observe the most dramatic individual results. Two speakers have particularly low rates of devoicing: the first, a 37-year-old primary-school teacher simply never pronounced a devoiced consonant within the time observed; the second, a 28-yearold secondary-school graduate teacher engaged in politics, is, with the former, the only informant in the Gembloux sample who presented no cases of devoicing before pause or vowel (in his case devoicing appears only at a low rate before VO). Yet devoicing scores are the more divergent when the right context of the consonant is a vowel: in this context, young speakers (except one man of working-class origin) have devoicing rates under 15% while other speakers’ percentages rise above 30% (except of course for the speaker who has no case of devoicing). Discussion The results presented above allow us to state that for many speakers in Wallonia, daily linguistic practices still differ from those recognized as standard French (those of the French radio and television for example) and that they are also quite diverse within Walloon Belgium itself. Despite the pressure of standardization, people in French-speaking Belgium maintain their own patterns of speech, even with respect to pronunciation where any divergence from the standard language is often criticized and almost never evaluated in a positive way (by contrast with lexical items particular to the region which may be valued).



Philippe Hambye

Our results not only confirm the existence of particular norms of linguistic usage, but they also help us to understand how distinctive they are and to what extent they might tend to converge gradually towards the standard. Two elements are particularly noteworthy. Firstly, we can note that the distance from the standard observed within the linguistic practices in Gembloux and Liège cannot be assigned only to the influence of substrate languages and to the low level of education of the speakers. If it were true that regional variation in Belgian French is due to dialect influence among speakers not aware of the standard language, then we would see devoicing with similar frequencies in the three cities and only among older and less educated speakers. Our results indicate that regional differentiation in Belgian French can be the product of a particular normative dynamic, even in a Francophone area deeply affected by the ideology of the standard. Secondly, it is worth emphasizing that the differences recorded are not only quantitative. Indeed, there is more than a contrast simply in the frequency of WFCD between Tournai on one hand, and Gembloux and Liège on the other. We saw that we could clearly distinguish phonological contexts favourable and unfavourable to devoicing. To the extent that Tournaisiens only show devoicing in phonological contexts which favour it, we can consider WFCD among them as being a phonetically conditioned phenomenon. Variation in this case corresponds to what Armstrong (2002: 13–18; Armstrong and Boughton 1998: 52–55) calls natural phonological variables (‘variables naturelles’), as opposed to arbitrary variables (‘variables arbitraires’). The latter type constitutes non-predictable forms of variation whereas the former is a consequence of natural tendencies in the articulatory system (like schwa or liquid deletion, assimilations, etc.). According to Armstrong, only the first type of variable is to be found in northern France French, because of the high level of language standardization in this area, and this would explain why phonological variation in this variety is mostly sensitive to the context and to the educational background of the speakers (as these are the main factors conditioning the speaker’s degree of resistance to these ‘natural’ tendencies and the adoption of careful pronunciation). What we observe in the two other cities in our sample, and especially in Gembloux, is a quite different phenomenon, and another type of variation, which is not purely the result of phonetic constraints and which may even go arbitrarily against some natural tendencies of the articulatory system (e.g. when devoicing occurs before a vowel). This means that speakers whose practices match with the more frequent pattern in Gembloux are not necessarily trying to speak standard French. Their pronunciation is not necessarily the reflection of their incapacity to avoid devoicing of final consonants and to control an articulatory tendency which is supposedly mastered by the more highly-educated speakers. Their way of speaking is in fact governed by a different set of norms and rules which makes their



Word-final consonant devoicing in Belgium 

French qualitatively (structurally) and not only quantitatively different from the standard language. Differences of this kind may play an important role in the perception of distinctiveness in the way people use French in Belgium. As they create sharper distinctions than purely quantitative differences, these may function as the basis for the social construction of ‘accents’ and linguistic varieties which themselves contribute to the production and reproduction of boundaries between groups. In this sense, linguistic differentiation as it has been depicted here creates and reflects at the same time differences between geographical entities which have a social meaning since they refer to social groups and social categories. It may thus for example contribute to the production and reproduction of the very idea of a FrancophoneBelgian distinctiveness or of the view of regional identities specific to Liégeois or Tournaisiens. This is not insignificant in times of very lively debates about the (Francophone) Belgian identity. Conclusion Geographical variation in French is clearly not today what it was for a long time before the massive standardization process of the 20th century. But the fact that the situation has changed does not mean it has disappeared. We have seen in this chapter that there are still salient differences in language use between the French spoken in Belgium and standard French and also significant differences within practices within Belgium. Movements of convergence and divergence seem thus to coexist within the standardization process still in progress. The evidence presented here confirms the need to go beyond the vision of regional variation in French as a contact-induced phenomenon. Through a narrow and detailed description of our data, we have been able to describe quantitative and qualitative differences which can help us understand how structural differences could be the linguistic backdrop to processes of social categorization (of accents, varieties and groups). At the same time, we have underlined the important limitations of our quantitative analysis as regards the issue of the explanation of speakers’ resistance to forces of standardization. In order to obtain a better understanding of the way differences in language norms and practices are socially produced, one has nonetheless to come closer to the linguistic socialization of the speakers, to their experience of the diverse ways of speaking French, and to their discourses about language. An analysis of the kind we have undertaken here is only the first step in that direction.



Philippe Hambye

References Armstrong, N. & Boughton, Z. 1998. Identification and evaluation responses to a French accent: Some results and issues of methodology. Revue Parole 5/6: 27–60. Armstrong, N. 2002. Nivellement et standardisation en anglais et en français. Langage et Société 102: 5–32. Baetens-Beardsmore, H. 1971. Le français régional de Bruxelles. Bruxelles: Presses universitaires de Bruxelles. Battye, A. & Hintze, M.-A. 1992. The French Language Today. London: Routledge. Bauvois, C. 2000. Des Belges et de l’assourdissement des sonores. In Actes du XXIIe Congrès international de linguistique et de philologie romanes, A. Englebert, M. Pierrard, L. Rosier & D. Van Raemdonck (eds), Vol. 3, 35–43. Tübingen: Max Niemeyer. Bauvois, C. 2002. Ni d’Ève ni d’Adam. Étude sociolinguistique de douze variables du français. Paris: L’Harmattan. Blanchet, P. 2001. Enquêtes sur les évolutions générationnelles du français dans le pays vannetais (Bretagne). Le français moderne 69: 58–76. Bouchard, P., Harmegnies, B., Moreau, M.-L., Prikhodkine, A. & Singy, P. 2004. La norme dans la francophonie périphérique: externe ou interne? Une étude expérimentale en Belgique, au Québec et en Suisse. In La variation dans la langue standard, P. Bouchard (ed.), 51–72. Montréal: Office québécois de la langue française. CALLIOPE. 1989. La parole et son traitement automatique. Paris: Masson. Durand, J. 1993. Sociolinguistic variation and the linguist. In French Today. Language in its Social Context. C. Sanders (ed.), 257–307. Cambridge: CUP. Durand, J. & Lyche, C. 2003. Le projet Phonologie du français contemporain (PFC) et sa méthodologie. In Corpus et variation en phonologie du français. Méthodes et analyses, É. DelaisRoussarie & J. Durand (eds), 213–276. Toulouse: Presses Universitaires du Mirail. Francard, M. 1993. L’insécurité linguistique en Communauté française de Belgique. Bruxelles: Ministère de la Culture, Service de la langue française. Francard, M. 2001a. “L’accent belge”: Mythes et réalités. In French Accents. Phonological and Sociolinguistic Perspectives, M.-A. Hintze, T. Pooley & A. Judge (eds), 251–268. London: CiLT/AFLS. Francard, M. 2001b. Le français de référence: Formes, normes et identité. Cahiers de l’Institut Linguistique de Louvain 27: 223–240. Francard, M. & Morin, Y.-Ch. 1986. Sandhi in Walloon. In Sandhi Phenomena in the Languages of Europe, H. Andersen (ed.), 453–474. Berlin: Mouton de Gruyter. Francard, M. & Franke, G. 2005. France et Belgique francophone. Deux pays qu’une même langue sépare?. In L’Europe et les Francophonies. Langue, littérature, histoire, image, Y. Bridel, B. Chikhi, Fr.-X. Cuche & M. Quaghebeur (eds), 119–137. Bruxelles: Peter Lang. Gadet, F. 1996. Variabilité, variation, variété: Le français d’Europe. Journal of French Language Studies 6: 75–98. Gadet, F. 2003. Is there a French theory of variation? International Journal of the Sociology of Language 160: 17–40. Garsou, M. 1991. L’image de la langue française. Bruxelles: Ministère de la Culture, Service de la langue française. Grégoire, A. 1956. Les vices de la parole, 4th edn. Namur/Paris: Ad. Wesmael-Charlier/Honoré Champion.



Word-final consonant devoicing in Belgium  Hambye, P. 2005. La prononciation du français contemporain en Belgique. Variation, normes et identités. PhD dissertation, Université catholique de Louvain. Hambye, P. 2007. Variation régionale de la prononciation du français en Belgique. Contacts de langues ou variation inhérente? In Actes du XXIVe Congrès international de linguistique et de philologie romanes, D. Trotter (ed.), Vol. 4, 363–374. Tübingen: Max Niemeyer. Hambye, P. 2008. Convergences et divergences. Quelques observations sur la standardisation du français en Belgique. In Standardisation et déstandardisation. Le français et l’espagnol au XXe siècle, J. Erfurt & G. Budach (eds.), 35–62. Frankfurt: Peter Lang. Hornsby, D. 2006. Redefining Regional French. Koinéization and Dialect Levelling in Northern France. London: Modern Humanities Research Association and Maney Publishing. Hornsby, D. & Pooley, T. 2001. La sociolinguistique et les accents français d’Europe. In French Accents. Phonological and Sociolinguistic Perspectives, M.-A. Hintze, T. Pooley & A. Judge (eds), 305–343. London: CiLT/AFLS. Klinkenberg, J.-M. 2000a. Le français en Belgique. In Histoire de la langue française 1945–2000, G. Antoine & B. Cerquiglini (eds), 701–718. Paris: Éditions du CNRS. Klinkenberg, J.-M. 2000b. W comme Wallonie, Bruxelles, Flandre. In Tu parles!? Le français dans tous ses états, B. Cerquiglini, J.-C. Corbeil, J.-M. Klinkenberg & B. Peeters (eds), 341– 350. Paris: Flammarion. Lafontaine, D. 1991. Les mots et les Belges. Bruxelles: Ministère de la Culture, Service de la langue française. Moreau, M.-L. & Bauvois, C. 1998. L’accommodation comme révélateur de l’insécurité linguistique. Locutrices et locuteurs belges en interaction avec des Français et des Belges. In Les femmes et la langue. L’insécurité linguistique en question, P. Singy (ed.), 61–73. Lausanne: Delachaux & Niestlé. Moreau, M.-L., Brichard, H. & Dupal, C. 1999. Les Belges et la norme. Analyse d’un complexe linguistique. Bruxelles/Louvain-la-Neuve: Ministère de la Culture, Service de la langue française/Duculot. Piron, M. 1985. Le français en Belgique. In Histoire de la langue française 1880–1914, G. Antoine & R. Martin (eds), 369–379. Paris: Éditions du CNRS. Pohl, J. 1983. Quelques caractéristiques de la phonologie du français parlé en Belgique. Langue française 60: 30–41. Pohl, J. 1985. Le français de Belgique est-il belge. Présence francophone 27: 9–19. Pooley, T. 1994. Word-final consonant devoicing in a variety of working-class French – A case of language contact? Journal of French Language Studies 4: 215–233. Pooley, T. 1996. Chtimi: The Urban Vernaculars of Northern France. Clevedon: Multilingual Matters. Pooley, T. 2000. Sociolinguistics, regional varieties of French and regional languages in France. Journal of French Language Studies 10: 117–157. Remacle, L. 1969. Orthophonie française. Conseils aux Wallons. 2nd edn. Liège: Les Lettres belges. Reuse, W.J. de. 1987. La phonologie du français de la région de Charleroi (Belgique) et ses rapports avec le wallon. La linguistique 23: 99–115. Smith, C.L. 1997. The devoicing of /z/ in American English: effects of local and prosodic context. Journal of Phonetics 25: 471–500. Snoeren, N. & Segui, J. 2003. A voice for the voiceless: Voice assimilation in French. In Proceedings of the 15th International Congress of Phonetic Sciences, M.J. Solé, D. Recasens & J. Romero (eds), 2325–2328. Barcelona: Universitat Autònoma de Barcelona.



Philippe Hambye Taylor, J. 1996. Sound Evidence. Bern: Peter Lang. Walter, H. 1982. Enquête phonologique et variétés régionales du français. Paris: Nathan. Warnant, L. 1997. Phonétique et phonologie. In Le français en Belgique. Une langue, une communauté, D. Blampain, A. Goosse, J.-M. Klinkenberg & M. Wilmet (eds), 163–174. Louvain-la-Neuve: Duculot.

Appendix Table 1.  Global results for each city Devoiced consonants (N)

Voiced consonants (N)

Total (N)

Devoicing rate (%)

181 144 100 445

213 212 274 679

  394   356   374 1124

45.9 40.4 26.7 39.6

Gembloux Liège Tournai Total

Table 2.  Results for each city contrasted by age group Devoiced Voiced consonants (N) consonants (N)

Total (N)

Devoicing rate (%)

Gembloux

AGE1 AGE2 AGE3

82 45 54

  46   78   89

128 123 143

64.1 36.6 37.8

Liège

AGE1 AGE2 AGE3

69 47 48

  47   73   72

116 120 120

59.5 39.2 40.0

Tournai

AGE1 AGE2 AGE3

25 36 39

  69   89 116

  94 125 155

26.6 28.8 25.2



Word-final consonant devoicing in Belgium 

Table 3.  Results for each city contrasted by educational background Devoiced Voiced consonants (N) consonants (N) Gembloux

Liège

Tournai

EB1 EB2 EB3 EB1 EB2 EB3 EB1 EB2 EB3

87 74 20 70 26 68 23 41 36

  69   89   55   50   34 108   66 116   92

Total (N)

Devoicing rate (%)

156 163   75 120   60 176   89 157 128

55.8 45.4 26.7 58.3 43.3 38.6 25.8 26.1 28.1

Table 4.  Results for each city contrasted by nature of word-final consonant Devoiced consonants (N)

Voiced consonants (N)

Total (N)

Devoicing rate (%)

Gembloux

/b/ /d/ /:/ /v/ /z/ /ʒ/

 2 39 14 19 53 54

 4 38 23 36 55 57

   6   77   37   55 108 111

33.3 50.6 37.8 34.5 49.1 48.6

Liège

/b/ /d/ /:/ /v/ /z/ /ʒ/

 1 25  3 20 58 57

 8 37  9 12 63 63

   9   62   12   32 121 120

11.1 40.3 25.0 62.5 47.9 47.5

Tournai

/b/ /d/ /:/ /v/ /z/ /ʒ/

 0 14  0 18 44 24

 9 75 18 31 90 51

   9   89   18   49 134   75

  0.0 15.7   0.0 36.7 32.8 32.0



Philippe Hambye

Table 5.  Results for each city contrasted by right context of word-final consonant Devoiced Voiced Total Devoicing consonants (N) consonants (N) (N) rate (%) Voiced obstruent Vowel Schwa / hesitation vowel Gembloux Sonorant Pause Voiceless osbtruent

 4 44 14 15 43 61

55 68 17 19 14 40

  59 112   31   34   57 101

  6.8   3.3 45.2 44.1 75.4 60.4

Liège

Voiced obstruent Vowel Schwa / hesitation vowel Sonorant Pause Voiceless osbtruent

 2 16  7  9 59 71

24 83 26 22 32  5

  26   99   33   31   91   76

  7.7 16.2 21.2 29.0 64.8 93.4

Tournai

Voiced obstruent Vowel Schwa / hesitation vowel Sonorant Pause Voiceless osbtruent

 0  5  1  6 11 77

48 84 46 34 33 29

  48   89   47   40   44 106

  0.0   5.6   2.1 15.0 25.0 72.6

Prosodic style-shifting as audience design Real-time monitoring of pitch range and contour types in Swiss French Jessica Sertling Miller

University of Wisconsin – Eau Claire This chapter focuses on a female speaker of Swiss French from the Vaud canton, whose speech was studied in three contextual styles: formal, informal, and very informal. A positive correlation between the magnitude of pitch range and informality of style was found in this context, with great variation occurring when the subject switched interlocutors, indicating possible prosodic styleshifting. In this study we also briefly investigate phrase-final rise-fall intonation contours that seem to be characteristic of Vaudois spontaneous speech, and that were observed more frequently when the speaker was conversing with a close relative than when she was addressing the investigator.

Introduction Bell (2001) defines style as “what an individual speaker does with a language in relation to other people”, and suggests that “style is oriented to people rather than to mechanisms or functions” (141). This focus on the interlocutor is what he calls audience design, as style shifts according to a speaker’s audience. In his own words, “audience design is generally manifested in a speaker shifting her style to be more like that of the person she is talking to” (143). Changes in style can therefore be regarded as an aspect of audience design, since those modulations are ways of marking one’s identity either to associate with or take distance from one’s conversational partner. Ervin-Tripp (2001) concurs: “Shifts of style appear to be like code-switching in invoking contrastive implications of the linguistic features.” (47) She continues: One shift is the change in language induced by a change in participants, because of the effect of speech accommodation (and divergence) on speech rates and speech variables. Dramatic examples of these changes can be heard when a speaker



Jessica Sertling Miller

is overheard talking to a child or to a relative from another region on the telephone. (Ervin Tripp 2001: 47)

The situation that Ervin-Tripp describes above echoes what happened during a pilot study investigating Swiss French prosody in two cantons of francophone Switzerland. One speaker from the Vaud canton seemed to switch intonational patterns depending on whether she was talking to a close relative on the phone or to the investigator, whom she had just met. This chapter investigates this styleshifting phenomenon, which we shall refer to as prosodic style-shifting, since one feature of the shift in question involves changes in various aspects of prosody, such as pitch range and tonal contour shapes. A phonetic description of prosodic styleshifting contributes to our understanding of the mechanics underlying this switch, which is perhaps not language-specific. The case of Vaudois French lends itself well to this investigation because of the possible variation between standard and non-standard tonal contours in the speakers’ inventory, as discussed later. Three speech styles associated with one speaker and two different interlocutors are considered here in order to observe how style-shifting can be systematically correlated with pitch range variations: formal, informal, and very informal styles. The formal style corresponds to the reading of a text; the informal style corresponds to a conversation with the French investigator, i.e. the present author, whom the speaker had only met once before briefly; the very informal style corresponds to a conversation with a close relative on the telephone. This analysis firstly investigates tonal contours that are perceptually linked to a Vaudois accent, i.e. speech that is typically identified in the francophone world as belonging to a speaker from the Vaud canton in Switzerland. Pitch range is also measured in major as well as in minor prosodic phrases. Major prosodic phrases will be called Intonation Phrases (IP), and minor prosodic phrases will be called Accentual Phrases (AP), following the terminology adopted by Jun and Fougeron (2000) in their autosegmental-metrical description of standard French prosody. The values obtained are then compared across the three speech styles described above. Research questions Speakers often alter their discourse depending on the formality of the situation. Joos (1959) outlines five different styles on a scale of formality, from the least to the most formal style: (1) intimate style, e.g. a private conversation between speakers who know each other extremely well; (2) casual style, e.g. a conversation between friends; (3) consultative style, e.g. a conversation between strangers; (4) formal style, e.g. a speech in an official setting; (5) frozen style, e.g. a cold form of



Prosodic style-shifting as audience design 

communication meant to dissuade participation. A different lexicon and varied syntactic structures can be employed, and different tonal patterns can emerge depending on the contextual style. Choice of style is constantly at play when speaking, and is closely linked to social environment and pragmatic intent. To adapt to certain contexts, speakers may alter their speech. As Irvine (2001) writes, “[styles] are part of a system of distinction, in which a style contrasts with other possible styles, and the social meaning signified by the style contrasts with other social meanings” (22). It is therefore to be expected that a Vaudois speaker would not speak the same way to a French-born linguist that she barely knows as to a close relative of hers, since the social situations differ greatly. The speakers’ choice of appropriate speaking styles is in fact mainly motivated by their audience. Bell (1984) argues that when speech situations and topics of conversation are comparable, the selection of speaking style is primarily a speaker’s response to a listener, and he calls this ‘audience design’. Speakers shift their style depending on how much they want to identify themselves to a specific group when interacting with their addressee. This would mean that a Vaudois informant is likely to use more non-standard Swiss features in a conversation in which he or she wants to be associated with the Vaudois group. As a consequence, the type of speech obtained during an interview will be linked to the type of relationship between the speakers and their interlocutor, even though it also depends on the topics approached and the setting of the conversation. Bell suggests that “speakers associate classes of topics or settings with classes of persons” (1984: 181). What type of style can then be expected from an interview where the speaker is conversing with an unfamiliar investigator? What sort of shift occurs when the informant talks to a close relative instead of the interviewer? And how does style change for a reading task? It is widely accepted that scripted speech is easy to distinguish from spontaneous speech. Studies have shown that listeners are able to tell apart with accuracy scripted stimuli from spontaneous. This was true in a study by Levin et al. (1982) even when a low-pass filter was applied to stimuli, which made words difficult to recognize. The study made clear that differences between scripted and unscripted speech are numerous, and various cues help listeners distinguish between the two. Such cues may include prosody. As far as standard French is concerned, there seem to be variations in intonation depending on speech style. Recent studies pointed out by Simon (2003: 5) have shown that it is difficult, if not impossible, to identify regional prosodic characteristics from read speech in French. Simon suggests that ‘reading prosody’ seems standardized for speakers of all varieties of French and quotes research by Boula de Mareüil (2003), who in his work on French spoken by people of North African descent synthesized read speech to find out which perceptual cues,



Jessica Sertling Miller

segmental or suprasegmental, give the most information enabling identification of the regional origin of the speakers. His results underline that in read speech, the fundamental frequency (f0, an acoustic property of speech measured in Hertz, which is correlated to the perception of voice pitch) does not provide enough information for listeners to successfully recognize the ethnic origin of the speakers. Thus looking at non-standard varieties of French in scripted speech might not necessarily contribute to the best identification of regional features. However, this does not mean that these local varieties are necessarily exempt from traces of regional features in that style, and these are certainly aspects of linguistic variation that need to be studied and understood. Recent observations indicate that Vaudois speakers seem to switch from a standardized type of French in formal contexts, such as reading tasks, to a more regional variety in informal contexts, such as casual conversations (Miller 2007). In careful speech situations, the French-speaking Swiss informants tend to use a type of speech close to what they consider to be the norm. This is consistent with remarks on the regional linguistic insecurity experienced by Swiss French speakers (Singy 1995; 1996; 2002; Prikhodkine 2002). Singy’s surveys on attitudes of Swiss speakers towards their own speech report that the Swiss feel that their variety is to some extent ‘inferior’ to that spoken in Paris (Singy 1996: 210). The French spoken around Paris is believed to represent the central norm in France, and is the variety to which French-speaking people worldwide are exposed to in the media. Informants in each fieldwork phase of the research on Vaudois intonation (Miller 2007) gave a negative evaluation of the way that they themselves speak. Some expressed the view that even French children have a more elaborate vocabulary than adult Swiss nationals, and admit that their own accent is associated with a lack of education. They also confessed to making fun of Swiss speakers who have a stronger accent than their own. This linguistic insecurity is certain to affect the informants’ speech during a recording session, and to be a source of linguistic variation. Compounding this with the Observer’s Paradox (Labov 1972), one can be fairly confident that speakers will show considerable shifting in various discourse contexts. In what follows, we investigate research questions relating to the effect of speech style on f0 patterns and to what extent they vary. Pitch range is measured and tonal patterns examined to see whether the location of pitch movements changes when a new style is adopted. Preliminary fieldwork observations indicate that Vaudois speakers may use standard French tonal patterns in formal situations and non-standard patterns in very informal contexts, which would contribute to the perception of a Vaudois accent. It is also conceivable that these patterns would combine standard and non-standard tonal patterns in situations that are neither very formal nor informal.



Prosodic style-shifting as audience design 

Preliminary observations Several reports suggest the existence of a recognizable Swiss accent (Bayard and Jolivet 1984; Knecht and Rubattel 1984; Métral 1977; Odin 1886; Singy 1995; 1996; Vouga 1984). Nevertheless, little phonetic evidence has been brought to light to date, perhaps owing to the difficulty of obtaining natural speech from Swiss speakers, who may be presumed to switch to standard French in a context where their speech is being monitored. The hypotheses formulated by those authors, based on informal intuitions, concern the existence of a pitch accent on the penultimate syllable of content words, and a slower articulation rate than in standard French. In addition, Léon (1971), quoted in Carton (1986: 252) and Simon (2003: 6), suggests that intonational regional markers are located at a precise point of the phrase, which seems to be at the end of a prosodic group. This could be true for Swiss French in the Vaud canton. The recent phonetic analysis of phrase-final Vaudois French intonational contours in scripted and unscripted speech has shown a recurring non-standard low-high-low tonal pattern in the conversational speech of several Vaudois informants (Miller 2007: 239), which could be perceptually associated with the Vaudois accent. Examples of contours from the recordings of a speaker nicknamed Nina, observed in three contextual styles (formal, informal, very informal) are presented below. These instances all occur at the end of IPs and were obtained following the methods described below. Figure 1 is an illustration of the formal contextual style with the AP le plus fort (‘the strongest’), which was obtained during the reading of a text. This representation is characterized by a small initial rise followed by an IP-final low boundary tone. This pattern can be interpreted as a LHiLL% pattern (Low-High initial followed by a Low boundary tone) within the autosegmentalmetrical framework used to analyze the prosody of Parisian French (Jun and Fougeron 2000; 2002). This AP configuration matches the widely reported pattern reported for standard Parisian French (Jun and Fougeron 2000; 2002). In our example, the values of f0 range from 125 to 175 Hz, making the pitch range in this instance 50 Hz. Figure 2 illustrates the informal contextual style with two IP-final APs: [ils sont un peu] [gonflés] (‘they are a little arrogant’). This sample was obtained during a casual conversation between Nina and the investigator. They were discussing the stereotypical views of French citizens in Switzerland. The first AP in the figure shows a fall-rise-fall configuration. This last fall movement could be interpreted as a low pitch accent (marked L*), yielding a contour typically represented as LHiL*, which is a possible pattern in standard French, although not the most common one (Jun and Fougeron 2000; 2002). The second AP, i.e. the IP-final AP, displays a rise-fall-rise contour, stretching over the last two syllables of the AP. This is not a reported standard



Jessica Sertling Miller

contour at the end of IPs in declarative utterances in varieties of French described so far in the literature. Here, there appears to be a first rise on the penultimate syllable [g¢)], lengthened and prominent perhaps due to an effect of insistence. There also appears to be an IP-final rise on the last syllable. Despite this rise, the utterance seems to have the perceptual connotation of an affirmation. These differences from standard French could be due to the pragmatic intent of eliciting a reaction or response from the hearer. In the utterance shown below, the speaker seems to be conveying with emotional intent her opinion that Parisians are arrogant. In this example, f0 ranges from 125 to 250 Hz. The pitch range in this instance is 125 Hz, greater therefore than in the formal contextual style shown in Figure 1. Figure 3 is an illustration of the very informal style with the IP-final AP j’ai la lessive (‘I have some laundry’), which was gathered while Nina was on the phone with her daughter. It should be noted that the subject walked away from the microphone to take the phone call. Consequently the acoustic quality of this sample is inferior to the others. This contour displays a fall-rise-fall tonal configuration on the last syllable of the phrase, which is not reported to occur in standard French at the end of major prosodic phrases. In this instance, f0 ranges from 175 to 300 Hz. The pitch range thus is 125 Hz, which is greater than in the formal contextual style shown in Figure 1, and equal to the informal contextual style shown in Figure 2. Nevertheless, there is a difference since the values in the very informal style are in fact higher than in the informal style. Not only do tonal patterns perceived as typically Swiss emerge in the informal and very informal styles (Figure 2 and Figure 3), but f0 values also seem to increase with informality of style, and pitch range tends to go up as well, especially from formal to informal, and from formal to very informal styles. Based on these preliminary observations, it appears that a contextual style shift may result in systematic modifications of this speaker’s pitch range as well as her selection of more Vaudois-sounding tonal patterns. The hypotheses that we attempt to test in this study are as follows: Hypothesis 1: pitch scaling varies according to contextual style. Hypothesis 2: the range of f0 values increases with informality.

Prosodic style-shifting as audience design 

Hz



175 150 125 L

H L

lә

L%

ply

fàt

0

0.93

Time (s)

Figure 1.  IP-final AP le plus fort (‘the strongest’) from the read sentence chacun assurant qu’il était le plus fort (‘each claiming it was the strongest’)

Hz

250 187.5 125 L i 0

Hi s¢

L* œ˜

pø

L

Hi g¢

Time (s)

L

H*

fle 1.20188

Figure 2.  Non IP-final AP ils sont un peu (‘they are a little’) and IP-final gonflés (‘arrogant’) in an informal context, in a conversation with the investigator about differences between the Swiss and the Parisians

Jessica Sertling Miller

Hz



300

237.5 175 L õe 0

la

Hi le

L

H*

L%

siv Time (s)

0.906526

Figure 3.  IP-final AP j’ai la lessive (‘I have some laundry’) in a very informal context, in a phone conversation with a close relative

Methods The data presented in this chapter were gathered during a pilot study investigating intonational patterns in different dialects of francophone Switzerland. Only one Vaudois speaker, Nina, was recorded during that stage, which at that point concentrated on intonation patterns and not on style-shifting. Two unplanned phone calls from her daughter during an otherwise controlled interview eventually altered the focus of this study. Because of the coincidental nature of the samples gathered then, it was not possible to reproduce this setting with other speakers during the limited amount of time available for fieldwork. While generalizations cannot be drawn from a study based on the samples of a single speaker, the conclusions reached from the latter can point to potential future investigations. Speaker Nina was 44 years old at the time of the recording. She was recruited during a local festival in a suburban village of Lausanne, where she was briefly introduced to the investigator. She agreed voluntarily to participate in the investigation, and chose to be interviewed in her own apartment. French is Nina’s only native language, and both her parents were born in francophone Switzerland. She did not study to university level, but has a diploma from secondary technical school. Furthermore,



Prosodic style-shifting as audience design

Nina has lived in the Vaud canton all her life except for one year spent in the German-speaking canton of St. Gall. However, she does not consider herself fluent in German, and it seems reasonable to assume that this year-long stay did not affect the way she speaks French. Data The recording session started with Nina filling out the consent form and biographical information sheets, with a simultaneous informal conversation to break the ice. Her first task was to read sentences on flashcards, and the text La Bise et le Soleil (‘The North Wind and the Sun’), commonly used in phonetic research. This task was a reproduction of part of a previous study of AP tonal patterns at different speech rates by Fougeron and Jun (1998). Finally, to elicit more spontaneous informal speech, questions were asked to initiate conversational discourse. The following topics were approached: life in Switzerland, cultural differences between France and Switzerland, as well as across Swiss cantons, and especially between French and German-speaking cantons. The interviews described above were planned in advance. What makes the study of Nina’s speech samples particularly noteworthy is the part of the interview that happened unexpectedly. The recording session was interrupted by two phone calls from Nina’s daughter, which were recorded as well. This yielded a third speech style in addition to the formal and informal contextual styles. As Nina was speaking to her daughter, she adopted a much more casual style, which is labeled here ‘very informal’. As a consequence, samples of three speech styles were recorded. This setting is similar to what Labov (1972: 89–90) experienced during an interview. His informant received a phone call during their meeting and after she picked up the receiver, she switched to a much more intimate style to communicate with the person at the end of the line. Such unexpected events have, therefore, been shown to provide a valuable glimpse into speakers’ least monitored vernacular speech styles. Segmentation protocol Samples from each of Nina’s three speech styles were selected for this study. The read text, La Bise et le Soleil, was chosen in its entirety and contains 45 seconds of recorded formal speech, with a total of 23 IPs. The informal speech style sample is an excerpt of Nina’s conversation with the investigator. 23 adjacent IPs were selected in the middle of the conversation. The investigator did not interrupt Nina at any time in the excerpt. Finally, to evaluate the very informal speech style, the entire recording of one of Nina’s phone conversations with her daughter was selected. It contains 21 IPs that matched the selection criteria. Those exclude from analysis





Jessica Sertling Miller

certain IPs from the spontaneous speech recordings, described as follows. The phrases with disfluent speech or filler words in phrase-final APs were not considered. In addition, phrases containing whispered sections were removed, since they interfere with the analysis of f0, which is only visible with voiced segments. Finally, interrogative phrases were discarded to focus solely on declaratives. The next step was the identification of the prosodic structure of each phrase. Using the speech analyzer Praat, the major prosodic boundaries were marked and each file was thus divided into IPs. For each IP, Praat’s automatic function established the points of the lowest and highest f0 values. The high values were labeled as H, and the low values as L. Some points selected by Praat were perturbations due to background noise and were removed during a manual check. These marks allowed us to later measure the highest and lowest f0 in each IP. After that, the last AP of each IP was located, since this is often where nonstandard tonal configurations are found in Vaudois French (Miller 2007). The position of major and minor prosodic boundaries was determined thanks to the syntactic structure of the phrases, and with the help of other cues such as accentual patterns, syllable lengthening, and place of pauses. Within that last AP, every syllable was delimited based on the perceptual visual aids provided in Praat. Those are the waveform, spectrogram, and the representation of formant frequency. These marks enabled us to keep track of different tonal targets in the last two syllables of each IP. A tonal coding was established to group those IP-final target points into categories. Labeling in that manner provides a way to track what kind of target point is being measured. The coding is as follows: 1. Li1 is a low initial target point preceding a tautosyllabic high target point in the penultimate syllable of the IP. It corresponds to an elbow before a rise. 2. L1 is a single low target point in the penultimate syllable of the IP. 3. H1 is a single high target point in the penultimate syllable of the IP. 4. Li2 is a low initial target point preceding a tautosyllabic high target point in the ultimate syllable of the IP. It corresponds to an elbow before a rise. 5. H2 is a single high target point in the ultimate syllable of the IP. 6. L2 is a single low target point in the ultimate syllable of the IP. It is often the last point in the IP. In summary, two analyses were performed. The first simply measured the highest and lowest f0 values in each selected IP. Those target points were labeled H and L. The second analysis focused on the last two syllables of each IP, with up to six possible marked targets, although they do not all necessarily occur together. Those target points were labeled Li1, L1, and H1 for the penultimate syllable of the phrase, and Li2, H2, and L2 for the final syllable. This labeling is not phonological. It was merely used to encode the points in Praat. Figure 4 exemplifies the coding.



Prosodic style-shifting as audience design 

Hz

250 187.5 125

Li1 H i 0

s¢

H1

Li2

H2

L œ˜

pø

g¢ Time (s)

fle 1.20188

Figure 4.  Example of coding in Praat. Tier 1 labels the last two syllables of the IP, tier 2 the lowest and highest points of the IP, tier 3 the syllable boundaries with the phonetic transcription

Results A Praat script was written to retrieve the value in Hz of each tonal target marked in the text grid of the files. The measures were automatically put into a text document, and manually sorted to be exported to SPSS, a statistical software program where a descriptive analysis was executed. On some occasions Praat retrieved values that corresponded to perturbations. The numbers that appeared high or low compared to the expected f0 range of a female speaker (approximately between 120 and 300 Hz) were then checked and the target points measured again manually to preserve accuracy. IP analysis The results relative to the measures obtained for the lowest and highest points of each IP (L and H) are presented here. For each IP, there is an L and an H target point. The difference between H and L was calculated and provided a value in Hz equal to the pitch range of each IP. SPPS computed the mean range of all the values obtained for each of the three speech styles, as shown below, so that pitch range values can be compared across speaking styles. N corresponds to the number of IPs selected for analysis in each file.

Jessica Sertling Miller

Table 1.  f0 values in Hz of each IP, combined into mean range values according to contextual style Contextual style

Mean

N

Std. Deviation

Formal Informal

117 100

23 23

33.60 59.93

Very informal

162

21

95.21

These values do not directly support the hypotheses framed earlier, but do not contradict them either. It seems that there is not a large difference in pitch range between the formal and informal style. If anything, the range decreases in the informal situation. However, the pitch range values clearly increase when the speaker switches to a very informal style. Additionally, while the mean f0 values do not progressively increase from a formal to a very informal style, the standard deviation does, and so does the general range of values. This is expressed visually below.

400

40 * F0 range



200

11

0 Formal

Informal

Very informal

Contextual style

Figure 5.  f0 range in IPs according to contextual style. The low and high ends of the whiskers represent respectively the lowest and highest values retrieved for each style. The symbols 11 and 40 stand for outliers



Prosodic style-shifting as audience design 

To determine the significance of the data obtained, a single-factor analysis of variance (ANOVA) was carried out with speech style as the only factor. This means that the results found are indicative only of this one speaker’s patterns, and cannot be generalized to a larger population. What one can gain from this preliminary analysis of pitch range variation in style-shifting is insight into significant interactions of factors that might later be tested on a larger scale. This analysis reveals a significant main effect for speech style (F(2, 64) = 4.95, p < 0.05). A Tukey posthoc test showed that the effect was chiefly due to differences between the informal and very informal styles (p < 0.05). The difference between the formal and very informal style is however almost significant as well (p = 0.077). There was no significant effect found between the formal and informal style. There is another way to categorize these styles and group them to account for the observed variation. What differentiates the formal and informal styles from the very informal style is the person whom Nina is addressing. In the first one, she is reading a text for the investigator, and in the second one she is having a conversation with the investigator. But in the third one, she is talking with a close relative. It seems that the switch from addressing an outsider to conversing with someone that she knows extremely well triggered the most significant changes in her pitch range. Thus, all seems to indicate that, as is often specified in the sociolinguistic literature, a style-shift could be a function of audience design (Bell 1984), among other factors such as topic and setting. In this case, there is evidence for prosodic style-shifting in particular. IP-final AP analysis Let us now turn to the analysis that measured six possible target points in the last two syllables of the IP. The values below were obtained by taking automatic f0 measurements of the values for each labeled target point in the last AP of every IP. The results are illustrated with the box plots in Figures 6–11. In almost all instances, there is a gradual increase in f0 range from formal to informal contextual styles. The only exception is for the H2 f0 target points, which have a slightly lower range value (272 Hz) than in the informal style (287 Hz). However, there is a rise in those values, and the box plot associated with the H2 f0 target points in Figure 11 illustrates this, with a minimum and maximum higher in the very informal style than in the informal style. This holds true in a majority of cases. A modulation in pitch range is probably not the only correlate of f0 variation with speech style, and changes in minimal and maximal f0 values could also play a part in speech style use and distinction.

Jessica Sertling Miller

300

Li1(Hz)

250

200

150

Formal

Informal

Very informal

Contextual style

Figure 6.  f0 range in Li1 points according to contextual style

350

300

250 L1 (Hz)



200

150

100 Formal

Informal Contextual style

Figure 7.  f0 range in L1 points according to contextual style

Very informal



Prosodic style-shifting as audience design 

H1 (Hz)

400

200

Formal

Informal

Very informal

Contextual style

Figure 8.  f0 range in H1 points according to contextual style

400 350

Li2(Hz)

300 250

*

7

200 150 100 50 Formal

Informal Contextual style

Figure 9.  f0 range in Li2 points according to contextual style

Very informal

Jessica Sertling Miller

300

L2(Hz)

250 *

3

19

200

150

Formal

Informal

Very informal

Contextual style

Figure 10.  f0 range in L2 points according to contextual style

500 19 400 H2(Hz)



300

200

Formal

Informal Contextual style

Figure 11.  f0 range in H2 points according to contextual style

Very informal



Prosodic style-shifting as audience design 

ANOVAs and Tukey post-hoc tests were performed on these data to evaluate the significance of the differences seen so far. Concerning Li1 f0 target points, it was observed that across speech styles, the difference in values is significant (F(2, 10) = 5.11, p < 0.05). The Tukey post-hoc test reveals that significance is however only found in the comparison between formal and very informal styles. It was found that the difference in values across speech styles for L1 target points is also significant (F(2, 34) = 6.32, p < 0.05). The Tukey post-hoc test reveals that significance is found in the comparison of formal and very informal styles, as well as between informal and very informal styles, but not when comparing formal and informal styles. In addition, the difference in values across speech styles for the H1 and H2 target points is widely significant (F(2, 37) = 10.42, p = 0.00 for H1; F(2, 30) = 7.80, p = 0.02 for H2). The Tukey post-hoc test reveals that significance is found in the comparison of formal and very informal styles, as well as between informal and very informal styles, but not when comparing formal and informal styles, echoing the findings for L1 target points described earlier. It seems that all H f0 target points show differences in pitch range across formal and very informal styles, and across informal and very informal styles, whereas the variation was limited across formal and very informal styles for some of the L f0 target points (Li1). One can hypothesize that peaks carry valuable perceptual information in French. As a consequence, they may be the most affected by a change of style when speakers want to convey certain nuances in their speech. Finally, the differences in values across speech styles for Li2 and L2 target points are significant (F(2, 22) = 4.26, p < 0.05 for Li2; F(2, 27) = 4.73, p < 0.05 for L2), the Tukey post-hoc test revealing that significance is found only in the comparison of formal and very informal styles. Table 2 summarizes the findings. An X is marked where the main effect for significant differences was found in the comparison of contextual styles (F–I = formal vs. informal; F–VI = formal vs. very informal; I–VI = informal vs. very informal). Table 2.  Significant differences across contextual styles for all target points CONTEXTUAL STYLES F–VI

I–VI

Li1 L1

X X

X

H1

X

X

Li2

X

H2

X

L2

X

Target points

F–I

X

 Jessica Sertling Miller

The results reported in this section all show significant main effects for style. Nevertheless, the Tukey post-hoc tests reveal that not all individual comparisons between styles are significantly distinct. No significant effects were found when comparing formal and informal styles. It appears that pitch range expansion mainly varies significantly and consistently across formal and very informal styles in the last two syllables at the end of IPs, extending uniformly to all target points in the IP-final AP. On some occasions, a difference between informal and very informal styles was also significant. This was true for L1 f0 target points, Hi f0 target, and H2 f0 target points, i.e. for half the points in the group, and notably all H tonal targets. One can speculate that H tonal targets carry valuable perceptual information intrinsic to their accentual nature in French (initial phrase accent or word-final pitch accent), and as such they also hold the most information relative to speech style. This could be why they are the points that show the most variation across styles. Summary and conclusion This chapter presents a case study of three different speech styles. These were obtained during a recording session with one female Vaudois speaker in her 40s. Although generalizations cannot be made from this study, it provides some preliminary evidence of what might be called prosodic style-shifting between various contextual styles. The subject’s pitch range was measured in each contextual style (formal, informal, very informal) and a positive correlation between pitch range and the informality of the situation was observed. In general, the more informal the situation, the wider Nina’s pitch range was. The results tend to support the prediction that pitch scaling varies according to contextual style, and that the range of f0 values increases with informality. Both were shown to be true in the discourse of Vaudois speaker Nina, more so from formal to very informal styles, however, than from formal to informal styles. More specifically, what is noticeable in these results is that there was little change observed between the formal and informal styles at all levels of the analysis, whether entire IPs or only selected target points in the last two syllables of the phrase were examined. Greater variation was seen in all analyses performed in comparisons involving the very informal style, which is when Nina switched to address her daughter. Variation in the prosody of Nina’s speech seems to come more from a change of interlocutor (from conversing with the investigator to conversing with her daughter) rather than a change of format (from reading to conversing). This is strong evidence supporting the hypothesis of prosodic style-shifting as audience design.



Prosodic style-shifting as audience design 

We can assume that when speaking to a close relative, people engage in more emotional conversations or convey their feelings more openly, thus using a larger tonal range for a greater involvement in the meaning of their utterances. The nature of those modulations and the reasons behind them can depend on social contexts and pragmatic intents. But there certainly must be several acoustic correlates at play in the distinction between scripted and spontaneous speech, and between speech addressed to different types of audiences. Prosody is one of those correlates: pitch scaling is affected, as well as the choice of tonal configurations, with patterns associated with the local variety of Vaudois French preferred when speaking to a close relative in Nina’s case. For example, more non-standard IP-final rises were observed when she was talking to her daughter. These high movements could logically also account for the greater pitch range variation observed in the very informal context. Based on those preliminary results, we can hypothesize that similar scaling effects accompanied by some kind of modification in tonal configurations would occur for other Vaudois speakers in similar situations, but perhaps too for speakers of different dialects of French, for the same reasons. This remains to be tested. To date, not much is known on the prosody of non-standard varieties of French, as a result perhaps of the challenges created by what we think is prosodic style-shifting as audience design. An outsider is unlikely to be able to record natural speech in formal fieldwork situations because speakers will probably tend to modify their speech accordingly, as did Nina. Protocols to obtain such samples should therefore involve as little intrusion as possible on the investigators’ part. Once more natural speech data is gathered, it would be interesting to carry out perception investigations to find out whether regional variation is better identified with the help of prosodic cues, segmental cues, or a combination of both. The study by Boula de Mareüil (2003) reviewed earlier suggests that f0 would not provide sufficient information to recognize the geographical origin of a speaker, but results could differ depending on what styles of speech are used. It may be that the identity of the speaker’s interlocutor would have some influence on the way the samples are perceived. References Bayard, C. & Jolivet, R. 1984. Des Vaudois devant la norme. Le français moderne 52: 151–158. Bell, A. 1984. Language style as audience design. Language in Society 13(2): 145–204. Bell, A. 2001. Back in style: Reworking audience design. In Style and Sociolinguistic Variation, P. Eckert & J.R. Rickford (eds), 139–169. Cambridge: CUP.



Jessica Sertling Miller Boula de Mareüil, P. 2003. Rôle de la prosodie dans l’accent maghrébin en français. Paper presented at the 5ème Journées PFC (Phonologie du Français Contemporain), 4–5 July 2003, in Toulouse, France. Carton, F. 1986. A la recherche d’intonations régionales. Actes du 17ème Congrès International de Linguistique Romane, Aix-en-Provence 1983 6: 247–257. Aix-en-Provence: Presses de l’Université. Ervin-Tripp, S. 2001. Variety, style-shifting, and ideology. In Style and Sociolinguistic Variation, P. Eckert & J.R. Rickford (eds), 44–54. Cambridge: CUP. Field, A.P. 2000. Discovering Statistics Using SPSS for Windows: Advanced Techniques for the Beginner. London: Sage. Fougeron, C. & Jun, S. 1998. Rate effects of French intonation. Journal of Phonetics 26: 45–69. Irvine, J.T. 2001. “Style” as distinctiveness: The culture and ideology of linguistic differentiation. In Style and Sociolinguistic Variation, P. Eckert & J.R. Rickford (eds), 21–43. Cambridge: CUP. Joos, M. 1959. The isolation of styles. Georgetown University Monograph Series on Languages and Linguistics 12: 107–113. Jun, S. & Fougeron C. 2000. A phonological model of French intonation. In Intonation: Analysis, Modeling and Technology, A. Botinis (ed.), 209–242. Dordrecht: Kluwer. Jun, S. & Fougeron C. 2002. Realizations of accentual phrase in French intonation. Probus 14: 147–172. Knecht, P. & Rubattel, C. 1984. A propos de la dimension sociolinguistique du français en Suisse romande. Le français moderne 52: 138–150. Labov, W. 1972. Sociolinguistic Patterns. Philadelphia PA: University of Pennsylvania Press. Levin, H., Schaffer, C.A. & Snow C. 1982. The prosodic and paralinguistic features of reading and telling stories. Language and Speech 25: 43–54. Léon, P. 1971. Essais de phonostylistique. Montréal: Didier. Miller, J. 2007. Swiss French Prosody: Intonation, Rate, and Speaking Style in the Vaud Canton. PhD dissertation, University of Illinois at Urbana-Champaign. Métral, J. 1977. Le vocalisme du français en Suisse romande. Cahiers Ferdinand de Saussure 31: 145–176. Odin, A. 1886. Phonologie des patois du canton de Vaud. Halle: Max Niemeyer. Prikhodkine, A. 2002. Lexique régional et insécurité linguistique. In Le français parlé dans le domaine francoprovençal: une réalité plurinationale, P. Singy (ed.), 139–163. Bern: Peter Lang. Simon, A. 2003. La variation prosodique régionale en français dans les données conversationnelles: Propositions théoriques et méthodologiques. Bulletin Phonologie du Français Contemporain 3: 99–113. Singy, P. 1995. Les francophones de périphérie face à leur langue: Étude de cas en Suisse romande. Cahiers Ferdinand de Saussure 49: 213–235. Singy, P. 1996. L’image du français en Suisse romande: Une enquête sociolinguistique en Pays de Vaud. Paris: L’Harmattan. Singy, P. 2002. Accent vaudois: Le complexe des riches? In Le français parlé dans le domaine francoprovençal: Une réalité plurinationale, P. Singy (ed), 165–186. Bern: Peter Lang. Vouga, J. 1984. L’accent vaudois. In La vie quotidienne, J. Vouga (ed), 289–315. Lausanne: Editions 24 heures.

The immigrant factor in phonological leveling Tim Pooley

London Metropolitan University This chapter investigates the relationship between immigration and the generally high degree of phonological leveling in France in the latter part of the 20th. century. It draws together the findings of a number of studies of the pronunciation of young people of Maghrebian origin in French cities (Paris, Grenoble, Marseille, Lille) in the late 20th century, with special reference to Lille in the later sections. Some common patterns emerge. Far from exhibiting a strong migrant variety, the speech of young people of Maghrebian origin conforms, in general, to the classic sociolinguistic pattern of their metropolitan French peers, the boys maintaining marked regional features, generally as minority variants, to a greater extent than the girls.

Immigration and vernacular French Large-scale inflows of foreign nationals into a community clearly have considerable potential to cause radical changes to its linguistic profile and practices, going far beyond the selected focus of this chapter, i.e. the effects of the migrant presence on the pronunciation of vernacular varieties of the mainstream language, French. In a previous study of a historic case of Flemish migration to Lille in the late 19th and early 20th centuries (Pooley, 2006a), I argued that migrants, having in the early stages adopted the local vernacular variety in somewhat less marked form than the native varieties of people of comparable social status, they and the areas where they settled in greatest concentrations, came over the course of a few generations to be perceived as the principal guardians of the vernacular. In this chapter I shall attempt to show that apart from a few differences of detail, a broadly similar pattern emerges from the research findings on the most studied migrant group of the late 20th century, the Maghrebians, as documented in cities such as Paris, Marseille and Grenoble. The final two sections focus on my own work in Lille, concerning which a preliminary report will be given on newly analyzed data recorded in 2005.



Tim Pooley

The findings presented here should be weighed against a background of generalized phonological leveling, which can be characterized as the increasing adoption by people of all social and regional backgrounds in France of a supra-local variety. Historically, this variety emanated from the Paris region, but in its expansion, it has lost some specifically Parisian characteristics, and may be construed as a regionally and increasingly socially neutral variety. Both the spread of this supralocal variety over the course of the last century and the regional pockets of resistance to it that remained at the turn of the 20th century are described in recent overviews for northern (Pooley 2006b) and southern France (Pooley 2007). The Maghrebians in Paris, Grenoble and Marseille Both academic studies and media discourses concerning contemporary youth vernaculars have increasingly focused on migrant groups, where Maghrebians are generally the largest ‘ethnic’ element in multi-ethnic communities. While media hype and certain filmic representations (e.g. Kasovitz’ La Haine ‘Hate’ in 1995) point to the formation of (a) new dialect(s) more or less incomprehensible to the majority of the population, such an implied claim does not resist the scrutiny of careful fieldwork that takes account of local situations (Lepoutre 1997: 430; Fagyal 2004: 43). Both the latter researchers studied Les 4000 estate in the La Courneuve in the northern suburbs of Paris. Lepoutre’s work is ethnographic in approach and strongly recommends careful micro-sociolinguistic studies to complement his description of the social world of adolescents in a multi-ethnic suburb in the 1990s, where Maghrebians constituted the largest migrant group. Fagyal’s study of the lexis of adolescents in the same area in the early part of the current century, failed to find convincing evidence of new-dialect formation despite the subjects’ perception that they have a particular way of speaking. With regard to phonology, Armstrong and Jamin (2002) have argued that the banlieues - the suburban areas around large French cities, characterized by low-cost housing and ‘sink’ estates - are the main source of vernacular innovation against a general backdrop of leveling to(wards) a supra-local norm (Pooley 2006b, 2007). Curiously, the first variationist study of the Paris banlieue, that of Laks (1978) in Villejuif to the south of the capital, focused on features such as the dropping of ~t] centre and [tab] table), and linking liquids (/t/ and /l/ after obstruents, as in [sa words like [epi] et puis ‘and then’ – phenomena which may now be considered compatible with supra-local norms. That said, Laks’s painstaking ethnographic work was rewarded with positive results, since he was able to show that integration into a network of young lads centered around the Maison pour tous in Villejuif corresponded to significantly heavier use of features selected for analysis.



The immigrant factor in phonological leveling 

Table 1.  Divergent phonological features in La Courneuve (Armstrong and Jamin 2002: 132) Feature

Examples

1 Use of back /a/ in all contexts la table [la tab], or c’est grave [sε grav] 2 /f/ raised towards /o/ la mort [lamot] la police [lapolis] ta mère [tamet], 3 Closing of /ε/ to /e/ before word-final uvular je suis vert [∫Ci vet] approximant j’ai peur [ʒe pø˜t] 4 Raising and lengthening of /œ/ to [ø˜] before /r/ /t/ realised as [t∫] and /d/ as [dʒ] before /i y w/ tu dis [t∫y di] toi [t∫wa] 5 Affrication of dental and velar stops je veux dire que [ʒvødʒi˜k∫ә˜]. /k/ realised as [k∫] before /a a i ε 6 y œ ø/ and after nasal e.g. qui [k∫i] donc [d¢k∫]. 6 Pharyngeal /r/ ta mère [tameA]

More recent studies have focused on features more clearly indicative of marked urban vernaculars, considering variants characteristic of traditional Parisian slang (known as Parigot) of pre-World War II times and of Dialectal Arabic, e.g. Jamin (2005) in La Courneuve. The dual origin of the features divergent from the supralocal norm is clearly illustrated by the vernacular features noted by Armstrong and Jamin (2002) reproduced in Table 1. Features 1 to 5 can in no way be attributed to the input of migrants of either southern European or North African origin. Ethnically speaking, they can be categorized as white vernacular norms. They are characteristic of traditional slang and feature, or have featured, in a number of regional varieties, which clearly predate significant levels of migration from the Maghreb. Back /a/ is what might be called a socially split variant, characteristic of both vernacular and somewhat oldfashioned standard varieties. In Parisian and Normandy vernaculars it is what James Milroy (1992) appositely called a lexically defined small-set variant, whereas in the northern (Nord–Pas-de-Calais) and western France (Brittany), it tends to occur in open word-final syllables. What appears to be happening in La Courneuve is that some speakers are treating it as a purely phonetic variant and using it (variably) in all contexts. Pre-rhotic close /o/ and raised /e/ are also to be found in older and marginalized vernaculars whereas the raising of /œ/ to [ø] before /r/, admittedly without lengthening, is arguably a (variably occurring) feature of the supra-local norm. Affrication after /k/ and /t/ is a feature that has been associated with Parisian vernacular for several centuries (cf. Rosset, 1911: 314) leaving only pharyngeal /r/ as an arguably Maghrebian feature. In this regard, however, current perceptions, of which speakers are very aware, are undoubtedly of greater import than historical accounts of which they are generally uninformed.



Tim Pooley

For feature 5 and to a lesser degree 1 and 6, Jamin, Trimaille and GasquetCyrus (2006) argue that such divergence from the supra-local norm is matched by homogenization across pluri-ethnic (but Maghrebian-dominated) areas of several cities. The use of back /a/ as a phonetic variant has been observed not only by Jamin in La Courneuve in the northern Paris suburbs, but also by Trimaille in Chorier-Berriat, a multi-ethnic area of Grenoble. That said, only Jamin’s data reflect the multi-ethnic character of the area investigated through the choice of informants, Trimaille having elected to restrict his analysis to subjects of Maghrebian origin. Pharyngeal /r/ has not only been noted in both Paris and Grenoble, but also in Perpignan by Pickles (2001). The findings of this latter study show this variant to be used by school students of Maghrebian origin significantly more than speakers of other ethnicities but not particularly those from social-problem areas. The affrication of /t/ and /d/ has been observed again in both the Paris and Grenoble studies, to which may be added the perceptual approach of GasquetCyrus (Jamin, Trimaille and Gacquet-Cyrus 2006) who adduces convincing evidence that the feature is perceived by the Marseillais as characteristic of their fellow citizens of Maghrebian origin, living in the northern suburbs (quartiers nord, hence the term accent QN). In the Paris and Perpignan studies, all the variants turn out to be minority features, that is, used in fewer than half of all possible cases by those who use them at all. In Grenoble, however, two of the three variants are strong majority features for some speakers who even use them in reading styles. Table 2 suggests that for the young people of various ethnicities in the northern Paris suburbs, affrication is an indicator as it is used more by those of Maghrebian origin than the Metropolitan French1 or members of other ethnic minorities, but there are no significant differences between reading style and interview style. Table 2.  Affrication of dental and velar stops in two styles in La Courneuve (Jamin 2005: 43) Ethnicity Metropolitan French (MF) parents Maghrebian parents Parents of other migrant origin

Interview Style N % 3,349 4,465 2,746

  6.4 21.6 13.6

Reading Style N % 579 741 379

  5.3 22.9 16.7

1. The term Metropolitan French is used in this chapter to refer to the inhabitants of urban areas of France who are ethnically French, in contradistinction to those who are of Maghrebian or other ethnic origin.



The immigrant factor in phonological leveling 

Table 3.  Affrication according to style, age and gender in La Courneuve (Armstrong and Jamin 2002: 133)

Males 15–25 Females 15–25 Males 30–50 Females 30–50

Casual Style N

%

Reading Styles N

%

1,369    387     18     52

22.5 13.4 0.6 1.4

111   26    1    1

17.3 8.3 0.2 0.2

For all groups, boys use affrication more than girls and adolescents and young adults (aged 15 to 25) use the variants more than mature adults (aged 30 to 50). Table 3 shows that it is fundamentally a significant minority variant for the younger age group immersed in street culture, while it is used only occasionally by the adults who have settled lives characteristic of mainstream values. This is symptomatic of the disavowal by most young people, once they leave school and enter the wider world, of the adolescent value system that they espoused so intently while attending school in the banlieue (cf. Lepoutre, 1997). While there are some grounds for claiming that in multi-ethnic urban areas, where the Maghrebians are the largest migrant group, certain phonological features and possibly some lexical items are used more by young people of North African origin than their peers from other ethnicities, neither affrication nor pharyngeal /r/ are, generally speaking, majority features for their users. As for lexical items, most, if not all, lexicographical studies signally fail to demonstrate differentiation of use by speakers with different social (particularly ethnic) characteristics. All in all, the findings for France, although by no means uninteresting, are a far cry from the new-dialect formation described by Fox (2007), for instance, in her study of the English of young people of Bangladeshi origin in the East End of London, where she presents several innovative vowel features that are strongly associated with this migrant community. Lille In Pooley (1996; 2000) I reported on the behavior of a class of 15-year-old school students (born in 1979 or 1980) in a special-needs section (SEGPA- Section d’Education Générale et Professionnelle Adaptée) in Marcq-en-Barœul. This study was based on a corpus recorded in 1995, the first of six (the others were recorded in La Madeleine (1997 and 2004), Roubaix (1998 and 2004) and Lille-Sud (2005)), designed to constitute a set of micro-level studies to investigate the linguistic



Tim Pooley

behavior and repertoire of multiethnic groups of young people attending school in various parts of the Lille metropolitan area (Lille-Métropole). While the interview sections of the corpora gathered prior to 2004 have been analyzed with regard to knowledge and perceptions of what in traditional terms was the local dialect – a variety of Picard – (Pooley, 2004a), this is the first interim report on the behavioral aspects in the use of French of the post-1997 corpora, using one of two classes recorded in Lille-Sud in 2005. To recall the main findings of the Marcq-en-Barœul study, it is perhaps first of all apposite to remark on the vitality of two marked regional features (i.e. nonsupra-local), i.e back /a/ [sa] ça ‘that’ and open /o/ [kft] côte ‘hill’ in an otherwise more or less completely dedialectalized (or preferably depicardized) vernacular. Secondly, the data showed strongly marked gender and ethnic differences, with back /a/ emerging as a significantly white masculine feature, while among the female subjects, there was little difference in the behavior of the Metropolitan French and Maghrebian girls. Indeed, the greater use of back /a/ among the Maghrebians by the girls compared to the boys was interpreted as being indicative of their greater integration into their gender peer group (see Table 7). The female group manifested a commendable level of interethnic harmony, whereas the male group manifested signs of interethnic rivalry and conflict. By way of indication, this is one of the few situations where respondents have manifested a positive value for the Le Pen Index, which involves indicating support for the Far Right leader of the National Front party to become president or approval of those who vote for him and his party. Curiously too, style-shifting patterns for open /o/ were completely reversed for the two ethnic groups, with the Metropolitan French adolescents showing the predicted pattern of greatest use of [kft]-type forms in Group Style (spontaneous conversation) as opposed to Word List, Reading Passage and Interview, whereas the Maghrebians used the greatest number of [f] in Word List Style and the least in Group Style, suggesting that they perceived forms like [kft] as standard. A Regional Loyalty Index (RLI) (Pooley, 1996; 2004a) based on responses about living in the region, the people of the region and their preferred choice of residence in adulthood proved sensitive to ethnicity with only French subjects getting the highest possible value. Among the Metropolitan French teenagers a high RLI value correlated significantly with frequent use of back /a/. The analysis presented on Lille-Sud is based on a partial analysis of half of the corpus. Two classes were studied, the Cinquième (corresponding to the second year of secondary education, born 1991 aged 13 in January 2005) and Quatrième (corresponding to the third year of secondary education, born 1990) and only analysis of data from the former is presented here. It is perhaps worth recalling that Lille-Sud was selected as a fieldwork site because of its reputation as a socially difficult ethnically mixed area right at the heart



The immigrant factor in phonological leveling 

Table 4.  Interethnic friendship indices in school (IFIS) and outside school (IFIO) LilleSud (2005) Index Value 2.5 2 1.5 1 0

IFIS

IFIO

2 3 2 3 4

1 1 1 1 10

of Lille-Métropole. The Cinquième class had 15 students in 2004–5, one of whom was absent, having suffered severe burns as a result of preparing a Molotov cocktail during the summer holidays, and one of whom produced no audible speech. Of the 13 other students, there were seven boys and six girls; seven were of Metropolitan French ethnicity (four boys, three girls) and five of Maghrebian origin (three boys, two girls; three Moroccans, two Algerians) and one girl of FrancoMoroccan parentage. The students of Maghrebian background generally had more settled home lives in or near Lille-Sud, while two of the French boys were being fostered and one French girl had recently returned to her native area after a stay of several years in Toulouse. From the institutional perspective, the class showed commendable levels of interethnic harmony and this was confirmed by the interethnic friendship indices in Table 4. 71% (counting the girl of mixed parentage in each of her guises) of the students named at least one friend of different ethnic background among their three best friends at school (Interethnic Friendship Index at School – IFIS). Outside school (Interethnic Friendship Index Outside School – IFIO), interethnic friendships were considerably fewer. In both cases, claimed interethnic friendships are more frequent than inter-gender friendships. In contrast to the Marcq-en-Barœul group, no students manifested the slightest positive reaction when questioned about Jean-Marie Le Pen and the National Front party. Linguistic and sociolinguistic comparison of 1995 and 2005 corpora While this section focuses principally on the two main regional variants of back /a/ in word-final open syllables and open /o/ in closed syllables in items which generally take closer variants in mainstream varieties, a preliminary analysis of the data warrant two series of observations. Firstly, one or two of the Metropolitan French informants occasionally used marked Picard forms that the 1995 data seemed to suggest were no longer used, in particular the non-vocalization of /l/ as



Tim Pooley

in [butεl] bouteille ‘bottle’. Cases of word-final consonant devoicing (WFCD) were also noted in the usage of a number of informants, including in reading styles. Secondly, preliminary listening suggests that the two ‘Maghrebian’ variants, discussed above, pharyngeal /r/ and affrication are used occasionally and to a greater degree by the boys. With regard to the use of back /a/, it was noticeable that strongly backed variants in the region of [%] or [a] were very much a minority feature, compared to the more generalized front realization [a]. On many occasions, a kind of intermediate form was used, transcribed as [!] with some degree of backing but by no means fully back. Comparing the two corpora (Table 5) shows the Lille-Sud group used strongly backed /a/ only in a minority of cases and significantly less than the Marcq-en-Barœul informants, even if tokens transcribed as [!] were included, as shown by the bracketed figures. Table 5 also suggests that phonological context is of considerable importance, with strong /a/-backing significantly favored in word-final and phrase-final (pre-pausal) position. Moreover, the variant shows clear signs of lexicalization, i.e. the phonologically arbitrary restriction to a small range of lexical items, particularly ça ‘that’ and là ‘there’ which are the most crucial sites because of their high frequency, but a few less frequent items, voilà ‘here you are’, déjà ‘already’ and toi ‘you’ score well above the overall mean. A number of structurally anomalous phenomena were observed, such as the difference between toi ‘you’ and moi ‘me’ or bas ‘low’ and là-bas ‘over there’ where the latter item never occurs with strong backing pre-pausally, whereas the firstmentioned scores above the mean. Open /o/, on the other hand, was used with greater consistency in the 2005 data than in those of 1995 and indeed it is a clear majority form in the corpus as a whole. Again, certain lexical anomalies may be noted: chose ‘thing’ favors the close variant, whereas quelque chose ‘something’ favors [f]. Some individual itemsare in Table 5.  Style shift in use of back /a/ and open /o/ (Marcq-en-Barœul, 1995; Lille-Sud, 2005) Style Word List Reading Passage Interview Group Conversation word-final Group Conversation word-final, phrase-final

Marcq-en-Barœul Back /a/ Open /o/ 19% 18% 24%

41%

39% 39% 39%

53%

Lille-Sud Back /a/ Open /o/ 7% (21) 9% (25)

49% 38%

4% (7)

74%

14% (27)

77%



The immigrant factor in phonological leveling 

Table 6.  Use of back /a/ (word-final phrase-final) and open /o/ according to gender in group conversations (Marcq-en-Barœul, 1995; Lille-Sud, 2005) Gender

Marcq-en-Barœul Back /a/ Open /o/

Boys Girls

42% 40%

60% 34%

Lille-Sud Back /a/ Open /o/ 21% (34) 5% (15)

77% 71%

Table 7.  Use of back /a/ and open /o/ according to gender and ethnicity (group conversation ) (Marcq-en-Barœul, 1995; Lille-Sud, 2005) Gender Boys Girls

MF 1995 Back /a/

Open /o/

42% 40%

60% 34% MF 2005

Boys Girls

Back /a/

Open /o/

22% 5%

74% 58%

Maghrebians 1995 Back /a/ Open /o/ 21% (34) 5% (15)

77% 71%

Maghrebians 2005 Back /a/ Open /o/ 20% 6%

90% 17%

variant, such as rôle ‘role’ [rfl] and gauche ‘left’ [go∫]. Although the open variant was slightly more frequent in phrase-final position, this is not significant as in the case of /a/-backing. In subsequent tables, all occurrences are considered together, while for back /a/, only strongly backed variants are taken into consideration. Table 6 shows classic gender differentiation patterns for both back /a/ and open /o/, albeit attenuated in the Marcq-en-Barœul data by the markedly differentiated behavior of the Maghrebian boys, who, being relatively new to the school, were out of kilter with the class norms. In the Lille-Sud data, back /a/ is a significantly masculine feature in a context where overall usage of the feature has decreased considerably. Open /o/, on the other hand, manifests both increased percentage frequency and a much lower level of gender difference. The results presented in Table 7 show that differential frequencies of usage for the two vernacular variants cross-cut with gender and ethnicity in contrasting ways in the 2005 data compared to those of 1995. In the Lille-Sud corpus, back /a/ manifests orthodox gender patterns with remarkably similar frequency rates across the two ethnicities, in contrast to the Marcq-en-Barœul findings, where the gender pattern was sharply reversed for the Maghrebians. For open /o/, the 2005 results suggest than ethnicity outweighs gender, insofar as the differentials between boys and girls are normal for both groups, whereas the Maghrebians used



Tim Pooley

[à] considerably more, even bearing in mind that the vernacular variant is the clear majority form for all. Moreover, and again in contrast to the 1995 study, the perception of the standard form is the same for both ethnicities, whereas the earlier results showed style-shift patterns suggesting that the Maghrebians construed the open variant as the standard form. As regards the Regional Loyalty Index, subjects in both corpora with a maximum score for three positive responses correlated with significantly higher use of back /a/ and open /o/. The crucial difference is that in the 1995 data, only Metropolitan French subjects had a maximum index value of 3, whereas in 2005 some Maghrebians manifested this highest level of attachment to the region (Table 8). For Metropolitan French subjects, it was hypothesized that indices pointing to favorable attitudes towards Maghrebians would correlate with relatively low scores for the regionally marked variants. Tables 9 and 10 give some support to this hypothesis with significantly greater use of open /o/ by informants with zero IFIS (Interethnic Friendship Index at School) and significantly greater use of back /a/ by those with a zero value for IFIO (Interethnic Friendship Outside School) (p