TheAcquisition of Sociolinguistic Competence in a Lingua Franca Context 9781783091447

Citation preview

The Acquisition of Sociolinguistic Competence in a Lingua Franca Context

SECOND LANGUAGE ACQUISITION Series Editor: Professor David Singleton, University of Pannonia, Hungary and Fellow Emeritus, Trinity College, Dublin, Ireland This series brings together titles dealing with a variety of aspects of language acquisition and processing in situations where a language or languages other than the native language is involved. Second language is thus interpreted in its broadest possible sense. The volumes included in the series all offer in their different ways, on the one hand, exposition and discussion of empirical findings and, on the other, some degree of theoretical reflection. In this latter connection, no particular theoretical stance is privileged in the series; nor is any relevant perspective – sociolinguistic, psycholinguistic, neurolinguistic, etc. – deemed out of place. The intended readership of the series includes final-year undergraduates working on second language acquisition projects, postgraduate students involved in second language acquisition research, and researchers and teachers in general whose interests include a second language acquisition component. Full details of all the books in this series and of all our other publications can be found on http://www.multilingual-matters.com, or by writing to Multilingual Matters, St Nicholas House, 31–34 High Street, Bristol BS1 2AW, UK.

The Acquisition of Sociolinguistic Competence in a Lingua Franca Context

Mercedes Durham

MULTILINGUAL MATTERS Bristol • Buffalo • Toronto

In loving memory of my cousin David Kosofsky

Library of Congress Cataloging in Publication Data Durham, Mercedes. The Acquisition of Sociolinguistic Competence in a Lingua Franca Context/Mercedes Durham. Second Language Acquisition: 75 Includes bibliographical references and index. 1. Sociolinguistics—Switzerland. 2. Interlanguage (Language learning). 3. Language and languages—Variation. 4. Second language acquisition—Switzerland. 5. Electronic mail messages—Social aspects. 6. Communicative competence—Switzerland. 7. Switzerland— Languages. I. Title. P40.45.S9D87 2014 306.4409494–dc23 2013041722 British Library Cataloguing in Publication Data A catalogue entry for this book is available from the British Library. ISBN-13: 978-1-78309-143-0 (hbk) Multilingual Matters UK: St Nicholas House, 31–34 High Street, Bristol BS1 2AW, UK. USA: UTP, 2250 Military Road, Tonawanda, NY 14150, USA. Canada: UTP, 5201 Dufferin Street, North York, Ontario M3H 5T8, Canada. Copyright © 2014 Mercedes Durham. All rights reserved. No part of this work may be reproduced in any form or by any means without permission in writing from the publisher. The policy of Multilingual Matters/Channel View Publications is to use papers that are natural, renewable and recyclable products, made from wood grown in sustainable forests. In the manufacturing process of our books, and to further support our policy, preference is given to printers that have FSC and PEFC Chain of Custody certification. The FSC and/or PEFC logos will appear on those books where full certification has been granted to the printer concerned. Typeset by Techset Composition India (P) Ltd., Bangalore and Chennai, India. Printed and bound in Great Britain by the CPI Group.

Contents

Acknowledgements

ix

1

English as a Lingua Franca 1.1 English as a Global Lingua Franca 1.2 Classifying English Users 1.3 From ESL and EFL to ELF and Beyond 1.4 The Sociolinguistics of ELF

1 1 5 6 11

2

Second Language Variation 2.1 Introduction 2.2 Sources of Non-native Variation 2.3 Type of Non-native Speaker: Interlanguage Versus High-level Competence 2.4 SLV and ELF 2.5 Previous Work on SLV 2.6 Where Does ELF Fit into SLV?

14 14 15 19 20 20 24

3

English as a Lingua Franca in a Multilingual Context: Switzerland 3.1 Languages in Switzerland 3.2 2000 Census Results 3.3 The Rise of English and its Causes 3.4 English as a Swiss Lingua Franca

29 29 34 39 45

4

Data 4.1 Introduction 4.2 Swiss Data – IFMSA-Switzerland 4.3 Languages Used in the Association 4.4 The Situation at Meetings (and Interviews) 4.5 The Mailing List 4.6 Type of English Used 4.7 Native English Data 4.8 Discussion

47 47 48 49 51 52 55 56 57

v

vi

5

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Methodology 5.1 Comparing Linguistic Varieties 5.2 Multivariate Analysis and its Use for Comparative Sociolinguistics 5.3 Categorizing the Results 5.4 Features Analyzed

59 59 61 62 64

6

Future 6.1 Introduction 6.2 Future Forms 6.3 Future Forms in the Source Languages 6.4 Previous Studies 6.5 Setting Up the Analysis 6.6 Results 6.7 Discussion 6.8 Conclusion

67 67 67 70 71 71 72 79 80

7

Relatives 7.1 Introduction 7.2 Relative Clauses 7.3 Relative Clauses in the Source Languages 7.4 Previous Studies 7.5 Analysis and Results 7.6 Multivariate Analysis 7.7 Discussion 7.8 Conclusion

82 82 83 86 88 92 104 109 109

8

Complementizers 8.1 Introduction 8.2 Complementizers 8.3 Complementizers in the Source Languages 8.4 Previous Studies 8.5 Analysis and Results 8.6 Multivariate Analysis 8.7 Discussion 8.8 Conclusion

111 111 112 113 113 114 128 131 132

9

Lexical Variation: Also, As Well and Too 9.1 Introduction 9.2 Also, As Well and Too 9.3 Additive Adverbials in the Source Languages 9.4 Previous Work on Additive Adverbials 9.5 Analysis and Results 9.6 Discussion 9.7 Conclusion

134 134 135 136 137 138 143 144

Content s

vii

10 Discussion 10.1 General Considerations 10.2 Features 10.3 Interpretations 10.4 ELF and Sociolinguistic Competence

145 145 146 148 152

11 Conclusion

155

References Index

158 165

Acknowledgements

As a native English speaker and a Swiss citizen, it’s no great surprise that I’ve ended up interested in English as a lingua franca in Switzerland, but without the teachers and mentors who made me the linguist I am today, I wouldn’t have been able to study it the way I have. Thanks especially to Jennifer Smith, Sali Tagliamonte and Peter Trudgill, who taught me everything I needed to know about sociolinguistics, language variation and change, and far more. I couldn’t have asked for better teachers. I am grateful too to many colleagues and friends for valuable discussions through the years which have helped shape my ideas about sociolinguistic competence and how it is acquired. Thanks to Robert Bayley, Jean-Marc Dewaele and Robert Mougeon for organizing and letting me present on panels on the acquisition of variation at Sociolinguistics Symposium 15 and NWAV 33, but also to Tess Fitzpatrick, Miriam Meyerhoff, Jonathan Morris, Vera Regan, Erik Schleef and Devyani Sharma and many others. Some of my thanks go to people who likely have long forgotten the help they gave me – first and foremost, all the members of the Swiss Medical Students Association (IFMSA-CH) who kindly gave me their time (and their emails) and allowed me to ‘use’ their association for the project. Thanks are also due to David Allerton, Yvonne Droeschel, Lucas Rosenberger and Richard Watts, as fellow members of the Fonds National Suisse team, and to the FNS itself whose research grant funded the initial project on English in Switzerland on which this book is based. The book was completed while I was working at the University of Aberdeen and then Cardiff University. Thanks to my colleagues Robert McColl Millar and Lisa Bonnici for their help while I was in Aberdeen and to all my new colleagues at Cardiff for making me feel so welcome. Thank you to the team at Multilingual Matters for general helpfulness, quick responses to emails and patience when my move to Wales delayed the initial submission of the manuscript, and also to the reviewer for very useful suggestions which allowed me to improve the book tremendously. I’m grateful also to the friends who kindly went through drafts of chapters when I needed fresh eyes on them: Janet Cruickshank, Nanna Haug

ix

x

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Hilton, Sandra Jansen and finally David Duff, who not only read far more than he bargained for, but who made the process much more fun. Some friends’ help is less tangible but just as valuable – thanks to my wonderful ‘cheerleaders’: Marina Betevic, Heather Brown, Isabelle Ensass, Katrin Sprohar and a whole host of tweeps, who were unflagging in their support and enthusiasm (or nagging as necessary). Thanks to my students for never once seeming too bored when I waxed lyrical about variation and acquisition. Finally, thanks to my family (especially my mom) for always believing in me and sharing my love of languages. Cardiff, 2013

1

English as a Lingua Franca

1.1 English as a Global Lingua Franca As English has increased in use across the world, so have attempts to account for its spread and to analyze the new ways it is used. While discussions about English in native, in foreign language and in lingua franca contexts are wide ranging and diverse, one aspect which has received relatively little attention so far is the sociolinguistics of English as a lingua franca (henceforth ELF). The present volume focuses on this topic and examines how non-native speakers of ELF deal with native sociolinguistic competence and whether they are able to acquire the variable patterns native speakers have. As early as the 18th century, English was seen to be on course to become a world language: English will be the most respectable language in the world and the most universally read and spoken in the next century, if not before the close of this one. (President John Adams, 23 September 1780, in Kachru, 1982: 2) From the vantage point of the 21st century, this impression seems remarkably prescient: English is not merely the language of England and other countries of the British Isles, but a language used natively across the globe by millions of speakers (Crystal, 2003). The specific processes involved in this spread have been examined in detail and numerous studies have presented and described the wealth of varieties of English worldwide (for example, Cheshire, 1991; Gorlach, 1998, 2002; McArthur, 1998; Trudgill & Hannah, 1982). Recent estimates suggest that there are around 400 million speakers of ENL and a similar number (around 430 million) who use English as a second language in countries where it has some kind of official status (Crystal, 2003: 67). The latter countries are often those which were affected by British colonization and in which English was established as an important second or official language, but not as a first language (for example India, Cameroon, Nigeria and Jamaica). 1

2

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

However, while these numbers are substantial, it is the foreign language speakers of English, rather than the first and second language speakers, who have made it the truly international and universal language of the 20th and 21st centuries. Ten years ago, Crystal (2003: 68) estimated that 750 million people spoke English as a foreign language with at least a ‘medium level of conversational competence’ and, more recently, Bolton and Graddol (2012: 3) found that nearly 400 million people in China, roughly a third of the population there, were English learners. Taking the figures for native, second and foreign speakers of English together, it can be said that at least 25% of the world’s population speaks English reasonably proficiently (Crystal, 2003: 69), a striking statistic given that nearly 7000 languages are spoken throughout the world (Lewis, 2009). While not quite reaching John Adams’ ‘universal’ distribution, it is clear that English is the most widely spoken language on the planet. This spread was not achieved purely through colonization or territorial expansion, but rather through increased globalization and the export of British and American culture, literature, music and technology. Much recent work on World Englishes (Brutt-Griffler, 2002; Crystal, 2003; Graddol, 2006; Jenkins, 2003; Schneider, 2007, 2011, in addition to those already mentioned above) has been concerned both with the reasons for its exponential growth, and with the consequences of this growth on the language of both native and non-native speakers. As the proportion of people who use English as a non-native language has grown, so have the number of reasons why it is seen as a valuable additional language. Many non-native speakers have learnt (or are learning) English not because they necessarily intend to use it with native speakers, but rather because it is now used as an international lingua franca in many areas of business, industry, medicine, science and education. This use of ELF is a relatively recent phenomenon, and while it is taken for granted today, it was noteworthy in the 1980s, as evidenced by Ferguson’s discussion of language use at European scientific and academic conferences: English is widely used on the European continent as an international language. Frequently conferences are conducted in English (and their proceedings published in English) when only a few of the participants are native speakers. At such conferences the English spoken often shows features at variance with the English of England but shared by the other speakers. (Ferguson, 1981: xvi) It is not just in Europe and at conferences that such language use occurs, but in some form or other in most parts of the world. This is not to imply that English is used everywhere (despite what some English-speaking tourists may think or desire), but rather that if a non-local language is used for conversation, it is likely to be English. The ubiquity of English affects its overall perception and its more general use.

English as a Lingua Franca

3

Before discussing the effects of ELF on speakers, native and non-native, I will address the important preliminary question of whether English is like other lingua francas. Broadly speaking, a lingua franca is a language used by speakers of different languages to communicate with each other (Meierkord, 2006). People have used lingua franca languages for trade, for diplomacy and for a wide range of other reasons throughout history. In some instances the language used as a lingua franca was the native language of one group and a second or third language of another group, but often lingua francas were chosen as neutral languages between speakers of different languages. Additionally, mixed language varieties such as pidgins and creoles were sometimes used as lingua francas. Pidgins are languages which arise through the contact of speakers of two or more languages using a language that is native to none of the groups. A pidgin is said to be a language with no native speakers, and a creole is a pidgin which has acquired native speakers (this is a somewhat simplified explanation of the difference between pidgins and creoles, but suffices for the purposes here; see Todd, 1991: 48ff). To give two brief examples of lingua francas: Latin was widely used as a lingua franca across the Roman Empire and French was a vital tool within diplomatic circles across Europe (and Russia) from the 17th century until the 20th century. Often the people using a lingua franca are not as proficient in it as native speakers are, but because the main purposes of lingua franca use have to do with communicative urgency and a desire to get one’s point across, this is not generally seen as problematic. Although there have been a number of important lingua francas historically, including the two mentioned above, and various languages are currently used as lingua francas in specific countries (for example Swahili in East Africa and Hausa in Niger and Nigeria), the situation of English is unique. Never before has one language been spoken by such a high proportion of people across the world and in so many different contexts. As noted, many of the earlier lingua francas were used primarily for trading or diplomacy, and the number and proficiency of the non-native speakers would have been relatively low. English is completely different in this respect, as research on ELF has underlined. English is a global lingua franca, the de facto choice across numerous bilingual and multilingual communities all over the world. Consequently, English as a language in the 21st century is characterized by its many different kinds of speakers and its different kinds of uses. To understand ELF fully, the models of use provided by native speakers and by foreign language learners are not sufficient and the topic must be approached from a new angle. Attention has therefore turned to the non-native contribution to the use of English: by examining this, researchers have gained new insight into where the language may be heading. The present study is a contribution to this new field of inquiry. Its focus is ELF in Europe and, more specifically, Switzerland, a complex linguistic community which clearly brings out the sociopolitical factors that often shape lingua franca usage.

4

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

To clarify differences between English as a native language (ENL), English as a second language (ESL), English as a Foreign Language (EFL) and ELF, this first chapter will present a general view of the field of World Englishes and EFL, before addressing the European situation more closely and discussing the merits of a sociolinguistic approach to the use of ELF. Although most research on English is related to native speaker varieties, there is a longstanding tradition of studies which attempt to account for the ‘other side of English’: that is to say, the side which ‘is concerned with English as the “other tongue”, or as a second language’ (Kachru, 1982: 1). These studies take as a starting point the fact that the purposes to which English is put within a country have an important bearing on the type of English used. The growing importance of English has affected its structure: the more English is used, the more likely it is to acquire new distinctive forms. Indeed, recent research conducted across Europe has come to the conclusion that English is in the process of acquiring a new pan-European form (see Erling, 2004; Seidlhofer, 2001, 2011) due to its increasing lingua franca use. One reason for this is that, in the past 50 years, the status of English in Europe has changed in many countries from a foreign language to a type of national language used for economic and communication purposes (Cenoz & Jessner, 2000; Hoffman, 2000; Phillipson, 2003). This explains why more people want to learn English, creating a vicious circle (as some people regard it) whereby as the use of English worldwide increases, the desire to learn it and use it grows too, which then further adds to its overall use and the number of people wanting to learn it. Some researchers see the spread of English as a further example of Britain’s colonizing tendencies or evidence of American globalization. In their eyes, this makes ELF a threat to the vitality of other languages, or at least as a threat to the use of other languages in a wider sphere (Phillipson, 2003; Skutnabb-Kangas, 2009). Although these concerns have some validity, they neglect the fact that in many instances it is the choice of the speakers themselves to use ELF rather than a top-down requirement. Strictly speaking, it is not up to linguists to judge people’s linguistic choices, whether it be to use the non-standard in one’s own language or to favour a single language as a lingua franca. Instead, the task of the researcher is to impartially observe, report and try to understand as much as possible about language as it is currently used by speakers around the world. Having introduced the notion of ELF, the sections below provide a broad classification of different types of English speakers and detail the challenges of finding suitable categories for speakers of ELF in the 21st century (Section 1.2), examine the shift to ELF and the implications of this kind of use for acquisition (Section 1.3), and finally treat the potential for new research on ELF using sociolinguistic methodology (Section 1.4). These will help to establish precisely what ELF is and what it is not, which in turn will ultimately inform the discussion of what features ELF itself might contain and how these might be transferred from native varieties.

English as a Lingua Franca

5

1.2 Classifying English Users Many models used to classify English users focus on native, second and foreign language categories and do not take international or lingua franca English into account, but, as pointed out by Seidlhofer (2011: 5), some of these traditional classifications (for example, Kachru’s circles of English model which is presented below) were never intended to be anything more than purely conceptual and cannot account for the global situation of English today, but are nevertheless useful starting points for analyzing alternative models of English use and the circle model will be presented in more detail here. Other models, Modiano’s (1999: 10) for example, deal with International English as an entity in its own right, but often in doing so lose distinctions of how English is acquired differently in different groups. English as an international language (EIL) is seen by Modiano as a variety which is made up of ‘features of English common to all native and non-native varieties’ (Modiano, 1999: 10), so while it is a useful concept to go beyond purely native speaker models, it does not explain how the shift to ELF took place and what the linguistic implications of this are. Kachru’s (1982: 38ff) system of categorization was developed to account for the different functions of Englishes and the ‘models’ which they follow. In his grouping there are three different circles of English-speaking countries: inner, outer and expanding, which can be roughly overlaid onto the concepts of English as a native langue (ENL), English as a second language (ESL), and English as a foreign language (EFL), respectively: (1) The inner circle refers to the countries where English has traditionally been seen as a native, or first, language; i.e. the United Kingdom, the United States, Ireland, Canada, Australia and New Zealand. These varieties are viewed as ‘norm providing’ (Kachru & Nelson, 1996: 78) as they often serve as the model for non-native forms of English. (2) The outer circle refers to countries where English is spoken as a second language and where English has essentially acquired an institutionalized form (Kachru, 1982). This includes countries such as India, Cameroon, Nigeria and Jamaica, and other English-speaking territories mentioned in Crystal (2003). For Kachru (1982: 38), these varieties are characterized by an extended range of sociolinguistic, stylistic and register uses, evidence of nativization in ‘both formal and in contextual terms’, and a ‘body of nativized [and localized] English literature’. (3) The expanding circle comprises all the countries where English is spoken as a foreign language and where the English used is a ‘performance variety’ (Kachru, 1982: 38). Kachru gives the examples of Iran and Japan, where English is taught at schools but does not have societal or cultural function, noting that ‘the performance varieties of English have a highly restricted functional range in specific contexts; for example,

6

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

those of tourism, commerce, and other international transactions’ (Kachru, 1982: 38). Additionally, these varieties, unlike those of the inner and expanding circle, do not have an ‘institutionalized status’. Instead, ‘[i]dentificational modifiers such as Japanese English or Iranian English are indicative of geographical or national performance characteristics’ (Kachru, 1982: 38). The inner circle varieties are clearly defined in Kachru’s characterization but the boundary between the outer and expanding circles is less transparent because ‘an institutionalized variety always starts as a performance variety, with various characteristics slowly giving it a different status’ (Kachru, 1982: 39). The distinction between institutionalized and performance varieties is tied both to attitudinal and linguistic processes, but also to ‘(a) the length of time in use; (b) the extension in use; (c) the emotional attachment of L2 users with the variety; (d) functional importance; and (e) sociolinguistic status’ (Kachru, 1982: 39). While useful as a conceptual starting point, Kachru’s model cannot be used to make hard and fast distinctions between varieties and speaker types and, more importantly, it is not fully able to deal with an extremely globalized world and with the use of ELF. In today’s society, there are a number of countries where English is regularly used intranationally as a lingua franca, but where it is clearly not (yet) an institutionalized variety. The situation of Switzerland, which will be discussed in Chapter 3, is one such case. The ‘other side’ of English has grown exponentially since Kachru’s first discussion in 1982 and it is clear that the circle which has grown the most is the one he aptly named the expanding circle, although his use of the term expanding does not quite match the situation currently found. For him, the circle was expanding in the sense that English was studied as a foreign language in more and more countries. While certainly true in this sense, the expansion has also involved a marked increase in frequency and the range of purposes for which English is used. In the 21st century, many countries which clearly do not fit into the original outer circle are using English (nonnatively) in ways that go far beyond what would be predicted in expanding circle varieties. They are using ELF, similarly in some ways, to outer circle varieties, but without the historical, cultural and institutional background of English, and as such are different from purely ESL and EFL situations.

1.3 From ESL and EFL to ELF and Beyond The transition of English from a native language, to a second language, to a lingua franca is important, because, as Seidlhofer (2011: 40–41) points out, ‘“English” does not simply transfer intact from one context to another – the “E” in English as a Native Language is bound to be something different

English as a Lingua Franca

7

in kind from the “E” in English as a Lingua Franca’ and it follows from this that the ‘E’ in in ESL is also a different English from the former two, so what exactly ELF is and how it is different from ENL, ESL and EFL needs to be addressed. Recent studies of ELF (Cogo & Dewey, 2012; Jenkins, 2009; Mauranen & Ranta, 2009; Meierkord, 2013; Seidlhofer, 2011) have concentrated on the disconnect between the traditional native versus non-native models, and tried to account for the increasing range of uses of ELF. Their main findings about lingua franca users of English will be examined here to establish what commonalities and differences there are within countries which have adopted or are adopting ELF. I will focus on claims about ELF use in Europe, although many are applicable to other ELF situations. Cogo and Dewey (2012: 11) deem that ELF needs to be defined on three levels: first in terms of its settings (‘contexts in which English is used as the principal contact language’); secondly in relation to its function (‘means of communication among different first language speakers’); and finally as a new research paradigm in sociolinguistics. A clear definition of ELF is crucial, because there are many mixed perceptions about what it is. Although the discussion above implies that what constitutes ELF is relatively easily defined, this is not fully the case. The disagreement about what ELF is does not come from within researchers of ELF themselves, but rather in terms of a gulf between what ELF research is and what researchers of native varieties of English and of second language acquisition (SLA) think it is. Cogo and Dewey (2012: 14) suggest that some of this confusion about ELF has ‘arisen partly as a result of misapprehensions about the nature of lingua francas generally (which are often assumed to be homogeneous when they are in fact characterized by plurality)’ and also because the ‘goals and focus of ELF research’ are sometimes misinterpreted. For example, there is the fact that many ELF speakers do not communicate on a frequent basis with native speakers even when some might be everyday users of English and that native speakers of English in ELF environments will also have to make concessions in terms of their language. English is the lingua franca of the 21st century and as well as concentrating on the reasons why and establishing how much more it is likely to spread, the emphasis needs to be on what this English is like. To do this, it is necessary to examine not solely how it distinguishes itself from native English, but also in what ways it is very similar to it. The first concern in any discussion of lingua franca languages is the rationale behind their use. For English, beyond the factors tied to globalization, the reason is that it often serves to dissolve linguistic and country boundaries and to put speakers on an equal (non-native) footing. Switzerland is a good example. There, as in many other countries, using a language which is foreign and politically neutral makes it possible to circumvent any issues of ‘favouritism’ (providing a ‘level linguistic playing field’ in Phillipson’s (2003: 169) terms).

8

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

This is one of the main characteristics of a lingua franca: by virtue of it being a non-native language for all (or at least most) of its speakers, there is a degree of tolerance and give-and-take in its use. As such, ELF needs to be examined as what Seidlhofer calls an ‘autonomous’ form of English, ‘separate from native-speaker English’ (Seidlhofer, 2011: 171). Seidlhofer (2011: 9) also rightly points out that, although some people see the desire to learn English simply as a wish ‘to communicate with its native speakers (or rather a subset of them whose English corresponds to the approved norms of usage)’, this is increasingly not the case because much ELF is used between different groups of non-native speakers and not with native speakers at all. The contribution and role of native speakers is important, but they do not ‘set the linguistic agenda’ in ELF communication and all users of ELF must ‘make adjustments to [their] local English variety for the benefit of [their] interlocutors’ (Jenkins, 2009: 201). While potentially a participant in lingua franca environments, the native speaker is no more important than the other participants. When making this point, Jenkins (2009: 202) also adds that it is important in ELF research to clarify the distinction ‘between difference (i.e. from ENL) and deficiency (i.e. interlanguage or “learner language”)’. The issue of supposed ‘deficiency’ will be discussed further, but first the ways in which English is acquired as a foreign and specifically lingua franca language need to be examined. In most of the expanding circle of English-speaking countries, regardless of whether English is used as a lingua franca on a national level or not, it is still acquired in a classroom in the first instance. I highlight this because, as I will discuss in the following chapter, many of the factors relevant to SLA also apply to ELF. Additionally, although the models learnt originally are those of native varieties, ‘learners of English as a foreign language assume the role of users of English as a lingua franca. As they move into contexts of use outside the classroom, EFL learners become ELF users’ (Seidlhofer, 2011: 187). That the transition from learner to user is intrinsic to the process is not anodyne and influences the motivations learners have and how they learn English from the start. Furthermore, almost none of the learners of English in Europe today are truly foreign language learners (or users) in a way comparable to foreign language learners in English-speaking countries. As already noted, English has become such a global language and is used so frequently for business, entertainment and in education that the motivations of students learning the language must be taken into account. The global importance of English is precisely the reason why English is being introduced in schools earlier and earlier in European countries, and the motivations behind and the attitudes towards English learning are often very different from those for other foreign languages. Students learning English in Switzerland, for example, know that they will use it almost regardless of what type of job they go on to have and wherever they go. This is not necessarily the case when learning other languages.

English as a Lingua Franca

9

Because ELF is seen to be moving away from native varieties of English, some ELF researchers have questioned whether it is still appropriate for the native model to be the main focus in teaching (Jenkins, 2009; Seidlhofer, 2011). What concerns me here, though, is not the ethics of which models to use, but rather the facts. To showcase how ELF is developing as its own variety (or varieties) alongside, but separate from, native speaker varieties, researchers often focus on the elements that ELF speakers use which are markedly differently from native speakers. A problem with this approach is that the attempt to examine salient differences frequently ignores the many similarities. Even if ELF demonstrates features divergent from native standard forms in some instances, the fact that the original acquisition model is towards a native target cannot be ignored, particularly when examining sociolinguistic competence. As Chapter 2 will demonstrate, the original models provided influence the ultimate form of language spoken. By comparing native models to the ELF forms used, my research will reveal which features ELF makes its own and which it adopts unaltered from native varieties. A large part of the debate on ELF has focused on how it could be taught and whether it is in fact desirable to teach it (rather than a native variety), and on aspects related to who English ‘belongs’ to if there are now more nonnative speakers than native ones. This is not my concern here. I take for granted that the ways of learning and using English have changed in the 21st century, and examine the consequences this may have had on the language as it is used and acquired. The discussion above is not intended to imply that researchers think that English is necessarily learned very differently from other foreign languages in Europe or to dismiss the fact that ELF is clearly often different from native speaker English and there are features that are more likely to occur in nonnative speakers. Rather it aims to underline what the initial learning process is likely to be and that the initial models are native varieties. This is important, particularly as the proficiency of ELF users is varied (Jenkins, 2009: 202); some might be using English while still studying it, but others use English on a regular basis at work and are not language learners in the usual sense. This, of course, presents an additional difficulty in the overall conception of what ELF is. While ELF spans a wide range of levels of proficiency, given my interest in the acquisition of variability, the focus here will be on speakers who are no longer formally studying English, but who are regular lingua franca users. This is not to say these users are no longer learning vocabulary (and to a lesser extent syntax and pronunciation), but rather that they are fairly proficient in terms of the variety of English they are using. The proficiency of regular users means that researchers of ELF are beginning to uncover commonalities in speakers from different backgrounds and, thus, can get an idea of what kind of features might be found in ELF. Scholars working on ELF have underlined that features of ELF speakers which differ from those of native speakers should not be assumed to be

10

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

errors, but ‘may instead be [. . .] legitimate ELF variant[s]’ (Jenkins, 2009: 202). It is too easy to home in on the differences and see them as deficiencies and forget in the process that the differences between ELF and ENL are no greater (and no more incorrect) than those between American and British English. In many ways, ELF is a variety of English like any other, more fluid and variable from speaker to speaker perhaps, but a variety all the same. The people who use ELF tend to use it not solely for perfunctory purposes or for trade and barter, but in similar ways and for similar purposes as they would use any other language which they were fluent in. This is one of the ways English distinguishes itself from earlier lingua francas and shows itself to be truly global. ELF users do basic business in it, but they also converse with colleagues and friends, make jokes, watch films and listen to music (albeit this is often in a native variety of English). It is not just a tool to communicate a few phrases, but a living, changing language. According to Seidlhofer (2011: 35), searching for signs of deficiency in ELF is one of the main conceptual problems for native English speakers when dealing with ELF users because there is an ‘inclination to perceive anything that does not quite meet NS expectations (based on individual intuition) as an “error”’. Native speakers justify this perception in two ways: first of all, the differences ‘can be taken as “barbarisms” or symptoms of “language disease” offensive to a native speaker’s sense of what is “right” and “pure”’, and secondly, in terms of ‘concern that the “errors” they diagnosed may “create false and unintended impressions”, i.e. that the disadvantaged non-native speakers will not succeed in making themselves understood’ (Seidlhofer, 2011: 35). This second concern, as Seidlhofer notes, is largely unnecessary since allowances are made for comprehension in ELF contexts. In fact, the greater comprehensibility of ELF over native English is one of its most interesting characteristics, as emphasized by Phillipson (2003: 167): ‘In many international fora, competent speakers of English as a second language are more comprehensible than native speakers, because they can be better at adjusting their language for people from different cultural and linguistic backgrounds.’ Non-native speakers from different backgrounds using English together are far more likely to try and interpret what others are saying and focus more on communicating rather than worrying about whether what is being spoken is grammatical. The non-native speakers’ different backgrounds mean that ‘participants using English as a lingua franca will need to cope with the result heterogeneity’ (Meierkord, 2013: 3). Their sensitivity to others’ linguistic needs is likely to be far greater than that of native speakers of English, who will not be so used to communicative difficulties. Finally, changes to do with how space and community are perceived and constructed need to be related to the development of ELF. In a globalized world, contact is not always face-to-face, but instead may be over the phone, over the internet, via email or through other forms of digital communication.

English as a Lingua Franca

11

Technological change has created an infinite community of English users. Seidlhofer (2011: 86–87) comments on this, stating that ‘at a time when many of us, and particularly those who are regular users of ELF, tend to spend more time communicating with people via email and Skype than in direct conversations with participants in the same physical space, the old notion of community based purely on local, non-mediated contact among people living in close proximity to each other clearly cannot be upheld anymore’. Speakers of ELF do not need to see each other regularly to be able to interact frequently or regard themselves as belonging to a community. This will be confirmed by the data presented in Chapter 4, where I analyze the use of ELF in the online practices of a student association in Switzerland. The members meet face-to-face only occasionally, but they communicate with one another on a very regular basis and as such are part of a stable, clearly defined community of ELF users. This book does not intend to identify what ELF users ‘get wrong’ but rather offers linguists and ESL/EFL teachers an opportunity to see which aspects are more easily learned and which are not. For linguists, this will make possible a better understanding of the processes of language acquisition in general and the acquisition of variation in particular, whereas for teachers it may underline which features might need more overt teaching (if one still aspires to a native model) and which do not (if they appear to be used similarly to native speakers already). The methods used in my investigation will be those of sociolinguistics primarily, with additional use of some methods of SLA, as will be discussed in Chapters 2 and 4.

1.4 The Sociolinguistics of ELF In her recent book Understanding English as a Lingua Franca, Seidlhofer (2011: 70) points out that sociolinguists have rarely concerned themselves with ELF, and as an example mentions that many volumes and corpora purporting to be interested in international Englishes or world Englishes completely ignore non-native or non-nativized varieties. She adds that ‘oddly enough, English as lingua franca as used across all of Kachru’s circles is excluded in practically all of the fascinating current research and discussion in sociolinguistics and descriptive linguistics in general’. From her perspective, ‘this refusal to take ELF and ELF speakers seriously is all the more perverse since it flies in the face of everything that sociolinguists have held dear all along: interest in the intricate relationship between linguistic variation, contexts of use and expressions of identity, insistence on the intrinsic variability of all language, and the natural virtues of linguistic diversity’ (Seidlhofer, 2011: 70). These claims are not unfounded and Seidlhofer is right to note that few sociolinguists have examined the processes of variation and language change

12

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

within ELF. To redress this situation, the present book is an attempt to apply quantitative sociolinguistic methodology to a corpus of ELF. This will allow me to establish what further insight about ELF and about the processes of language variation more generally can be gained. I will examine features of inherent variability in native English and determine which aspects of it are acquired and maintained by non-native speakers, and which aspects are more likely to be modified. Although the lack of sociolinguistic research on ELF partly ‘lies in a disinclination to question received wisdom’ (Seidlhofer, 2011: 70), it is also due to the lack of SLA research on ELF. Seidlhofer (2011: 11–12) notes that ‘[i] ronically, English is the language most studied in SLA research – but in the guise of ESL/EFL, not ELF’. For her, the lack of SLA research into ELF is a sign ‘that there is blatant mismatch between the purpose for which English is most learnt in the world, namely ELF use, and what is focused on in SLA, namely ESL/EFL’. Research on SLA has not caught up with the spread of English and is still looking at English language acquisition as if it were comparable to the acquisition of other foreign languages. The interface between these two areas is, however, what will reveal most about the patterns of EFL. This is despite the fact that such research has only recently begun to be examined more broadly. The long-perceived mismatch between sociolinguistics and SLA is partly to blame for the paucity of studies on ELF from a sociolinguistic perspective. In this book, I hope to demonstrate there is indeed much to be learned by sociolinguistics, SLA and ELF researchers from the study of sociolinguistic features of ELF. Concretely, what has the discussion above shown in terms of how different the acquisition of English in a lingua franca context is likely to be? The hegemony of the English-speaking music and film industry means that children will have heard songs in English or seen British or American films (dubbed or subtitled) long before they begin to learn English at school; contact with some form(s) of English is the norm for most people. The ubiquity of English will continue throughout the acquisition period and beyond, as it is simply impossible to avoid English even if one chooses not to speak it in many places. Students will hear words and expressions which they will then often attempt to adapt to their own speech, even when these forms are not taught in class. Its use as a lingua franca also means that English learners are more likely than other foreign language learners to use it with other non-native speakers. This final point demonstrates why it is vital for SLA researchers to reflect on the lingua franca side of English: language which is used by groups of non-native speakers together will develop in different ways from a language which has only the native model as a focus. This is tied to processes of language contact and language acquisition, but also to the fact that ‘for non-native ELF speakers, being able to use the language like native speakers and without traces of their L1 is increasingly perceived as unnecessary, unrealistic, and at least by some, as positively undesirable’

English as a Lingua Franca

13

(Seidlhofer, 2011: 50). Findings by Rindal and Piercy (2013) in Norway echo this: a large minority of the adolescents in their research do not want a British or an American accent, but rather aspire to a neutral one. An understanding of speakers’ motivations to learn and use English will provide additional insight into what level of attainment they may reach. The motivations go both ways, of course, in the sense that the widespread use of English will be a positive influence on learners of English, while at the same time meaning that learners feel (rightly) entitled to learn English in a way that is useful to them rather than aiming purely towards a native model. This means that it is possible to examine features of ELF and determine how different (or similar) they may be to native speaker output, but also to establish how much internal consistency there may be (across different groups of ELF speakers). This means researchers can observe ‘the unfolding of familiar processes of language variation in language use, but extended in non-canonical, creative ways’ (Seidlhofer, 2011: 108). This is precisely what this book sets out to examine: to what degree can variation in the features found in ELF be said to be ordered and understandable processes, even in the cases where they might differ from native speakers’ patterns? I will try to answer this this question by studying features that are inherently variable in native varieties of English and which are part of native speakers’ sociolinguistic competence, and by establishing, first of all, whether the non-native speakers match the patterns of native speakers, and secondly, because non-native speakers of English of three different linguistic backgrounds are examined in this study, whether these speaker groups are similar to each other. Cases where the non-native groups are similar, but do not follow native patterns, might demonstrate the presence of a specifically ELF use of a feature. This book, then, seeks to reconcile the often differing approaches taken by sociolinguists and SLA researchers: by quantitatively comparing the patterns and rates of native speakers to ELF non-native speakers, I will be able to establish whether they are really that different. The aim is not to prove either that ELF is completely different from native varieties or to demonstrate that it is not that different at all, but rather to use tools from various linguistic subfields to uncover what ELF can tell us about how language is used as a whole. The methodology and the data used will be discussed in the two following chapters and the results of the analyses are presented in the second part of the volume. An overall discussion of the findings and their implications for further research as well as for our knowledge of ELF, SLA and sociolinguistics more generally will be dealt with in the final chapters.

2

Second Language Variation

2.1 Introduction The claim that sociolinguistics in general, and language variation and change (henceforth LVC) in particular, have not given much thought to the processes and forms of ELF is true for the most part, as discussed in Chapter 1. However, the dearth of research into the sociolinguistic aspects of ELF must be considered in light of the fact that the study of non-native varieties through a fully sociolinguistic perspective is relatively new. Until recently, sociolinguistic researchers primarily examined native rather than non-native varieties and, conversely, many second language researchers did not explicitly set out to examine whether and how non-native speakers acquired sociolinguistic competence, even if attention to speech, formality and other sociolinguistic factors were included in their analyses (Ellis, 2008: 119ff; Gass & Selinker, 2008: 259ff). The oversight, then, lies in both sub-fields and spreads far beyond discussions of ELF, although, obviously, some researchers have focused on aspects at the interface of the two (see Section 2.5 for a summary of the research that has most clearly brought sociolinguistic and SLA insights and methodology together). As this chapter will detail, there are several reasons for the traditional lack of research that is directly on the interface of the two fields, but none is insurmountable and they often relate to conceptual rather than methodological differences. Moreover, many of these reasons are not fully applicable to the situation of ELF or to varieties spoken by highproficiency non-native speakers. Nonetheless, these points can help explain why L2 sociolinguistics has only recently1 been seen as integral to the fields of sociolinguistics and SLA and can provide a starting point for the discussion about how sociolinguistic competence is acquired in ELF contexts. This will allow me to demonstrate why a better understanding of the acquisition of sociolinguistic competence is useful for ELF, sociolinguistics and SLA researchers. As a whole, this book will investigate the type of features that have been often left aside in SLA research, in order to learn what these can reveal about the acquisition of variability and of sociolinguistic competence and how it can be related to ELF and language acquisition research. 14

Second L anguage Var iat ion

15

The importance of sociolinguistic competence in high-level users is obvious: if non-native speakers are to be judged to have fully mastered the target language, they must not only show that they have mastered the syntactic and phonological aspects of the non-native language, but also display that they have acquired the variable rules of native speakers, both for features where the variation is stylistically motivated and those where the variation is internally constrained. This sociolinguistic competence belongs more broadly to native speakers’ communicative competence (Hymes, 1972: 281) and is an intrinsic part of the mastery of a language. An examination of the communicative and sociolinguistic competences of ELF users will permit a deeper insight into what it is possible to learn in lingua franca contexts and establish how learning in this type of situation can be similar to or different from classroom, immersion, immigrant or other acquisition environments in terms of sociolinguistic competence, as well as in other respects. By using the methodology of comparative sociolinguistics (see Chapter 5 for a full discussion of this) and concepts related to second language variation (SLV) I will be able to test how non-native speakers deal with inherent variability and to establish the types of variable constraints that they are able to acquire and the ones they are not.

2.2 Sources of Non-native Variation Interestingly, one of the main reasons for the traditional lack of research straddling LVC and SLA is linked to differences in what researchers mean by the term variation. While for LVC researchers, variation is generally related to speakers’ communicative and sociolinguistic competences, in the sense that speakers may use one form (variant) or another depending on social or internal factors, many SLA researchers focus primarily on the variation in terms of developmental forms versus ‘correct’ ones, that is to say, cases where non-native production is at odds with native speaker production. Features that allowed researchers to understand the effects of cross-linguistic influence and language learning processes made up the bulk of SLA studies that examined variation, while those which focused more specifically on how non-native speakers deal with the underlying patterns of the target language were studied less often (an early exception to this is Adamson & Regan, 1991). However, the emphasis on mistakes obscures the fact that non-native speakers are sensitive to native sociolinguistic competence even while acquiring a language, as a number of recent studies have demonstrated conclusively (Mougeon et al., 2010; Regan, 1996; Regan et al., 2009). These studies focused on cases where non-native speakers have acquired the variability already existent in the target language. This kind of variability is fully noticeable only when it is studied both quantitatively and qualitatively. Since both types exist in non-native output, they must

16

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t Variability in target language (sociolinguistic competence) Non-native variation Variability related to learning (cross-linguistic influence)

Figure 2.1 The two sources of non-native variation

be distinguished before the differences between SLA and LVC can be discussed further. Of the two types of variation found in non-native speakers, one is related to aspects of sociolinguistic competence in the target language, whereas the other is related to processes associated with language learning, either through the cross-linguistic influence of the source language or because of general learning processes (Figure 2.1; Odlin, 2003: 436ff). Variability associated with the target language has been examined in some second language research, but there has been little attempt to categorize the two different types in a completely explicit way. Regan (1996: 178), for example, discusses variation in interlanguage as being on two axes. For her, ‘progress along the vertical axis can be seen as progress along the developmental line in the acquisition of native structures of the language whereas progress along the horizontal axis is seen as the acquisition of sociolinguistic competence or the acquisition of dialect differences’ (Regan, 1996: 178). Howard et al. (2013) and others refer instead to type I and type II variation, with type II variation being related to ‘native-speaker sociolinguistic and sociopragmatic variants’ (Howard et al., 2013: 341). The terminology used in this book is slightly different, as it attempts to provide less opaque labels. Variation related to native speaker sociolinguistic competence is termed target-based variation (TBV), whereas nonnative-like variation is called learning-related variation (LRV).

Learning-related variation (vertical variation) In variation of the learning-related type, non-native speakers produce grammatically inaccurate constructions (Example 1), but also constructions similar to those a native speaker would produce (Examples 2–3). (1) In Geneva, we command our new computer, with a scanner who is also a fax and a photocopieuse (colours and black and white)! (2) For this, I write here the way for helping students who want to stop to smoke. (3) when you find an activity which could replace smoking. (all f, French, email).2 In the examples above, the variation in terms of relative pronoun selection is linked to what is permissible in French. In French, relative pronoun choice is

Second L anguage Var iat ion

17

not governed by whether a subject is animate or inanimate, whereas in English it is. Animate subjects such as students can be followed by who in English, whereas scanner cannot, and which can be used only for inanimate subjects (such as activity or scanner). In French, qui would be used for both animate and inanimate subjects. The lack of distinction in French most likely leads French native speakers to use the form reserved for animate subjects with an inanimate one sometimes (as in Example 1), but it may also be due to language learning processes separate from the source language. It is a case of variation rather than simply fossilization or a mislearnt rule, as there are also cases, as in Examples 2 and 3, where the Standard English structure is used instead. Cases of LRV are, on the whole, straightforward for researchers to discern, as one of the variants will be grammatical to native English speakers, whereas the other one is not.

Target-based variation (horizontal variation) On the other hand, TBV designates variability which derives from the target language as in the examples below: (1) The first people that come up with a german and a french version, please send it out. (2) I am sending it in advance, so that you can prepare yourselves and also the people who will not be in Bern are able to express their thoughts! (both b, Italian, email). In the above examples, the same subject (people) of a relative clause is used with two different relative pronouns by the non-native speaker. This is because Standard English allows for the use of both that and who with an animate subject such as people. Non-native speakers will have been exposed to both variants while learning English and thus have both forms in their own English use repertory. In this instance, the researcher is faced with two variants, both of which could be used by a native speaker of English; the task is then to establish whether the non-native speakers share the rates and patterns of native use. It was important to resolve this terminological polysemy, as a lack of clarity can make it difficult to establish which type of variation is being examined. For example, most of the chapters in Bayley and Preston’s Second Language Acquisition and Linguistic Variation (1996) present research that examines LRV rather than TBV.3 These help us understand the processes of SLA better, simply not those related to sociolinguistic competence, as these can only be uncovered by examining patterns of TBV. One reason for the neglect of TBV in this kind of study is that, while SLA research has demonstrated that interlanguage contains a lot of variability, the general consensus of SLA researchers is that it is often free

18

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

variability, particularly in the initial stages (Ellis, 2008: 136). The assumption that most variation is haphazard or related to L1 and interlanguage factors rather than determined by rules of the target language means that these studies do not examine the type of variation that is neither free nor problematic: variation affected by sociolinguistic competence. As one of the main premises of research into sociolinguistic variation is that variability is constrained by variable rules shared by people in the same speech community, it is unsurprising that the fields as a whole have not shown much synergy when dealing with these concepts. Research into the acquisition of sociolinguistic competence differentiates itself from the type of variation more generally studied in SLA by concentrating on analyzing how nonnative speakers deal with the variability present in native speakers (TBV). This is very different from an emphasis purely on the processes that mean that non-native speakers sometimes produce forms that would not be used by native speakers. Additionally, in some cases, the notion of free variability is taken too far by SLA research and even features which have been shown by variationists to be socially or internally constrained are seen as free variability: I alternate haphazardly between ‘who’ and ‘that’ as subject relative pronouns in non-restrictive relative clauses. (Ellis, 1985: 80) The alternation in the choice of relative pronouns does not involve free variability, as has been demonstrated by a number of researchers (Ball, 1996; Guy & Bayley, 1995; Tagliamonte et al., 2005) and in Chapter 7 here. Consigning sociolinguistic variability to free variation means that features which might enlighten SLA have often been left aside in SLA research. Moreover, the focus on the most salient differences between native and non-natives and the concomitant dismissal of features that appear to be used similarly to native speakers can lead researchers to highlight mistakes when in fact the overall picture is one of similarity. While completely appropriate for speakers with low levels of competence, given the high level of proficiency found in many regular ELF speakers, the emphasis on errors is problematic. For example, Meierkord’s (2004) analysis of non-native speaker sentences underlines the pitfalls of examining solely non-native differences: far from finding that the non-native speakers in her data produced primarily syntactically marked utterances, she found that, overwhelmingly, the utterances were completely regular (88% overall and 95% for the proficient expanding circle speakers; Meierkord, 2004: 119). It is clear then that, while learningrelated features are salient to SLA and ELF researchers alike, quantitative analysis examining both correct and incorrect forms often reveals that the latter are less frequent than initially appears. ELF researchers are also prone to an overemphasis on mistakes at times and have often turned their attention to the differences between ENL and

Second L anguage Var iat ion

19

ELF rather than the similarities. For example, Seidlhofer mentions that who/which confusion is frequently found in ELF speakers and claims that is due to the speakers preferring ‘the clearly marked relative wh-pronouns to that, which is more ambiguous in terms of word class’ (Seidlhofer, 2011: 107). However, as Chapter 7 will demonstrate, the instances of who/ which confusion in the Swiss ELF data only account for 2% of the total number of tokens. Most non-native groups used that at similar rates to native speakers, so that far from being an example of a new lingua franca feature, in Switzerland at least, the non-native speakers have almost fully acquired the internalized rules of relative pronoun selection that native speakers have.

2.3 Type of Non-native Speaker: Interlanguage Versus High-level Competence The emphasis on LRV in SLA is in part due to the types of learners examined: as the main focus of SLA studies is on how people acquire a language that is not their mother tongue (Ellis, 2008; Tarone, 1988), it is not surprising that many researchers choose to focus on the initial stages and processes of acquisition. Research on SLV and on ELF, on the other hand, tends to concentrate instead on proficient users, so this is a key way in which SLA and ELF research differ. The concept of interlanguage, which is seen as an intermediary stage of the L2 and which contains sizeable amounts of crosslinguistic influence of the L1 (Tarone, 1988), is not fully apposite for ELF research, particularly with regard to sociolinguistic competence, because it implies a temporary stage of language quite different from what many ELF users have. While it may be difficult to examine TBV in the L2 of speakers acquiring a language amid high levels of LRV, it is not the case for proficient non-native speakers because much of the initial non-systematic variability will have disappeared. When discussing sociolinguistic competence, a distinction needs to be made between language learners and language users. The latter are speakers whose second/third/lingua franca language use has gone far beyond the initial acquisition stages and thus can no longer be seen as an interlanguage. These speakers may have a number of non-native or possibly idiosyncratic features in their speech, but their language is relatively stable (but not invariable) and fluent. Both the speakers examined in the present study and those examined by other SLV researchers fall into this category of high-level second language speakers. These language users may no longer be formally learning a new language, but their language use can nonetheless inform understanding of general acquisition processes if studied with the appropriate tools.

20

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

2.4 SLV and ELF Despite the differences mentioned above, research at the interface of sociolinguistics and SLA is potentially relevant to both fields and can further our understanding of both fields in ways that are more difficult when using the methodologies of one alone. Given that one of the main tenets of sociolinguistics is that our language is intrinsically influenced by who we are in contact with, it is impossible to discuss what ELF is like linguistically without a full conceptualization of how it is spoken and who it is spoken with. Examining variable patterns means taking a slightly different angle from the majority of ELF researchers: rather than focusing on features that can show ELF to be a variety distinct from native models, features of inherent variation will be studied. These are features which native speakers use variably, yet predictably, and the aim will be to uncover whether the nonnative speakers have been able to acquire them. In this respect the aims here are closer to those of SLA researchers as the primary focus is on a comparison of native and non-native patterns, rather than solely on how ELF is used. A broader understanding of what users of lingua francas are and are not able to acquire easily will nonetheless increase understanding of how English may change as more non-native speakers acquire it and use it as a lingua franca. It may also reveal more about what kind of target ELF speakers aspire to. Research along these lines may also inform the discussion about how ELF should be taught (as discussed in Seidlhofer, 2011): variable features which are found to be acquired easily could be perhaps targeted less in the classroom than those which are more likely to cause native speaker comprehension problems (since the native speakers are known to show gatekeeper tendencies at times). The concluding discussion of this chapter will return to this, but first the main findings of studies examining non-native aspects of sociolinguistic competence will be presented to highlight what is already known and what more can be uncovered by examining ELF.

2.5 Previous Work on SLV Although research that overtly discusses the interface of LVC and SLA is restricted when contrasted to the fields overall, an increasing number of recent studies have attempted to remedy this lacuna and a ‘new strand of research’ has worked on consciously setting itself apart from ‘the bulk of previous research in Second Language Acquisition’ (Mougeon & Rehner, 2001: 398). Moving away from a focus on ‘aspects of the target language where native speakers display invariant language usage’ (Mougeon & Rehner, 2001: 398) and on LRV, this new research has attempted to establish to what extent nonnative speakers acquire TBV.

Second L anguage Var iat ion

21

Studies examining non-native acquisition of sociolinguistic competence can be divided into two broad categories: those with an explicit focus on highlevel non-native speakers’ acquisition of TBV and how it fits within an SLA framework, and those whose primary interest is uncovering how foreign immigrants integrate linguistically into local communities. While both groups are similar in methodology and in their interest in inherent variation, the latter studies tend not to explicitly reflect on the implications of their findings for the process of acquisition. All merit mention here, however, as the overarching findings can contribute to tentative predictions about how a lingua franca context will affect the acquisition of sociolinguistic competence. Although not intended to be an exhaustive list, the studies below provide an idea of the range and type of research examining SLV4 and their main conclusions will then allow me to discuss the implications for ELF. Among studies in the first category, Mougeon, Nadasdi and Rehner examined non-native French speakers in immersion classrooms in Canada (Mougeon & Rehner, 2001; Mougeon et al., 2010; Rehner et al., 2003), while Nagy et al. (2003) and Sankoff et al. (1997) also studied non-native French speakers in Canada. Dewaele, Howard, Lemée and Regan have all looked at the French of students from Ireland and the UK, particularly in terms of study-abroad contexts (Dewaele & Regan, 2002; Regan, 1996; Regan et al., 2009). Beyond French acquisition, Bayley researched the variation patterns of Chinese speakers, either learning Chinese as a heritage language in the US, or learning other languages abroad (Bayley, 1996; Langman & Bayley, 2002), while Adamson and Regan (1991) studied the acquisition of [ŋ]/[n] variation in Asian immigrants in the United States. Studies in the second category comprise research examining non-native English incomers to the United Kingdom and Ireland in recent years. These studies have tended to partly eschew the SLA focus on how language is learned and concentrate instead purely on how community integration is linked to native-like variable patterns. Cheshire, Kerswill and their collaborators have examined what they have called Multicultural London English – an emergent variety of English in London which is used and learned by both native and non-native children (Cheshire et al., 2011). Other research has examined Polish immigration: Meyerhoff and Schleef (2012) have studied schoolchildren in London and Edinburgh; Drummond (2011, 2012) examined recent immigrants to Manchester and what motivational factors influence the acquisition of a local accent; and Nestor et al. (2012) concentrated on the situation in Ireland.5 These few examples underline that a significant proportion of the research conducted into SLV in non-immigration contexts has been concerned with French variability. While findings for non-native variation in French and Chinese can suggest what might be found in non-native speakers of English, they cannot tell us precisely how non-native speakers deal with inherent English variability in an ELF context. Thus far there has been very

22

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

little research into non-native English variability from an acquisition perspective (other than Adamson & Regan, 1991). The present volume aims to redress this by establishing how three different linguistic groups deal with native English sociolinguistic competence and examining the context of lingua franca use.

Acquisition of informal variants With respect to what these studies uncovered, the overarching finding is that sociolinguistic competence is not as easy to acquire as other features of the target language (see also Hopp, 2007; Sorace & Filiaci, 2006; White, 2010, for further examples of ‘interface’ difficulties). Many non-native speakers fail to attain the stylistic range of native speakers and have a tendency to overuse formal variants or not use the more informal ones at all. Almost every study into non-native acquisition of variation revealed that native patterns were difficult, if not impossible, for learners to match. Sometimes time abroad improved the rates and brought the speakers closer to the native underlying constraints, but in other cases the emphasis on the formal variant was a permanent feature of the learners’ speech. The underuse of informal variants can be partly accounted for by the classroom materials used: these variants are almost never presented in the textbooks and are often not used by teachers in class. Researchers found this meant that students had no real input of the informal form, which in turn made them less likely to be able to use it in a native-like way. While classroom contexts are useful to help non-native speakers learn ‘what is categorical in the target language’, this context does not provide students with the ‘knowledge (conscious or unconscious) of what is variable in NS speech’ (Regan et al., 2009: 137). Sociolinguistic competence is seen to be difficult to acquire because ‘whereas there exist numerous dictionaries and reference grammars to support the teaching of lexis and syntax, there are no such reference books to support the teaching of sociolinguistic variation’ (Lyster, 1996: 167). Additionally, because ‘students learn an academic register of the L2 but not its vernacular’ (Lyster, 1996: 167), they often are not even aware of the variation. The lack of reference books which teach sociolinguistic variation is not surprising: for the most part even native speakers are not aware of the variable rules they use every day, but it is precisely this which makes the acquisition of sociolinguistic competence such a valuable aspect to examine to understand language acquisition better. The fact that the application of these variable rules is almost totally subconscious for native speakers means that they are more difficult for non-native speakers to notice and acquire. Classroom materials cannot fully account for the overuse of formal forms, however. Research into SLV has not only been conducted in classroom environments, and results suggest that full native-like sociolinguistic

Second L anguage Var iat ion

23

competence can be difficult to reach even in immersion and year-abroad contexts despite some increase in informal variants.

Context of acquisition Unsurprisingly, SLV studies revealed that a focus on TBV is relevant to understanding more general processes of SLA and that different learning contexts affect the acquisition of sociolinguistic competence in similar ways to other aspects of language learning. Proficiency and ability to reach nativelike patterns of use are related to the contexts in which the target language is acquired. Traditional classroom, immersion classroom, naturalistic and study-abroad contexts provide different opportunities for acquisition and impact on learners’ ability to acquire TBV (Howard et al., 2013: 314). For Regan et al. (2009: 135) ‘the context of acquisition is crucial for the acquisition of this area of language [i.e. sociolinguistic competence] for advanced learners’. Examining results of their own and other research on the acquisition of French variables and comparing four contexts, Regan et al. (2009: 135) uncovered a ‘cline in gains in sociolinguistic competence’, with the least gains in a traditional classroom environment and the most in naturalistic settings. Howard et al. (2013: 346) echo this, summarizing the cline as: Naturalistic context > study abroad > immersion > regular classroom Mougeon et al. (2010: 155) point out that it is not surprising that ‘FL2 learners eventually develop a sociolinguistic repertoire that includes many of the same variants that are used by FL1 speakers’ in situations where they have an opportunity for interaction with FL1 speakers outside a classroom setting. Similarly, the advanced learners Regan et al. (2009: 141) studied ‘restructured their interlanguage so that the factors which affected variation were ever closer to those which constrain NS variation’. The studies examining Polish immigration to the UK and Ireland uncovered a number of cases where nonnative speakers had managed to acquire the native patterns. This underlines how more naturalistic contexts are better for the acquisition of sociolinguistic competence because they offer more opportunities for real contact with native speakers and consequently more input of the informal variants.

Contact and input Input and context are closely related: ‘more informal interactive contact with native speakers is related to gains in nativeness’ (Regan et al., 2009: 139) and additionally, there is ‘robust evidence that the acquisition of variation patterns of the L2 speech community was significantly positively affected by contact with native speakers’ (Regan et al., 2009: 139). This underlines that contact with native speakers as well as the context of acquisition play a

24

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

crucial role in the acquisition of sociolinguistic competence. Howard et al. (2013: 344) underline the ‘highly positive effect’ of native speaker contact: ‘learners who report high levels of contact increasingly approach nativespeaker levels of use’.

Other factors Sociopsychological factors have also been found to affect acquisition: extraversion and level of integration into the community also play a role in the acquisition of sociolinguistic competence. Dewaele (2004) found that extroverts were closer to native norms than introverts in some of his research, and Drummond (2011, 2012) found that the desire to stay in the UK increased rates of local forms to levels closer to native speakers. Finally, feature-related factors affect whether non-native speakers match the patterns of native speakers. These are factors such as frequency of occurrence and status of the variants; however, because they are feature specific they will only be dealt with in the analysis section and the final discussion.

2.6 Where Does ELF Fit into SLV? How can the findings of recent studies examining SLV be applied to an ELF context? By and large, the consensus is that non-native speakers struggle to acquire the native patterns of stylistic variation, particularly in traditional classroom learning situations. However, the results underline that one of the reasons behind this ‘non-acquisition’ has nothing to do with the learners themselves, but rather because the input provided by teachers and by textbooks does not fully represent native patterns. Can lingua franca use compensate for classroom teaching? In what ways is the acquisition process likely to be affected by the manner in which English is used? How will English’s status and use as a lingua franca influence and affect inherent variation patterns? The discussion of previous findings made it amply clear that a large part of the (non-)achievement of the non-native speakers in terms of sociolinguistic competence correlates with the contexts in which they acquire and use the language: the Irish students who spent a year abroad in France were closer to native speaker patterns than the immersion students in Canada who did not have many opportunities to use French outside a classroom environment. Furthermore, Polish immigrants to the UK and Ireland were closer still to native norms. Where does ELF fit with respect to the various learning contexts and the degree to which they help or inhibit native-like patterns? It is rather difficult to determine a priori because, whereas, in some ways, it is more natural than classroom learning and so could be compared to the study-abroad or immigrant contexts, in others the fact that there is very little native speaker input means that ELF speakers may be less likely to acquire native-like patterns.

Second L anguage Var iat ion

25

The use of ELF could be said to be closer to naturalistic or study-abroad contexts than to immersion or traditional classroom contexts primarily because language use is broader than that found in a classroom environment. The situation of ELF is one where the speakers have made the language their own and are able to use it creatively: they are not language learners, but more accurately language users, which may translate into a more native-like use of variation patterns. Moreover, there is the additional possibility of a focused use of patterns, different from native speakers, but similar across groups of speakers with different native languages. The extended opportunities for use which are found in the more naturalistic contexts are also present in ELF, because language use takes place (almost) wholly outside the classroom. ELF speakers will use English more frequently and more regularly than in a classroom environment. At the same time, the context of use cannot be separated from the input and there will be much less native speaker input for ELF speakers than there was in the studies of year-abroad or immigrant English even if the frequency and range of use will be more favourable to the acquisition of sociolinguistic competence than in immersion and traditional classrooms. Additionally, the type of teaching materials used in classroom contexts which were lacking sociolinguistic variation are the same ones ELF speakers would have encountered, so if native-like patterns are found then they may be related to ELF use rather than to actual classroom acquisition. In sum, as one of the features of ELF is the fact that much of the communication takes place away from native speakers, it is possible that lingua franca users will be less likely to acquire sociolinguistic competence than second language learners who spend time abroad or who have contact with native speakers of the target language. However, this must be counterbalanced by the fact that the ELF speakers use English in a wide range of contexts and this may prove to be important in the acquisition of variable patterns even if there is less native speaker input. Before turning to other ways in which input and contact with other speakers can help predict outcomes of language acquisition, two further points must be mentioned with respect to contexts of use. Regan et al. (2009) found that the benefits of the year abroad were maintained even when the native speaker inputs are no longer daily available, which means that if the benefits extend beyond solely native input and are related also to greater opportunities of natural language use then it is worth underlining that ELF speakers maintain a constant stream of input from other ELF speakers. Similarly, Mougeon et al. (2010) noted that the pupils who had opportunities to spend time with Francophones were more likely to have acquired native patterns, which is relevant in the sense that ELF speakers are exposed to English from sources other than the classroom. That the contexts of use are mainly with other non-native speakers may have other consequences on the form ELF takes and the patterns acquired

26

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

may turn out to be very different from native speakers, but similar between source language groups. This is a point not always made by previous ELF researchers but it is nonetheless relevant because the process of tertiary hybridization may be in operation in the elaboration of ELF. When it comes to understanding ELF, its elaboration and its potential patterns, it is vital to remember that not only is ELF used with native speakers and with people who share a mother tongue, but by virtue of it being used as a lingua franca there is a mix of source languages. This potentially mixed input may have consequences on the output. Tertiary hybridization, adapted from the field of biology and first used in a linguistic context by Whinnom (1971), is a process through which new non-native varieties are formed. The notion of linguistic tertiary hybridization attempts primarily to account for the linguistic restructuring found in creoles (with respect to how diffuse forms become a focused and structured grammar), but it can equally be applied to situations of lingua franca contact. For Whinnom (1971: 105–106), a new pidgin or a new variety is ‘unlikely to arise in a stable form’ in the communication of two speakers (or two groups of speakers) when it is the native language of one of the speakers, i.e. ‘a French-“based” pidgin [. . . in] the communication of an English-speaking schoolboy with a French schoolboy’. Rather, it is in cases where the two speakers (or groups of speakers) are using a third language which is native to neither that a new form may arise, i.e. ‘a French-based pidgin could easily arise from the communication of an English schoolboy with a German schoolboy’ (Whinnom, 1971: 106). The essential difference is that ‘in the second [situation], the target-language, French, is removed from consideration’ (Whinnom, 1971: 106). In Whinnom’s view new language varieties can arise only in cases where there are two or more language groups speaking a third language, because in these cases: Neither speaker has any model on which to improve his performance in French, nor any motive to improve it. Their motive can only be improved communication with each other, the inhibitions of the native who fears ridicule from speakers of the target-language are removed, and one can easily see how and why certain essential pidgin features could be produced. (Whinnom, 1971: 106) This has many echoes with ELF situations. As in Whinnom’s example, one of the main aspects of ELF is communicative urgency, as well as a mixing of speakers with different mother tongues and the fact that native speakers are largely absent. His interest in the processes of language learning and language contact in contexts beyond native speakers led Whinnom to reflect on how new forms arise and how these might differ from what would develop in contexts

Second L anguage Var iat ion

27

of simple, monolingual SLA. Continuing his example of a German–English schoolboy French pidgin, he states: though both [schoolboys] would have acquired and would maintain phonemes foreign to their native systems, their mispronunciations of French words, intelligible to each other and reinforced by repetition, might soon render the ‘French’ unintelligible to a native speaker. (Whinnom, 1971: 106) Although such extreme cases as Whinnom’s example are not likely to be found in ELF situations, the idea that new features arise from a mixed input is useful. In the present study, the concept of tertiary hybridization will be particularly relevant if the analysis reveals cases of variable patterns which do not belong to native English but which are found in the output of several non-native groups. This adds a further dimension to the study of sociolinguistic competence in contexts where the non-native groups do not appear to have acquired native competence but nonetheless share similar patterns. If acquiring sociolinguistic competence is part and parcel of being a full member of a speech community, then it may be that the norms within the ELF community are different from those of native speakers of English and that non-native speakers will aim for the former rather than the latter. Additionally, if the patterns are truly different from native models then it may be that over time they are transmitted to other speakers (or maybe recreated by other speakers with similar backgrounds). These points underline how the situation for ELF will be different from much previous research into SLV: across immersion, year-abroad and the various ways of acquiring a new language through immigration, the native variety target was present in some form.6 While initially present for ELF as well, particularly in the Swiss context studied here, actual ELF use takes place after the initial acquisition stages and is primarily with other non-native speakers. Another aspect of ELF use and input is the fact that there will be more rather than less variation than in other studies of SLV: ELF use is almost defined by the variety of its users and there are native speakers, non-native speakers, speakers with high levels of proficiency, speakers with low levels of proficiency. The great variety of forms and the potential lack of a fixed model may affect how and whether speakers acquire sociolinguistic competence. This is not to imply that extreme variability is necessarily a hindrance to a focused, clear variety: on the contrary, in cases of new dialect formation, it is precisely the presence of extreme variation from speaker to speaker that allows the dialect of the next generation to resolve into a new and focused form (see Kerswill, 2002; Trudgill, 2004). The final discussion will deal with this in more detail and establish how features of inherent variability can reveal whether a variety has reached a focused stage. What has this discussion shown about the implications for the acquisition of sociolinguistic competence in a lingua franca context? First of all, it

28

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

has shown that it is rather difficult to make conclusive predictions at this stage, because the situation of use is very different from previous studies. Because ELF speakers are users and not merely learners of English, it is likely that some of the gains found in more naturalistic settings will be replicated even if the bulk of the input is not directly from native speakers. Recall that ELF speakers do not for the most part aspire to a native model and instead belong to a broader community, and because ‘acquiring variation patterns in native speaker speech seems to need contact with the community’ (Regan et al., 2009: 137), the study may uncover similarities among non-native groups but which are different from the patterns of native speakers. The points above were taken into account in the elaboration of the book and the selection of features for analysis: features that either were explicitly taught in the classroom (in some way) or for which it was clear that some degree of variation was present in class were deliberately targeted. Because previous studies found that informal features were less likely to be acquired, it seemed useful to cast the net more widely and also to examine features whose variation was not primarily related to register and style. This study aims to expand on previous research in other ways: Howard et al. (2013: 352) noted that research into the acquisition of sociolinguistic competence needed to examine a wider range of L1s and L2s. The ELF users in this study are native speakers of French, German or Italian to broaden the range of L1 speakers. Additionally, because many earlier studies focused on French variation, the focus on ELF offers further opportunities to examine another L2 as well. This will make it possible to uncover whether or not ‘the acquisition of such variation may be in some ways language-specific’ (Regan et al., 2009: 143).

Notes (1) Both the recent editions of the Cambridge Handbook of Second Language Acquisition (Herschensohn & Young-Scholten, 2013) and of the Oxford Handbook of Sociolinguistics (Bayley et al., 2013) include chapters on the acquisition of variation, which underlines how the subject is gaining more visibility across the two fields. Earlier handbooks (e.g. Doughty & Long, 2003) covered variation but not the acquisition of sociolinguistic competence. (2) All examples from the data are presented in the following way: speaker code, native language and source type. (3) Regan’s paper on the acquisition of ne by Irish university students of French is a notable exception. (4) For a more in-depth overview of research in this strand, see Bayley and Regan (2004) and Howard et al. (2013). (5) Sharma’s research on London Asian speech (Sharma, 2011; Sharma & Sankaran, 2011) could be associated with this category as well; however, because many members of the community studied regularly spoke English before coming to the UK or are English native, it is not directly comparable to the other studies listed. (6) Although for Multicultural London English the proportion of non-native speakers was such that the effect of native speakers was partially minimized.

3

English as a Lingua Franca in a Multilingual Context: Switzerland

3.1 Languages in Switzerland When thinking about countries where English is used as a lingua franca, Switzerland might not be an obvious choice at first glance: it is often cited as a country of many languages and many multilinguals (Edwards, 1995; Grin, 1999; Hoffmann, 2000; Rash, 1998). The actual situation is subtly different, however, and, as will be demonstrated, English has become one, if not the main, language for communication between Swiss people with different native languages. For my purposes, Switzerland’s societal if not individual multilingualism is a boon when examining English as a lingua franca (ELF), as the linguistic practices (and choices) of three language groups, namely French, German and Italian speakers, can be studied. This chapter will provide an overview of the linguistic situation in Switzerland, focusing first on the national languages and their distribution in the country (as well as their intranational use), before turning specifically to English. The situation of English within the country will be presented in terms of its increased use in the second part of the 20th century, but also with respect to the reasons why it has become an intranational lingua franca. I will examine the lingua franca status of English in Switzerland to determine when this status was gained and how pervasive the use of English is. Taking as a starting point the fact that a number of studies had already found that English was used extensively in the country (Cheshire & Moser, 1994; Droeschel, 2011; Dürmüller, 2001, 2002; Rosenberger, 2010; Watts & Murray, 2001), the chapter will analyze census data to obtain a more complete picture of national and ‘immigrant’ language use. Despite being a country with slightly over seven million inhabitants, Switzerland has four national languages: German, French, Italian and 29

30

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Romansh. Of these four, all but Romansh are also official languages1 and are used in the government and for federal administration. The official status of the three languages dictates that they should be given equal representation in the government and that any Swiss citizen should, in theory, be allowed to use any one of these languages for governmental or administrative purposes (Rash, 1998: 30). While the discussion will tend to focus on German, French and Italian as monolithic entities, each of these languages in fact comprises several varieties, some thought to be more Standard than others. German is the most widely spoken language and the results of the most recent census reveal that it was the main language of 4,640,359 people (Swiss Federal Statistics Office [hereafter OFS], 2002).2 However, most Swiss Germanophone speakers have one of the Swiss German dialects as their main language/mother tongue rather than High German (Rash, 1998: 49ff). Swiss German dialects belong to the Alemannic branch of Upper German, meaning they are rather different from the Standard German (also called High German) spoken in Germany (and Austria). The Swiss German situation is one of the canonical examples of a diglossic situation (Ferguson, 1959), in that (a Swiss version of) High German is used for writing and in formal situations, but a Swiss German dialect is used in informal situations (Ferguson, 1959; Rash, 1998: 49ff). Lüdi and Werlen (2005: 36–37), using the results of the census, find that, at home, over 90% of German speakers use a variety of Swiss German and not High German, and that of the remaining speakers, half use a mixture of Swiss and High German. I will return to the implications of the Swiss German diglossia on the use of ELF in Switzerland later in this chapter. Italian, which was reported as the main language of 470,961 people in Switzerland in the 2000 census, is similar to Swiss German with respect to diglossia: 47% of respondents report that they use either just the Italo-Grison dialect (as named in the census; most linguists would call it the Ticinese dialect) or a mix of dialect and Standard Italian at home. The differences between dialect and Standard Italian, while probably as great as those found between Swiss and Standard German, are not viewed in the same way, however, because in Italy, as well, local dialects are still used in the home in a number of regions alongside Standard (or regional) Italian (Moretti, 1999). The Italo-Grison dialects are Lombard dialects and very similar to nearby Northern Italian dialects. The shift from Italo-Grison dialect to a more Standard Italian is relatively new: in the 1990 census 58% of the population reported using the dialect of a mix of dialect and Standard at home (Bianconi, 1995; Ludi & Werlen, 2005: 40). French was the main language of 1,485,056 people in Switzerland in the 2000 census. While there are French patois spoken in Switzerland, the situation is not entirely comparable with that of the German and the Italian speakers: most French speakers speak a dialect of Swiss French, but it is not

English as a Lingua Franca in a Mult ilingual Conte x t : Sw it zerl and

31

as different from French French as Swiss German is from High German. Only a few speakers speak a dialect that is substantially different from French. Most of the Francoprovencal dialects spoken in Switzerland are obsolescent now (Evolène being the main exception: see Elmiger & Matthey, 2006) and only 1% of the French-speaking population of Switzerland reported using patois rather than French in the census (Lüdi & Werlen, 2005: 39). Although the three languages are on an equal footing nationally, the figures above make it clear that their relative distribution within the country is not. German has by far the greatest number of native speakers, with nearly two-thirds of the Swiss population (63.9%) reporting it as their mother tongue at the time of the most recent national census in 2000 (see Figure 3.1).3 French native speakers make up approximately 20% of the population of Switzerland, whereas Italian speakers account for just under 7% (OFS, 2002). Figure 3.1 includes the other languages reported, making up nearly 10% of the whole. The geographical distribution of the languages must be presented to fully understand the linguistic tapestry of Switzerland. Switzerland operates on a principle of linguistic territoriality, so rather than all three (four including Romansch) languages being spoken throughout the country, they are, for the most part, segregated in linguistic regions which are almost completely operational in a single language (Lüdi & Werlen, 2005: 13). Although there are a few cantons (cantons are federal divisions, similar to British counties or American states) which are bi-(tri-)lingual and a few towns are similarly bilingual, this is more the exception than the norm; within a Germanspeaking area one would expect to hear German primarily rather than French or Italian, and French in a French-speaking area. Figure 3.2 provides a map of Switzerland with the linguistic regions clearly marked on it. Romansch 0.5

Italian 6.6 French 19.5

Other languages 9.5

German 63.9

Figure 3.1 Relative size of the national languages of Switzerland (reported mother tongue speakers) (adapted from OFS, 2002)

32

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Figure 3.2 Linguistic regions of Switzerland (Wikipedia Commons)

This means that, on the whole, the inhabitants of Switzerland live in monolingual communities (see Table 3.1 for figures) despite there being a number of cantons encompassing two or more linguistic regions and a few bilingual towns and villages where most of the inhabitants use two languages daily. The towns of Fribourg/Freiburg and Bienne/Biel are two examples. Table 3.1 underlines the extent to which the linguistic divisions are maintained; in the German-speaking area, for example, fewer than 5% of people speak a national language other than German as their main language. As can be seen in Table 3.1, except for the Romansh language area, the percentage of people speaking the language of the linguistic area is always above 80%. Thus, broadly speaking, the main language of individual speakers is the same as that of the canton or region in which they live. The lower percentage found in the Romansh-speaking area (i.e. 69%) can be explained in two ways; firstly, because Romansh is used by such a small fragment of the overall population in Switzerland, and secondly, because the Grisons is the canton which displays the highest number of different linguistic regions. There are Italian-, German- and Romansh-speaking areas in the Grisons and most Romansch speakers are in fact bilingual with German (or Swiss German) (Lüdi & Werlen, 2005: 12). Figure 3.1 reveals that nearly 10% of the population listed a language other than the four national ones as their main language. What are the other languages spoken in Switzerland, then? The 2000 census demonstrated that,

English as a Lingua Franca in a Mult ilingual Conte x t : Sw it zerl and

33

Table 3.1 National and non-national languages by language area, census 2000 a Language area

Resident population

Speaking the language of the linguistic area

Speaking another national language

Speaking a non-national language

Total

In absolute figures 6,208,918 4,519,601

In % 85.2 86.6

In absolute figures 422,553 245,621

In % 5.8 4.7

In absolute In figures % 656,539 9.0 455,913 8.7

1,404,482

81.6

137,507

8.0

178,376

10.4

266,730

83.3

32,287

10.1

21,230

6.6

18,105

68.9

7,138

27.2

1,020

3.9

Switzerland 7,288,010 5,221,135 German language area 1,720,365 French language area 320,247 Italian language area 26,263 Romansh language area

Source: Lüdi and Werlen (2005: 14); OFS (2002). Notes: a Most of the results in this chapter come either from the website of the Swiss statistics office or from figures provided to me by members of a cantonal statistics office, and as such provide a far more reliable and complete picture of the Swiss situation than it would have been otherwise possible to obtain.

although German, French and Italian are the top three languages spoken in Switzerland, Serbo-Croatian,4 Albanian, Portuguese, Spanish, English and Turkish are all used more often as a main language than Romansh, as shown in Table 3.2 (OFS, 20025). I will come back to the case of English as a main language in Switzerland in the discussion below. Table 3.2 Top 10 languages spoken in Switzerland 1 2 3 4 5 6 7 8 9 10 Source: OFS (2002).

German French Italian Serbian, Croatian Albanian Portuguese Spanish English Turkish Romansh

4,639,762 1,484,411 470,961 103,350 94,937 89,527 76,750 73,422 44,523 35,072

34

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

These results are not entirely surprising given that, in 2000, around 21% of the population of Switzerland was not in fact Swiss (OFS, 2007: 15). Most of these non-Swiss are not recent or temporary immigrants, but rather people who have been in the country for over 10 years (54% of non-Swiss respondents). Of course, not all of these non-Swiss people would claim a language other than German, French or Italian as their main language: Lüdi and Werlen (2005: 15) point out that the policy in Switzerland for foreigners has been one of integration and assimilation, so that by the second generation most incomers to Switzerland have shifted to the national language spoken in the area they live in even if they may still use other languages as well. Many of the people reporting a non-Swiss language as their main language would in fact be able to speak one of the national languages as well. Additionally, a sizeable portion of non-Swiss inhabitants come from France, Germany and Italy and consequently would not be as ‘visible’ in terms of foreign languages (OFS, 2007: 26).

3.2 2000 Census Results While the distribution and number of languages spoken as main languages in Switzerland is vital to understand the overall situation, my main focus is on what the speakers of the various national languages do when in contact with one another. Obviously, in a small country there will necessarily be many times when the inhabitants of the different linguistic regions will need to communicate with each other. Good examples of this are national companies which will, sometimes at least, bring members from different regions together. What are the linguistic competences of Swiss people in terms of one another’s national languages and in terms of extraterritorial languages? The results from the 2000 census are again invaluable: in addition to the question asking for the main language (and which only allowed a single response), a further question focused on what other languages people used at work, at school and at home. As I will show below, the census results also underline the degree to which Swiss citizens are for the most part monolingual, and highlight what many journalists, academics and laymen have sensed for years: the growing importance of English as a language in Switzerland, for international, commercial reasons, but also as a tool for intranational communication (Dürmüller, 2002: 116; Watts & Murray, 2001). Questions 8(a) and 8(b) of the 2000 census dealt with language use (see Figure 3.3, which is a translation of the French version; OFS, 2002). Note that the second question allows multiple responses. Just as the first question raised issues of what might be reported as ‘main’ language as opposed to mother tongue, the concept of what ‘usually’ represents in terms of frequency across respondents is difficult to judge in

English as a Lingua Franca in a Mult ilingual Conte x t : Sw it zerl and

35

8. Language For children who do not yet know how to speak, please indicate the mother’s tongue. People who speak Frioulian or Ladino should tick ‘Romansh’ and not ‘Italian’. (a) What language do you think in and do you know best? (only one possible answer) German

French

Italian

Romansh

Another language, if so which one: ………………………………………..

(b) What language(s) do you usually speak? (several answers possible) Students should not indicate the language which they are learning but only those they usually speak at school. Alemanic dialect

German

Roman Patois

French

Ticinese dialect or ItaloGrison

Italian

Romansch

English

Other languages

At school, at work At home, with family

Figure 3.3 Language questions in the 2000 census

the second question.6 Some respondents may have had stricter views about what this meant and consequently the results may be lower than actual language use. Furthermore, the decision of the census to ask students not to indicate the languages they are studying is very interesting. Its relevance is tied to the fact that there has been a shift towards English as the first non-native language taught at school between the end of the 20th century and the start of the 21st in some cantons. Traditionally, children in Switzerland would have learnt one of the other national languages first at school. Which language depended primarily on which canton they lived in: the French cantons tended to teach German and the German ones French, whereas the Italianand Romansh-speaking cantons varied depending on locality. From 2000 onwards, some of the German-speaking cantons decided to bring non-native languages in sooner and chose to introduce English before French. This has been the subject of much debate within the country, but clearly underlines a possible shift towards English. These decisions have received extensive press coverage in the Swiss media, as they were seen by many French speakers to be a slight to their language; in both 2004 and 2005, Le Temps, one of the main French-language Swiss newspapers, devoted a large section to the decisions of the Swiss German cantons (Busslinger, 2004; Cossy, 2005). The shift towards English in school is of course too recent to have influenced the linguistic choices in the census or in the data collected for the present analysis, but it is useful to mention it as it adds to a full understanding of the linguistic situation and how it is evolving (see Droeschel, 2011, for a fuller discussion of the decisions taken in terms of the educational curriculum). The results of the second census question will be considered from two angles: first of all to establish how often people in Switzerland report using

36

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

languages other than their main one, and secondly, to examine which languages are used, particularly with regard to English. The results presented here are a combination of findings published by the OFS (2002) and Lüdi and Werlen (2005), and the raw figures which were provided to me by the Vaud statistics office. A mix of results was used because the official report was not released immediately, and also because the figures need to be examined slightly differently at times for the present study rather than examining them solely using the more general census results available. Figure 3.4 presents the results in terms of (non-)reported monolingualism: by tallying the number of speakers who reported that they never usually used any language other than their native one (for the purposes of this analysis the Standard and dialectal varieties of German, Italian, and French, respectively, were taken together) the percentage of Swiss people who use more than a single language is obtained. The results are presented by cantons in decreasing order of monolingualism. This figure makes it clear that most Swiss people do not regularly use a language other than their own: taken as a whole, almost 60% of respondents use just one language. This does not necessarily entail that they are truly monolingual; the school system ensures that all students have studied one, if not two, languages other than their mother tongue for at least five years. It does mean that these people do not really feel that they use languages other than their native one on a day-to-day basis. It also means that 40% do use other languages regularly, either at home or at work. The cantons which have the highest proportions are for the most part those which also have the highest percentage of foreigners (see Figure 3.5). The Grisons does not have a particularly high percentage of foreigners, but does have high rates of multilingualism, which can instead be explained through Romansh and German bilingual use. Of course, a high percentage of foreigners and higher rates of multilingualism should not be taken merely 90 80 70 60

%

50 40 30 20 10 0

Figure 3.4 Percentage of monolingualism in Switzerland (by canton)

English as a Lingua Franca in a Mult ilingual Conte x t : Sw it zerl and

37

40 35 30

%

25 20 15 10 5 0

Figure 3.5 Percentage of foreigners

to imply that it is solely the foreigners who use more than one language: Zurich, Geneva and Basel all have a number of international companies where multilingualism, or at least the ability to speak a common lingua franca, is a definite advantage. What languages do Swiss speakers use when not using their native language? For the most part, within the family realm, the majority of people are monolingual (Lüdi & Werlen, 2005: 29); for work, however, the situation is very different, and Lüdi and Werlen calculated that in every linguistic region the average number of languages spoken by people working was around or above 1.5. Looking at the languages used in the various regions, it is clear that, although English is not the only non-local language used, it is an important one and one which has increased since 1990. In the German-speaking region, English was given as one of the languages used for work 17.4% of the time in 1990 and 23.4% in 2000 (Lüdi & Werlen, 2005: 47).7 In the French-speaking region it has gone from 13% in 1990 to 17.7% in 2000 (Lüdi & Werlen, 2005: 51) and in the Italian-speaking region it has increased from 7.7% in 1990 to 11% in 2000 (Lüdi & Werlen, 2005: 55). These numbers are substantial and show how important English is within Swiss companies and workplaces. Not all jobs show such a high use of English of course, but all do show some. In the German and French regions, English is the second language used in the workplace after the local language, while in the Italian-speaking region it comes fourth after the local language. Table 3.3 shows the rates of the national languages and English across linguistic regions (based on figures provided in Lüdi & Werlen, 2005: 47, 51, 55). Figure 3.6 presents the breakdown by individual canton (Lüdi & Werlen, 2005: 116).

38

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Table 3.3 Language use in the workplace

German-speaking region French-speaking region Italian-speaking region

German

French

Italian

English

98.0a

19.7 97.9 16.9

11.1 6.8 98.6

23.4 17.7 11.0

15.4 22.0

Notes: aThese are actually the rates for Swiss German as the results demonstrated that a lot of people in the German-speaking area used Swiss German rather than High German in the workplace.

Unsurprisingly, there is a degree of overlap for monolingualism and low English use in the workplace, but Figure 3.6 underlines how frequently English is nonetheless used in Switzerland even in the cantons with lower rates. The highest rates are found in the financial hubs and in the areas with a strong tourism industry (Lüdi & Werlen, 2005: 66). Geneva’s role as one of the main seats of the United Nations can also explain the high rates of English found in the workplace there. Additionally, the disparity between cantons with high rates of English use and those with low use can partly be explained by viewing it as a rural versus urban dichotomy. The picture of language used in schooling is similar, notwithstanding that students (of all levels) were asked to report only on the languages they used and not those they were studying. Overall 14% of answers of languages used in education mention English (Lüdi & Werlen, 2005: 70). This is true particularly, but not exclusively, at the tertiary levels where the rate increases to 40%. Table 3.4 presents the overall rates of English and the national languages across linguistic regions (based again on Lüdi & Werlen, 2005: 75, 77, 79). 40 35 30

%

25 20 15 10 5 0

Figure 3.6 Percentage of English in the workplace

English as a Lingua Franca in a Mult ilingual Conte x t : Sw it zerl and

39

Table 3.4 Language use in schooling (languages used, not studied)

German-speaking region French-speaking region Italian-speaking region

German

French

Italian

English

N/A 9.3 7.3

17.6 N/A 6.6

4.3 2.3 N.A.

16.7 9.2 4.6

While Tables 3.3 and 3.4 make it clear that national languages are used in education and in the workplace outside their local areas, it is the use of English which is particularly relevant here. The high rates for English underline that, although some speakers use each other’s national languages, others must be using English with each other instead. As Lüdi and Werlen (2005: 52) note, in some professions, English is used for internal communication and this is at the expense of the national languages.8 The census results make it clear that the years of learning the respective languages of people from other Swiss regions have not gone to waste, but to understand the full picture of the linguistic situation in Switzerland, English needs to be included.

3.3 The Rise of English and its Causes The results reflect the growing importance of ELF in Switzerland and help demonstrate that English is used by a significant proportion of the population for inter-regional contact. They also establish the extent to which Switzerland, although multilingual as a whole, is a country made up of many monolingual speakers. The rise of English in Switzerland will now be examined to gauge when the increase in English occurred and reflect on the extent to which it can be tied to its use as a lingua franca. It can only be thought to be used as a lingua franca if it can be demonstrated that it is used by French, German and Italian speakers with one another and not merely to communicate with people who do not speak one of the national languages. Although the census results focused on the use of English in the workplace and education, these are not the only places where it is used. The following chapter presents a medical students’ association, which is the case study for the subsequent analyses. It will demonstrate that English is not only used in business, but also by medical students from across the country when communicating with one another. These medical students found it easier to use English to communicate with each other quickly than to use French and German. This is partly because English is also the language of medical science worldwide: many of the important journals and textbooks are written in English, and even within Switzerland many medical research groups use English, although university courses in medicine are in French or German, according to the linguistic region.9 However, the medical students’ and others’ choice to use English in Switzerland is not just tied to concerns

40

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

of the workplace and education and extends to more social and personal interactions of an ELF type, as should be clear from the discussion so far. The importance of English in Switzerland is relatively new, but the use of English in Switzerland stretches back much further: Switzerland was on the tourist circuit for both British and Americans from the Romantic period onwards (e.g. Byron and Shelley spent time in various parts of Switzerland and several towns in the French-speaking area are mentioned by Louisa May Alcott and Henry James). This use was, however, purely tied to tourism and geared towards English speakers, with Dürmüller (2002: 115) noting that ‘before World War II the presence of English could be felt only where the citizens of the Anglo-American countries – then above all Britain – came into contact with Swiss citizens’. This clearly is not a lingua franca use of English. This use of English with non-Swiss people still occurs to some extent today, as Switzerland places a large emphasis on tourism. However, since the end of World War II, particularly since about 1960, the situation regarding the place of English among the languages used in Switzerland has drastically changed . . . People in Switzerland are now often more familiar with English than with the languages of their compatriots. English has definitely changed its status, it has moved from the fringes to the centre, from the status of a foreign language to that of an additional language with LINGUA FRANCA function, i.e. a language that can be used for special purposes and for wider communication. (Dürmüller, 2002: 116, emphasis in original) The rates of English use found in the census underline that it is precisely the latter type of use that is occurring in Switzerland now. If the date of 1960 is taken as the period of shift from English as a foreign language to be used with foreigners (i.e. tourists) towards English being used as a lingua franca, then it is reasonable to assume that there is now a second generation of Swiss speakers using English in this way. This should not be taken to imply that English itself is passed from one generation to the next: parents do not generally speak in English with their children.10 However, the attitudes of one generation may well be transmitted; even if the English spoken in Switzerland is continuously being renewed and each generation is likely to have the same learning difficulties as their parents (i.e. the French speakers will make similar mistakes to their parents as they too will have cross-linguistic influence from French). The youngest generations in Switzerland (those who would have grown up in the period where English is thought to have acquired more lingua franca status and those born from then onwards) will have been exposed to English much more than the older generation and are much more likely to use English. The census results focusing on languages used in education make this clear: as English is taught earlier at schools, it will also be more frequently used when these people go

English as a Lingua Franca in a Mult ilingual Conte x t : Sw it zerl and

41

into the workplace. The results of the national census demonstrated that English is overwhelmingly used in the workplace and education and not at home. The use of English by Swiss speakers is restricted to specific areas and situations where groups with a mix of French, German and Italian speakers need to communicate, and underlines its status as a lingua franca. Thinking of Switzerland’s long-time use of English for tourism, it could be said that it may be a latent reason why English so readily became an intranational lingua franca. Many Swiss would have been accustomed to using English with tourists and it would have been a small (but significant) step to use it with their compatriots. With the worldwide increase in English use, it is unsurprising that Swiss speakers may have felt it more worthwhile to favour English (in learning and in use) over other languages. Having made it clear that English is operating as a potential lingua franca in Switzerland, what remains to discuss is how this came to be the case. What triggered the rise of English in Switzerland and to what extent is this comparable with other ELF countries? A full understanding of the reasons behind its increase may also help better comprehend the potential shift to EFL use in other European countries where English never had colonial language status either. Beyond the worldwide increase of English, which we have touched on in the first chapter and which definitely comes into play for the Swiss situation and considerations tied to the linguistic marketplace value of English for business and industry (Bourdieu, 1991), there are a few causes which are more restricted to the Swiss situation which can explain why a non-national language such as English may have come to be used as a lingua franca in Switzerland, where there are of course a number of other languages which could have been chosen instead. Four main reasons can help us understand the rise of lingua franca English specifically in Switzerland: neutral language choice, ‘economy of expression’, language learning and comprehension, and diglossia. These factors, unlike the more general globalization pressures towards English, have clearly influenced its spread as a lingua franca in Switzerland and it may be that some are restricted to countries with a multilingual background such as Switzerland. The discussion of these reasons will also clarify the Swiss context of use of ELF and possibly help to explain potential outcomes in terms of the acquisition of sociolinguistic competence.

Neutral language choice People in Switzerland seem particularly conscious of the linguistic needs of one another and of speakers of non-official languages, possibly because of Switzerland’s general awareness of linguistic diversity. This translates into the choice of English as an intranational lingua franca over one of the national languages.

42

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

English is in some instances chosen in Switzerland because speakers are conscious that if they were to employ a single Swiss language for communication within a multilingual group, they would be putting the native speakers of that particular language at an advantage over the speakers of the other languages. A number of Swiss English speakers (as in the extracts below) justified their use of English by claiming it was more fair, as choosing German over French or vice versa11 would be giving an unfair advantage to the native speaker of the language selected. The following two extracts come from interviews with two Swiss French speakers collected as part of the overall project. Both are talking about what language(s) they would use with Swiss German speakers: (1) They usually can speak French much more easily than us German but it’s-, somehow it’s unfair to – to let them speak our language and then us- [. . .] Yes we don’t speak theirs and it’s quite unfair in a sense. (2) I never felt very easy when they spoke French because I knew that they had difficulty to speak French and I thought it was unfair for me to ask them to speak French so I would prefer English. Using English puts everyone at an equal and democratic (dis)advantage and, as such, is a more neutral language choice. As it is a foreign language for almost all Swiss people, the three groups are assumed to have an equal facility in using it. The choice of English as a neutral language is also tied to factors of language comprehension as I will discuss below, but the comments suggest that the desire for fairness is firmly entrenched in the Swiss linguistic psyche and is linked to the national linguistic policies which seek to set the three official languages on equal grounds (politically, even if not numerically). This sets the situation of English use in Switzerland apart from many other countries in Europe, where the choice of ELF may be tied to convenience but generally not fairness.

‘Economy of expression’ Concerns to do with choosing a neutral language can also be related to issues of what could be termed economy of expression: although the general policy within the country is to allow the use of several languages, in practice it can be easier to use just a single language. As an example, an extract from a medical student brochure which outlines the courses students in all of Switzerland need to take during their studies is presented below. The writers of the brochure decided to write it in English, rather than in French and German, and explain their decision by citing economy (i.e. it would reduce their workload). The Working Group gladly accepted the suggestion of the Joint Commission of Swiss Medical Schools to prepare the Catalogue in

English as a Lingua Franca in a Mult ilingual Conte x t : Sw it zerl and

43

English, in order to reduce the workload. (Swiss Catalogue of Learning Objectives for Undergraduate Medical Training: Bürgi & Bader, 2001: 9, emphasis mine) This is linked to a general avoidance of having to state things twice. Presenting everything in both French and German demands much more effort on the part of the speaker, or more likely writer, than using only English. A further example comes from the data collected for the present analysis (see Chapter 4 for a detailed discussion of the data): in an email, one member of a mailing list advises other members to use English when writing, noting that because ‘Switzerland is a multilanguage-speaking country’, members should remember that when sending an email they should ‘write it in english, or to send it both in french and german’. If one adds to this the sheer quantity of information provided in English in Switzerland for the benefit of foreigners, it seems clear that many Swiss people’s contact with English is likely to be as significant as their contact with other Swiss languages. If material is presented only in French and in English, a German speaker might look at the English version rather than the French one. While using both languages may be even more neutral than choosing a third one, it is less efficient, so it is clear that the combination of this and the desire for neutrality can go some way in explaining why English has increased as a lingua franca in Switzerland so rapidly. As well as being a neutral language choice, English can be seen to be a default language in Switzerland.

Language learning and comprehension The third factor in the use of English in Switzerland is tied to language learning. Although the choice to teach English as the first foreign language in schools is a recent development, evidence suggests that English has always been used with greater competence than other second languages in Switzerland. Discussions during the interviews conducted as part of the research, as well as anecdotal evidence, lead to the conclusion that many French speakers feel that they left school better equipped to use English than German, despite the fact that they have often studied German for longer. This was echoed in discussions with German speakers; English was easier than French for them. The reasons for these feelings towards French and German are difficult to judge (beyond diglossia, which is detailed below) and could reflect attitudes towards the two languages as much as concrete differences in language teaching. Alongside attitudes, there is also often additional exposure to English through music, television and film. Disentangling the factors tied to attitudes in Switzerland towards French and German as second languages lies outside my scope here. Suffice to say that these attitudes translate into higher rates of English use.

44

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

The effect of a better level of English is also visible in terms of comprehension: the case study which will be used to examine the acquisition of sociolinguistic competence in a lingua franca context provides a very clear example of this (for a full discussion, see Durham, 2003). At a time when members of the association studied were still using their own languages primarily when emailing each other, an issue arose which had to be dealt with immediately. Emails in French and German went back and forth until, finally, one member wrote: ‘it’s about time that things get clear in this meeting story’ and went on to summarize the events as he saw them, in English. The next few emails on this subject were in French and German. Then this member wrote again in English, summarizing the discussion and ending his message by saying ‘Thank you for your answer, so the actual position is: [. . .]’. After using both French and German, the members of the association had to come to terms with the fact that it was difficult for the French speakers to understand all the German messages and vice versa. English proved to be a more readily accessible foreign language for both groups and, as will be shown in Chapter 4, became the lingua franca within the association. Members were able both to write in English and understand English if someone else used it to a far greater extent than other nonnative languages.

Diglossia Finally, as mentioned earlier, a crucial element of our understanding the linguistic situation in Switzerland rests on what varieties of German are spoken within the country. The fact that French and Italian speakers in Switzerland learn High German at school, when it is common knowledge (and further underlined by the results of the 2000 census) that the majority of people in the German-speaking part actually use a Swiss German dialect instead, is not anodyne. This means that, even having studied German, a national language, for up to eight years, Swiss French and Swiss Italian speakers would most likely not be able to understand a conversation between two Swiss German speakers. Although Swiss German is closely related to Standard German, there are enough differences between the two to make it nearly impossible for a French or Italian speaker with only knowledge of High German to understand it. This no doubt ties in with the feeling of French speakers that their English is better than their German. This means that French and Italian speakers often cannot understand their German-speaking compatriots’ language even if they would like to, which leaves them with no option but to request another language. Because English, on the other hand, does not pose these kinds of issues and is taught almost for the same amount of time as German in the other regions, it is unsurprising that it is favoured over German/Swiss German.

English as a Lingua Franca in a Mult ilingual Conte x t : Sw it zerl and

45

3.4 English as a Swiss Lingua Franca These various factors help to explain how English has become one of the languages used by Swiss people to communicate with each other at a national level and why it has become a quintessential lingua franca in Switzerland. The Swiss use English as a common language among groups who do not speak (or want to speak) each other’s languages. Clearly, this use is very different from how it is used by native speakers and helps explain why most Swiss people think that the variety of English they use is fairly different from Standard English (Watts & Murray, 2001). English has moved from being a language used with tourists to one used with other Swiss nationals and, with that, it has gained a clear lingua franca function within the country. In situations where people are dealing with speakers from a range of linguistic backgrounds, it is easiest to choose a single language and English appears to be the most readily understood and accepted language in mixed language groups, the main reason for this being that it is a non-native language for all. Like some other ‘expanding circle’ countries in Kachru’s (1982) model of English, the current shift towards ELF in Switzerland means that it is slowly acquiring some outer circle characteristics. While it clearly does not have many of the hallmarks of outer circle countries yet, it may be that with time and increased use it will, or it is possible that it will coalesce with other nonnative lingua franca varieties of English spoken in Europe and a single panEuropean variety of English will be created. This is a good position from which to examine more carefully what form the English used takes, particularly with reference to how aspects of sociolinguistic competence are dealt with. The Swiss speakers use English primarily with one another and not with native speakers of English although the model of English they are aiming at is still Standard English (as evidenced by the teaching at school as if English were still a foreign language). What consequences will this have on their use of forms that native speakers use variably? Will they follow (or attempt to follow) the native models, will each linguistic group follow their own pattern or will the fact that they are using English with one another mean that they share similar patterns, albeit ones different from native speakers? These questions will be examined through the various analyses and discussions. Also, what consequences will the more specific factors tied to English’s shift towards a lingua franca in Switzerland potentially have on speakers’ sociolinguistic competence? The choice to use English because of a desire for neutrality has aided its increased use and will undoubtedly continue to do so. Beyond the considerations that English is seen as a neutral language choice because it is ‘culturefree’ in ELF discussions, the specific consequences of using a language for this

46

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

reason have not been dealt with in respect of the acquisition of sociolinguistic competence. No doubt, however, sensitivity towards others’ needs could conceivably also translate to a better awareness of sociolinguistic patterns more broadly, but this will have to be verified through analysis. The factor of language comprehension is also important because of its effect on sociolinguistic competence: if English is selected because it is seen as an easier non-native language than French or German, then it must be that the level of English spoken will be fairly high. The lingua franca speakers of English in Switzerland are high-level users and not beginners. The analysis will be able to uncover whether this affects their use of inherent variation patterns. While the linguistic situation in Switzerland has a number of points in common with other ELF countries, the specifics of the shift to English within the country and the inter- and intranational reasons behind it make it a valuable place to examine both ELF in use and how sociolinguistic competence has been acquired by lingua franca speakers and whether this differs from non-native speakers in traditional classroom settings, in immersion classroom settings and in year-abroad settings.

Notes (1) Romansh is the native language of only 0.5% of the Swiss population (OFS, 2002). Although it has recently been given more status, it is not on a par with the other languages in terms of importance. Due to its extreme minority status, Romansh speakers of English were not considered in the analysis. (2) The OFS changed its method of data collection following the 2000 census and no longer includes questions about language in its surveys, making the 2000 census the most recent one available for our discussion. As Chapter 4 will make clear, the data analyzed were obtained within the same time frame so this is not an issue. (3) As will be discussed below, the census asked which language the respondent thought in and spoke the best. This can be interpreted as mother tongue, but may also apply merely to someone’s main language. (4) Serbian and Croatian were grouped together in the census. (5) See also Lüdi and Werlen (2005: 17ff) for information about the distribution of the non-national languages within the country. (6) In French it was given as ‘habituellement’. (7) Because respondents could choose more than one response the totals for this and subsequent figures add up to more than 100%. (8) Because the report is written in French I have chosen to paraphrase it rather than cite directly. (9) In the EU (of which Switzerland is not a member), it has even been said that ‘[t]he trend of using English as the teaching medium at continental universities can be seen in almost all scientific areas. It often applies only to single courses, and occasionally to a whole degree. . . . The trend towards an entire degree being taught in English is most visible in business schools’ (Phillipson, 2003: 77). (10) Although note that 4.4% of responses in the census ticked English as one of the languages used at home (Lüdi & Werlen, 2005: 27). (11) Italian would not generally be chosen as the lingua franca in mixed language groups as only a relatively small proportion of people outside Ticino learn Italian at school.

4

Data

4.1 Introduction Examples of English in Switzerland abound (see, for example, Cheshire and Moser’s 1994 study on advertisements in English), but in order to examine sociolinguistic competence, an internally consistent corpus of data is necessary, so that any results can be said with certainty to apply to the whole linguistic group and not just to a particular speaker. Although a large proportion of Swiss speakers use English, the subsequent analysis will examine the linguistic practices of a single group as a test case. Data taken from speakers ‘on the fly’ raise a number of issues for second language data, as interspeaker differences can be substantial if the speakers’ English levels are diverse. Choosing a group where everyone can be expected to have at least the same amount of English language instruction minimizes this issue. The data used in this book come from a medical student association, which means that all the members can be expected to have a roughly equivalent level of English. The group was selected in part because of this and because of their willingness to participate in the study, but also due to the fact that they provided a very good example of a self-imposed decision to use ELF. In addition, the group offers a good mix of speakers of the three main Swiss languages. Although examining the output of a single group means that a full picture of English use in Switzerland cannot be presented, the linguistic situation of one particular group of Swiss English speakers can be accurately shown. An understanding of the processes involved in the use of English will be gained by examining the languages used by this group of speakers and the circumstances in which they are used. These findings can then be used to infer whether similar strategies could have been adopted in other Swiss situations. The bulk of the data from this association is made up of emails: this is the primary method of intranational communication within the association. This underlines the importance of not merely thinking about why non-native speakers might choose to use ELF, but also establishing what mediums they use. If, as is the case here, the need for a lingua franca arises 47

48

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

primarily in computer-mediated communication, then it is precisely email data that need to be examined. This is true even taking into account the fact that emails tend to be more similar to written data than oral data. If one considers merely the importance of the internet in spreading the use of English and as a primary cause behind its current lingua franca status, which it no doubt is, then the case for using such computer-mediated data is obvious. Another consideration in the decision to use primarily email data is tied to the observer’s paradox (Labov, 1972): people monitor their language more when they are being observed and this effect is likely to be even stronger for non-native speakers for whom concerns of correctness will come into play (Eisenstein, 1989; Tarone, 1988). The email data are more natural, particularly in the sense that this is how the Swiss students examined here use English. As Section 4.3 will make clear, a large portion of the data was sent before the association was contacted for participation and consequently was not subject to the observer’s paradox. The English used in the emails is an exact reflection of the members’ natural use. As well as introducing the medical association, this chapter also presents a corpus of native English data from a roughly comparable association in the United Kingdom which will allow for a better comparison with the nonnative data in the analysis chapters.

4.2 Swiss Data – IFMSA-Switzerland The International Federation of Medical Students’ Associations – Switzerland (hereafter IFMSA-CH) is an association of students who are studying at the various medical schools in Switzerland at the universities of Lausanne, Geneva, Berne, Zurich and Basel (IFMSA, 2003; IFMSA-CH, 2003).1 The purpose of the association, in the words of one of its members, ‘is to enable international cooperation in professional training and the achievement of humanitarian ideals’ (b, Italian, email). The members of the association help other Swiss medical students find placements in hospitals abroad and organize the Swiss internships (or clerkships as they are called within the association) of students from abroad. In order to organize these clerkships, the members communicate with each other and with members from related associations2 in other countries, generally via email, but also by attending general assemblies in Switzerland (about twice a year) and abroad (once a year). Many of the members of the association have a subsidiary role, contributing infrequently to the mailing list and rarely, if ever, attending the national and international meetings. Although I will be focusing primarily on the members who had a more central role within the association, and who, consequently, contributed to the mailing list more frequently, these ‘quieter’

Dat a

49

Table 4.1 Distribution of IFMSA-CH speakers by gender and mother tongue

Female Male Total % of speakers

German

French

Italian

5 6 11 37

8 7 15 50

2 2 4 13

members are nevertheless important, as the emails are addressed to the whole association and so linguistic choices and uses had to be appropriate for the entire group. Members come from a variety of linguistic backgrounds: French, German and Italian speakers are all represented in the association. The association had about 80 members at any one time over the period when the data were collected (changing as students graduated and new fledgling doctors took their place), but the results presented in the next chapter are only from the members who emailed enough in English to make it worthwhile to include them in the corpus.3 The data are then made up of the emails of 30 members who contributed most frequently to the list. Taking their native language into account, the active members can be broken down as shown in Table 4.1. Table 4.1 makes it clear that the distribution is not completely balanced in terms of members’ native language: there are fewer Italian members studied than French or German ones. This is similar to the actual distribution of languages in Switzerland, although here we have more French than German members. Crucially for my purposes, however, the distribution in terms of actual data is much more balanced (as will be discussed in Section 4.5) as a number of the Italian members held the main roles within the association at various times. The participation of members is not completely balanced within the language groups either, but it is sufficiently so to make it possible to assess that the use of English within the groups is largely similar.4 Previous studies (see, for example, Guy, 1980) have found that variability seen at a community level tends to apply to the individual as well; this was largely the case of the features studied here. The frequency at which members emailed was in part tied to their role within the association; the core members were more involved in the administration of the association and consequently emailed more frequently than other members.

4.3 Languages Used in the Association The primary focus of this book is the members’ English use, but it is useful to briefly examine their overall linguistic choices before turning to a more detailed analysis of English. This is particularly relevant because the

50

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

diverse linguistic backgrounds of the members of this Swiss association meant that the choice of which language or languages to use when communicating with each other is one that had to be made fairly early on in the email communications. Before examining the effect ELF use may have on the acquisition of sociolinguistic competence within the association, I need to demonstrate that English is indeed being used as a lingua franca. The universities which the members attend are either German speaking or French speaking, as the Ticino (TI), the one Italian-speaking canton, did not have a medical school. For the most part, the members chose to go to a medical school in a university in which courses were taught in their native language; thus the French speakers went to the French-speaking universities (Lausanne and Geneva) and the German speakers chose the Germanspeaking ones (Berne, Zurich and Basel). There were Italian speakers at both French- and German-speaking universities. Note that regardless of whether the students were at a French- or a German-speaking university, the amount of English used was the same. On a day-to-day basis within the local exchange offices (LEOs), the members used their university language; it is only when members communicated with members from other universities that the question of lingua franca came into play. This need to choose an intranational lingua franca differentiated IFMSA-CH from other IFMSA member countries. In many of these countries, one language would be used for all country internal emails and information, and another language, generally English, for country external information (anything having to do with the association as a whole, as English is the official language of the worldwide association). Swiss members, on the other hand, needed to select which of two (or more) languages to use for internal communication. Despite the fact that all the students spoke either French or German as their university language, at the time of data collection, English was the main language used by members of the association. This was the case both for face-to-face interaction at the twice-annual national meetings and for written communication, i.e. in the emails which were sent to a general mailing list for the whole association. Although some of the communication between the group was in French or German (and rarely Italian), the bulk was in English. This was a decision which the association had gradually come to once more members began using it and contributing to it. Although a lengthy presentation is not fully germane to the discussion here, Durham (2003) examined the linguistic choices on the mailing list in detail and found that the two main reasons for the selection of English as the primary language within the association had to do with (a) a feeling that it would be beneficial as practice for the international meetings and (b) the fact that it was a neutral language choice and that members of different linguistic groups were more comfortable using it than another national, but non-native language. Section 4.4 presents the situation at the meetings attended as part of the project, while

Dat a

51

Section 4.5 introduces the mailing list in more depth and details the amount of data used for the analysis. The general feeling in Switzerland about the position of English visà-vis the Swiss national languages, as discussed in Chapter 3, must have played a part in the language choice as well. IFMSA-CH’s choice to use English as the main language of intranational communication is in no way unusual for the country. The language choices of the IFMSA-CH students are consistent with the general trend of seeing English as a necessary tool to further oneself in society, and the members would have been aware of the importance of knowing English for their studies. Note, however, that medical instruction in Switzerland is not in English. Some students can, and indeed do, manage without speaking English at all. Students in Switzerland attend medical school as undergraduates, whereas it is primarily in postgraduate studies that English is used in Swiss universities (Dürmüller, 2001).

4.4 The Situation at Meetings (and Interviews) In order to get a better feel for the linguistic uses of the IFMSA-CH members, in addition to studying their emails, data were collected at one of the national meetings and through interviews (in English) of two of the French members of the association. These interviews provide us with more oral data from the French speakers, as there was less from them during the meeting, and it also made it possible to overtly discuss these members’ thoughts on the use of English in Switzerland and within the association. Both the meeting and interviews provide oral data which can then be compared to the emails, making it possible, in some instances, to establish the extent to which the findings of the emails are applicable more broadly. While language use at the meeting was not influenced by my presence, the fact that the interviews were conducted in English undoubtedly was. However, because I had previously met the people interviewed at the national meeting, I was able to establish that there was not a noticeable difference in the level of formality of these speakers’ English between the meeting and the interviews. The meeting, in May 2002, was conducted primarily in English, and when German was spoken, there were requests from the French members to change into another language, underlying the extent to which English is used as the lingua franca within the association. There were no requests to change from French when it was used; however, conversations afterwards suggest that French was not always readily understood by some of the German speakers. The main speakers at the meeting (who, for the most part, were the same as the main emailers in the study), were quite comfortable in switching from one language to another. In fact, some speakers used quite a

52

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

lot of code-switching, with sentences starting in German, switching into French, and ending in English, as in Example 1 below. (1) Also ich habe die das Spiel organisiert und der, et et pour ça il y a eu quelques problèmes because I have now three people to present to you. (b, Italian, meeting)5 This is especially interesting, given that very little code-switching was found in the emails (as will be discussed below). The meeting was held in Basel, which was the de facto headquarters of IFMSA-CH, and was very well represented in terms of languages. About 50 people attended, roughly half of them from the French-speaking universities and half from the Germanspeaking ones. The meeting underlines that, within the association, English is used in circumstances when global understanding is sought. In smaller groups or, as discussed below, in emails to specific individuals, English is not as important, as it is easier for one speaker to ask for further explanation or clarification. These findings support the hypothesis that English is used as a lingua franca in Switzerland: it is used in cases when the speakers could not otherwise make themselves understood by speaking their own languages.

4.5 The Mailing List The email data used for analysis come from two periods of data collection. In the first period, which went from 1999 to June 2002, every single email was collected and tagged for language (English, French, German or other) in order to establish what the main language used was (English was found to be used 75% of the time, 660 messages out of 882). The second period from June 2002 to the beginning of 2003 only collected emails sent in English for a total of a further 144 messages. For the present purposes, only the English emails from both periods are examined and the French and German emails are left aside (again, see Durham, 2003, for a full analysis of the linguistic choices of the IFMSA-CH members). From 1999 to 2003, the mailing list was IFMSA-CH’s main method of intranational communication. The association began using a Yahoo mailing list in May 1999 to ensure that every member received information as quickly as possible and to avoid having to send messages individually. Members sent messages from their own account to the mailing list address; these messages were then automatically forwarded to all other registered members.6 In the course of 2003, the mailing list was transferred to another mail server and members could easily select which members to send their messages to, which meant that there were fewer association-wide messages and data collection was stopped.

Dat a

53

More messages were exchanged at times when events needed to be organized (transport and registration for conferences) than when each LEO was working individually, but on the whole the mailing list was used frequently. Although the messages were of varying lengths, ranging from just a few words to many pages of reports from various international meetings, most were 200 words or longer and thus provide enough data to analyze speaker patterns accurately. Similarly, the frequency of an individual’s emails does not necessarily correlate with the amount of data, as some speakers emailed infrequently, but at length when they did. This is why it is often possible to form a clear picture of the English of members who only sent a handful of emails. The following email is an example of the language production of the Swiss speakers. (2) Hi guys, you remember at the past meeting in Bern a red map that ‘b’ gave to ‘c’, concerning the activities of SCOPH and the presentation of their activity? Well, ‘c’ [. . .] has lost it somewhere. Has anybody of the participants of the meeting taken it home? Please, answer quickly! I need that map desperately! A lot of THANKS (c, Italian, email) Features which mark the email as being written by a non-native speaker of English have been underlined (‘map’ is most likely a calque from the German ‘Mappe’ meaning folder). The main point is that the emails are relatively informal and that members feel free to use their ‘own’ variety of English in them, without being unduly concerned with its accuracy as long as the message is clear. The usefulness of the mailing list in investigating language choice derives from the fact that each individual member is relatively free to decide what language to use, as can be seen in the following comment: (3) I repeat: if you don’t feel writing in English, but you feel like sharing something with us, please write in your language, . . . it is better that not writing anything! (b, Italian, email) The English emails in the corpus and the data from the meeting and interviews as a whole come out to slightly under 100,000 words (and over 800 emails) and are distributed in the following way across the three language groups: roughly 39,000 words from Italian members, 35,000 words for the French members and 25,000 words for the German members. The Italian members contribute a substantial part of the corpus: this is unsurprising in light of the fact that some of the Italian members held the main positions in the association over most of the data collection period.

54

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

This meant that they emailed more frequently than other groups, but there is more to it than that. The Italian members, more than the other groups, used English over French and German in their messages. More than the French and German members, they would have been particularly sensitive to the need to ensure wider understanding: whether messages were in French or German, they would not have been in the Italian speakers’ native language. This does not imply that the Italian members could not cope with other languages: of the three groups in the association, it is the Italian speakers who appear to be the most multilingual. This no doubt because they need to be linguistically competent enough to attend medical school in one of the other national languages. Before examining the English used by the members in more detail, it is vital to address whether the email use genuinely represents an instance of English being used as a lingua franca and, if so, whether the members of the three different linguistic backgrounds are all participating in this use. In terms of the first question, the answer is very clearly yes. The patterns and frequency of use of English underline this. The members are consciously choosing to use English and overtly mention this in their metalinguistic commentary in the emails. They have decided that English is the best language to ensure association-wide understanding and practise this both in their online communication and at the national meetings. Moreover, this is a choice that they came to over time and jointly: in the very incipient stages of the mailing list members tended to use their university language. The gradual shift towards English came when it became obvious that it offered the most in terms of ease of understanding and of use. Another reason why it is likely to be beneficial to use a non-native language is that this will put all the speakers at an equal disadvantage; if the French speakers and the German speakers both used French, for example, then the French speakers would be more at ease, of course. It is likely that the mode of communication, the mailing list, also influenced the choice of language. As long as communication (face-to-face or otherwise) is mainly directed to specific people, there is no need to worry about widespread understanding. However, when the aim is to communicate to a broader, multilingual audience, as the internet makes it easy to do, neither French nor German is able to serve as the main language in the Swiss context, and it becomes necessary to use English. Although two or more languages can be used for face-to-face interaction (as was demonstrated in some of the interactions at the meeting), in the case of email, the use of more than one language was impractical and confusing, and precipitated the eventual choice of a single lingua franca, namely English. One difference between the students’ choice to use English and that of Swiss companies which use English as their official language is that the members of the association chose it for primarily practical reasons rather than economic ones. The students’ use of English is presumably purely

Dat a

55

Table 4.2 Proportion of language use by writer native language in emails 1999–2002 Native language of writer

% in English

% in French

% in German

Total number of emails

French German Italian IFMSA total

60 65 89 75

37 4 7 15

3 31 4 10

262 209 411 882

determined by the fact that it is the most accessible language. Moreover, English is the language of medical science around the world, and should the students decide to pursue medical research then English would be a necessary language for them. From the comments of the students on the mailing list as well as those who were interviewed, it seems that the members did not use English outside the confines of IFMSA very often: in fact, the mailing list was seen as a place where members could practise English. Even if their English use was somewhat restricted, in that it was primarily used for emailing but not often for conversations, the members were quite likely to be in contact with English on a regular basis through the mailing list. There was roughly an email a day sent out to the members. This means that even if the members did not email themselves, they read messages in English every day and thus would have had more contact with it than with a language (French or German depending on the region they are from) which they only used occasionally. The frequent opportunities for English use are important in terms of the acquisition of sociolinguistic competence: the findings discussed in Chapter 2 underlined that with increased input speakers became closer to native patterns. The answer to the second question, namely are members of all backgrounds using English, is again yes. An analysis of the data shows that all three groups use English and that, when taken alongside their use of other languages, it is used at roughly similar rates by all. Table 4.2, which is adapted from Durham (2007), shows the proportion of language use in the three language groups in the first period of data collection. It is clear that the German and French members make use of their native languages on occasion, but English is the first language used for emailing throughout.

4.6 Type of English Used In all the emails analyzed in this study, the main concern of the students is to get their point across, not to write in perfect English. There is a communicative urgency, rather than a desire to create flawless sentences, and members are quite willing to admit that their English can be imperfect.

56

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

When they do not know the exact word in English, they will often replace it with the corresponding one from their own language. Idiomatic forms are translated literally, sometimes accompanied by the phrase, ‘as we say in French/German’. The students are aware that their English is different from the Standard English they hear on television or the radio or in movies, but this does not trouble them as it serves its purpose. In fact, when the association was first approached to get permission to use the mailing list for study, quite a few of the members were keen to see what their own special brand of English was like. The initial contact with the association also spawned a number of comments by the students on their use of English. In terms of cross-linguistic communication, it may be advantageous for the addressees that writers use a language that is not their native one as it ensures a greater chance of comprehension. To put it simply, a French speaker will be more likely to understand a German speaker’s basic English than the same speaker’s native German (if only because the German speaker lacks the knowledge to make use of more complicated forms in English). In the interviews, as well as in a number of the emails, the members commented that they were aware that their English might not be as rich as a native speaker’s but that it was still suitable for communication. (4) I thinker that the English er which, no – which is used all around the world just to communicateer for theer basis of communication it’s a very poor English, I mean, I really don’t have a lot of vocabulary in English [. . .] it’s incredible, that’s unbelievable in English is that we can speak a lot but we use always the same words. (w, French, interview)

4.7 Native English Data To ensure maximum comparability with the non-native data (which come primarily from emails), a similar corpus of native English emails was collected. This means non-native speakers’ variation patterns will be able to be examined alongside those of native speakers. The purpose of this second corpus is to serve as a control group to the IFMSA data, allowing me to compare the results for the non-native email data with native email data. Any results from previous studies looking at native speech cannot be compared to the non-native results without first establishing that native speakers do not change their distribution of the forms in another register. For the features where native studies already exist, the native email corpora will act as a bridge between the native oral and non-native email data, verifying that any differences encountered are due to genuine differences between the native and non-native speech patterns and not to register differences.

Dat a

57

Because it was not possible to obtain emails from IFMSA-CH’s sister association in the UK for comparison, the mailing list of a sports society at a British university was used instead.7 Although not strictly exactly identical, the native mailing list selected contained emails of a very similar nature to the IFMSA-CH ones. Because the university association chosen was one that I had been directly involved in, I knew that, as in the case of the IFMSA-CH mailing list, emails were sent to organize events and provide information for the club’s members. The style of the emails was, as on the IFMSA-CH list, informal as members knew each other well. The corpus itself is relatively smaller than the Swiss corpus as a whole (around 40,000 words in total) and the number of emailers is more restricted as well, but the fact that data come from a single language group means that this is not an issue.

4.8 Discussion Having shown that English is used as a lingua franca on this mailing list and that there are sufficient data to examine the use of the French, German and Italian speakers, the next section will focus specifically on how the nonnative speakers deal with the variability which is inherent in English and establish to what extent they have sociolinguistic competence comparable to native speakers. This will be achieved by examining a set of features known to vary in native speakers of English and establishing what the distribution is for the non-native speakers. In cases where the non-native speakers do not appear to have acquired the native variable rules, the next question will be to determine whether the three groups are similar and if so whether this may signal a specifically lingua franca type use rather than acquisition issues tied to the source languages.

Notes (1) In 2006 the association merged with other medical student groups and subsequently changed its name to SWIMSA. The information presented here relates to the association as it was at the time of data collection, so between 1999 and 2002. (2) The worldwide federation, IFMSA, celebrated its 50th year of existence in 2001 and has members from all parts of the world. (3) The amount of data provided by members was very variable, but to be included in the study members had to have emailed at least five times in English. This does mean, of course, that there is not data for all the speakers in all the features. (4) For all features, the rates of individual speakers were considered to ensure there were no differences in speakers of the same language, but these will not be presented in the analyses. Because of the uneven distribution of the data, it would have been confusing to present speakers individually for each feature. (5) Gloss: [German] I organized the game and [French] because of that there were a few issues. . .

58

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

(6) The format of messages varied depending on the sender. Some members were able to make use of standard word document features (bold, italics, colours) in their emails. (7) The emails were sent by both members of the association in York and members throughout the UK.

5

Methodology

5.1 Comparing Linguistic Varieties There are a number of ways to examine whether the features in the lingua franca used by non-native speakers of English in Switzerland match the sociolinguistic competence of native English speakers. I will discuss why the methodologies of comparative sociolinguistics offer the best option and will provide an explanation as to how the multivariate analyses in the subsequent chapters can be interpreted. I will also briefly introduce the features studied and elaborate on the rationale behind their choice. The most obvious way establish whether the non-native speakers have acquired sociolinguistic competence is simply to compare features used by native speakers and by the non-native groups and see whether they pattern in the same way. Although some features, such as final stop deletion or conjugation regularization, are thought to be vernacular universals of English (Chambers, 1995: 242–243) and have been found to have developed independently in a number of different English dialects, many features exist in different dialects only through direct contact or transmission. The past participle form of got found in some varieties of American English (i.e. gotten) has its origin in earlier varieties of British English and is still used in relic areas of the British Isles (in Buckie in Scotland, for example; Smith, 2000: 175). In a similar way, comparing the English spoken by the three Swiss groups to one another and to native English can help uncover whether these groups exhibit the sociolinguistic competence of native speakers and, if not, whether they are similar to each other (which would suggest that a Swiss lingua franca variety might have emerged). However, surface comparison of linguistic features does not always provide sufficient evidence for postulating the existence of a single language variety, as similarity on the surface may disguise underlying differences. This is especially true for features which are variable: although two language varieties may appear to share a common feature, only an in-depth analysis of 59

60

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

how the feature patterns in the two varieties can reveal whether it is used in the same way; this is why variationist sociolinguists ‘seek to discover usage patterns in the relative frequency of occurrence or co-occurrence of structures, rather than simply in the existence or grammaticality of those structures’ (Poplack & Tagliamonte, 2001: 7). Variation theory is able to examine the underlying patterns of a feature and it is precisely these patterns that can help us uncover features shared across varieties, particularly when using the methodology of comparative sociolinguistics. Comparative sociolinguistics has been used to determine historical relationships between varieties of the same language and it can be used to compare features across varieties to determine if they share the same underlying patterns and constraints even in situations when the surface patterns appear to be very different. ‘Differences in overall rate of variant occurrence may be due to any number of (non-linguistic) factors’ (Poplack & Tagliamonte, 2001: 92), but crucially ‘the conditioning of variability (i.e. the configuration of factors affecting the occurrence of the variant forms), as well as the direction of their effects, are deeper constraints, remaining constant regardless of the extra-linguistic circumstances’ (Poplack & Tagliamonte, 2001: 92). If two varieties present the same conditioning and direction of effects when examined using statistical analysis in the form of a multivariate analysis, then it is highly likely they come from the same source variety. The methodology of comparative sociolinguistics has been used most impressively by researchers attempting to uncover the sources of African American vernacular English (Poplack & Tagliamonte, 2001), but it has also been used to examine generational change (Tagliamonte, 2002) and to compare British dialects (Tagliamonte & Smith, 2005). By comparing the results of separate statistical analyses (Section 5.2 will discuss how the results are reached and how they should be interpreted), the methodology enables researchers to pinpoint shared patterns. As Poplack and Tagliamonte see it: The operating principle of the comparative method is as follows: if two or more languages share a feature which is unlikely to have developed by accident, borrowing, or as the results of universals, then that feature is assumed to have arisen only once and to have been transmitted from a common ancestral source. The goal is to recover as much as possible of that source from a comparison of the descendants, as well as to determine the changes that have taken place in each. (Poplack & Tagliamonte, 2001: 96) If multivariate analysis can prove that a feature has similar variable rules in two varieties, then this feature is judged to have come from a common source variety. This presupposes that similarity in variable patterns cannot be due

Methodology

61

merely to chance: although surface similarity can be misleading, shared underlying factors and hierarchies are thought only to occur in related varieties. For my purposes, this means that if any or all of the non-native groups share the patterns of the native group, then it is possible to conclude that the variation has been acquired.

5.2 Multivariate Analysis and its Use for Comparative Sociolinguistics From Labov (1966) onwards, variationists realized that language is influenced by multidimensional factors and needed to find a way to model this. In many cases, studying factors individually does not allow the linguist to grasp the overall picture of variation, as factors are often impossible to disambiguate through normal distributional means. Multivariate analysis, or multiple regression, allows researchers to examine a number of different factor groups together and determine which ones are significant and are influencing the selection of the variants when all are considered simultaneously (Bayley, 2002: 124; Tagliamonte, 2012: 126; Young & Bayley, 1996: 253). Although multiple regression is a common statistical technique, ‘the programs known as VARBRUL1 have been used most extensively in sociolinguistics because they have been deliberately designed to handle the kind of data obtained in studies of variation’ (Bayley, 2002: 124). Multivariate analysis allows researchers to obtain the variable rules of a feature; these are a ‘mathematical construct mirroring the systematic choice mechanism of language’ (Tagliamonte, 2006: 266). Over the past few years, many researchers have begun using mixed model effects in their analyses (Johnson, 2009) in addition to multiple regression, but the methodologies of comparative sociolinguistics mean that multivariate analysis techniques are sufficient here. Multivariate analysis operates by running a ‘step-wise procedure of [. . .] multiple regression’ (Tagliamonte, 2006: 140) which uncovers the relative significance of each factor in the model and their distribution.2 Scores, generally called factor weights, range from 0 to 1; values above 0.500 are said to favour the use of the factor and those which are below 0.500 are said to disfavour it. (The first feature analyzed using multivariate analysis will provide a more detailed illustration of this.) In order to interpret the results, the ‘relative position of the factors weights, vis-à-vis each other’ (Tagliamonte, 2006: 145) need to be examined. The contribution of each factor group can be determined through its range, which involves subtracting the factor with the lowest score in a factor group from the factor with the highest score. Factor groups with high ranges are those which operate the strongest effect on the occurrence of the variant being studied.

62

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

In comparative sociolinguistics, the multivariate analyses of the same feature are compared across several varieties and three aspects are particularly important: Three lines of evidence deriving from the variable rule analysis may be adduced in interpreting its results: (statistical) significance of effect (at the 0.05 level), magnitude of effect, as determined by the range between the highest and lowest factor weight in a factor group, and hierarchy of constraints, or ordering of factor weights within a factor group. (Poplack & Tagliamonte, 2001: 92) The statistical significance is the least reliable of these aspects, as it is the most susceptible to outside influence (differences in the number of tokens, for example). Poplack and Tagliamonte (2001: 93) explain that ‘when a variable is affected in the same way by the same set of factors in several studies, a study with a larger number of tokens will tend to detect more of these factors as statistically significant than one with fewer tokens’. Because this book, like Poplack and Tagliamonte’s research, compares several varieties at once, statistical significance will be considered, but always alongside range estimates and hierarchies of constraints which are affected less by token numbers. Applying this methodology to non-native variation will be particularly useful in determining whether or not non-native speakers have acquired native patterns. If speakers from all three backgrounds can be shown to have the same constraint rankings and magnitude of effect for a given variable, then it will be supposed that they are using the same variety of English. Moreover, by comparing their results to those of native speakers, I will be able to determine whether their patterns are unique to Switzerland or show the acquisition of native sociolinguistic competence.

5.3 Categorizing the Results The methodology used in comparative sociolinguistics will be slightly modified for the purposes of the current research, given that it deals with native and non-native groups and because, as well as target-based variation (TBV), there are occasional instances of learning-related variation (LRV) which will need to be dealt with. The variation of each feature will be analyzed using multivariate analysis (when possible) and compared across the four groups (the three Swiss non-native groups and the native group). This will allow me to ascertain whether the non-native patterns are due to the acquisition of native speaker patterns or whether they are due to second language learning processes. If the non-native groups share the patterns of the native speakers then the variability can be judged to have been acquired,

Methodology

63

as following Poplack and Tagliamonte’s methodology the similarities are unlikely to be due to chance. On the other hand, if the Swiss groups have not acquired native sociolinguistic competence, this will point to one of two results: either the patterns demonstrate the presence of a Swiss-wide pattern (and possibly a Swiss English as a lingua franca) or each group will pattern differently because of source language influences. This is schematized in Table 5.1. The three variation pattern types for individual features will be determined as follows. •

•

Case A A feature will be categorized as case A (native-like acquisition of variability) if an analysis, using the methodology of comparative sociolinguistics, can conclude that the patterns of variation in the English of all three language groups are closely comparable to each other and to the native speaker control group. In such a case, the assumption will be that the non-native speakers have the same variable rules as native speakers. Case B Variables that do not pattern like native speaker variables, but in which all three language groups vary in the same way, are categorized as case B. This category will demonstrate the presence of a potentially (Swiss) ELF variable. If a variable patterns in the same way for all three language groups, then it is a likely point of convergence, especially if it can be demonstrated that the patterning does not match what would be found for non-native speakers of the Swiss languages in countries (i.e. France,

Table 5.1 Potential cases of variability Case

Patterns found

Outcome

Types of variation it is found with

Case A

French, German and Italian present similar variation patterns to native speakers French, German and Italian variation patterns are unlike native speakers but similar to each other French, German and Italian vary unlike native speakers and unlike each other

Sociolinguistic competence acquired

TBV

Focused form (or common problems of acquisition)

TBV, LRV

Variation influenced by native language only

TBV, LRV

Case B

Case C

64

•

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Italy and Germany) where English does not have a comparable lingua franca use. Case C Variables in which none of three language groups has the same patterns in English and furthermore none of them presents the same patterns as native English are categorized as case C. In this case there has been neither acquisition nor focusing but simply the cross-linguistic influence of the source languages or learning-related processes. In cases such as this one, the variation may turn out to be individual or free and speakers from the same linguistic background may not have the same patterns.

Features belonging to case B are considered to be potential rather than bona fide features of a single variety of Swiss English, because the patterns found might be shared with other non-native varieties of English as well. The identical patterns could be indicative of other processes, such as pidginization, general difficulties in learning a specific aspect of English syntax (common to all non-native speakers and not just the Swiss speakers) or due to some type of linguistic universal at work. Features found in case B are undoubtedly cases of focusing but determining whether it is restricted to Swiss speakers is beyond the scope of this book.

5.4 Features Analyzed Having established what the analyses of features might uncover, I will now introduce the rationale behind the selection of the four features chosen for analysis. Each will be discussed in detail in their respective chapters, but a few points are nonetheless useful here. All the features selected were ones which demonstrated TBV and which had, to a greater or a lesser extent, been studied in the output of native speakers of English already. Unlike previous studies on the acquisition of sociolinguistic competence, the variants were generally not distinguished primarily on the basis of formality, although there was often a stylistic element to the variation. Some features selected had two variants, whereas others had three or more, and some were selected in part because one of the variants was a zero form to establish how this affected acquisition. When picking suitable features for analysis, attention was also paid to choosing a mix of features which were known to be explicitly taught and features which either were not presented or were introduced as free variation in language teaching.

Future tense (will versus going to) The chapter on the future will examine the variable use of the future forms will and going to (Examples 1–2). Although there are differences in

Methodology

65

the distribution of will and going to in different native varieties, including a difference of overall distribution of the two variants between British and North American English, many of the factors which condition the variability have been found to be shared across varieties (Tagliamonte, 2002: 751ff). (1) I will try to keep the facts up to date. (o, German, email) (2) this is gonna be a sweet ride on the night train. (o, German, email) Both of the variants for the future were explicitly taught in class, which will allow me to establish whether overt teaching of variability has an effect on non-native patterns.

Relative pronouns This chapter will deal with the selection of relative pronouns by nonnative speakers. It considers the choice between the pronouns, who(m), that, which and the zero form (Examples 3–5). (3) but send me a list of people who want to go there. (*, French, email) (4) or at least the people that might concerne it! (k, German, email) (5) and the number of people Ø you are paying for. (h, German, email) Although there are a number of prescriptive rules which affect the use of the various relative pronouns and these are taught in Swiss schools, many of the underlying effects of variation (aspects tied to animacy, syntax, adjacency and definiteness) are not taught. This analysis will allow me to establish if these underlying factors nevertheless affect non-native speaker patterns of relative pronoun selection.

Complementizer that The selection of complementizer forms provides speakers with the option of either using an overt form (that) or a zero form (Examples 6–7). (6) I think Ø we will be 20 medical students. (f, French, email) (7) I think that he will be great in helping you. (b, Italian, email) Native speaker use is affected by factors such as lexical verb, grammatical person and tense (Tagliamonte & Smith, 2005: 301). Non-native speakers are not taught that these factors affect variability; in fact the two variants are presented largely as free variation. This chapter will consider whether the lack of overt teaching of the variability hinders non-native speakers’ acquisition of sociolinguistic competence.

66

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Additive adverbials also, as well and too This chapter will focus on a case of native speaker lexical variation, namely the variation between the three additive adverbials also, as well and too (Examples 8–10). (8) I think the grammar is changing as well now. (S, French, interview) (9) Actually, this message is also a big HELP request. (j, French, email) (10) it was a situation that I shared, too. (f, French, email) The variation of the three forms has been found to be linked to style and to register. Furthermore, one of the variants, also, is acceptable in a wider range of positions than the other two forms, which will allow me to uncover whether non-native speakers were likely to extend the use of the other two variants to positions not acceptable to native speakers. The analysis will consider both whether the non-native groups share the patterns of native speakers and whether the three groups have similar problems in terms of placement. This is another feature where the variation is presented as free variation. The analysis of these four target-based variables will provide valuable data on their own, but examined together they will also make it possible to establish a further point: whether overt teaching of variability facilitates the acquisition of the native-like patterns. While the variation in that and zero complementizers is not taught at all, the variation in relative pronoun use is only partially taught and finally the variation between will and going to is presented to some extent. Will there be an effect of this teaching in the results? Ultimately, the overall results will offer suggestions into the extent to which ELF use influences the acquisition of sociolinguistic competence.

Notes (1) These programs are Varbrul and Goldvarb; see Young and Bayley (1996) for an indepth discussion of how to use the two programs. (2) Please see Tagliamonte (2006: 128–157) and Young and Bayley (1996) for a more detailed presentation of the processes involved in multivariate analysis.

6

Future

6.1 Introduction Unlike French and Italian, English does not have a morphological future tense, but instead makes use of other syntactic strategies to convey futurity (Bybee et al., 1991). There are a number of ways to express the future in English, arising from the fact that ‘futurity, modality and aspect are closely interrelated’ (Quirk et al., 1985: 213). Researchers who have examined the distribution of these various forms have demonstrated that variability is constrained by a range of factors (Berglund, 1997; Poplack & Tagliamonte, 1999; Torres Cacoullos & Walker, 2009a). Moreover, much of the research found cross-dialectal and cross-speaker similarities with respect to factors conditioning the variability. This underlines that some of the rules governing the distribution of future forms are shared by all native speakers of English and consequently means that futurity is an excellent feature to examine in terms of non-native acquisition of sociolinguistic competence. Also, the ‘rules’ affecting the use of two main future forms – will and going to – are taught in Swiss schools, which means that I can also attempt to establish the degree to which overt teaching can help with the acquisition of native patterns of variation. This chapter will first present the forms used to express future in English, and briefly introduce the future tense in French, Italian and German, before focusing on the analysis and discussion of the results.

6.2 Future Forms Will future The modal auxiliary construction is generally associated with prediction and is often said to be ‘the most common way of expressing futurity’ in English and to approximate ‘a colourless, neutral future’ (Quirk et al., 1985: 214). The modals used are will, shall or ’ll (Examples 1–3). The form will can be used with all persons, while a ‘strong teaching tradition has upheld the 67

68

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

use of shall [. . .] with a first person subject in formal style’ (Quirk et al., 1985: 214).1 (1) Our next national meeting will take place. (e, French, email) (2) we shall vote new bylaws for IFMSA-Switzerland. (b, Italian, email) (3) I’ll be quiet. (h, German, email)

Will + progressive infinitive This construction ‘combines reference to a future time with the “temporal frame” associated with the progressive’ (Quirk et al., 1985: 214) (Example 4). (4) I will be arriving on Saturday morning. (m, German, email)

Going to future Be going to followed by the infinitive is ‘frequently used to express futurity, especially in informal speech’ (Quirk et al., 1985: 214) (Example 5). Its general meaning is ‘future fulfilment of the present’, but it can also be used to express ‘future fulfilment of present intention’ and ‘future result of present cause’ (Quirk et al., 1985: 214). (5) I don’t think she is going to come in Lausanne. (S, French, email)

Simple present future The ‘next most common’ construction after the will future is said to be the simple present (Quirk et al., 1985: 215); however, it is primarily restricted to dependent clauses. (6) Tomorrow (7th of April) is the World Health Day. (g, French, email) (7) If I don’t get answer from you beautiful people, I’ll start calling you home at every hour of the night. (c, Italian, email)

Other future forms Quirk et al. (1985) discuss other future forms such as the progressive present, and be to. These appear less frequently than the other forms, however, and will not be considered here. Quirk et al.’s judgments about the relative frequency of the future variants have been mostly corroborated by empirical data. Poplack and Tagliamonte (1999: 326) found that in native spoken Ottawa English, the most mainstream of the varieties they considered, the will and the going to forms made up 74% of the overall distribution, whereas the progressive present and the simple present future together accounted for only 26%.

Future

69

Although a number of the constructions to express future tense can be found in non-native English, this study will focus only on the two most common future constructions, the will future (including the will + progressive variant) and the going to future. Analysis of these two variants will make possible a more direct comparison with the results of Poplack and Tagliamonte (1999) and Berglund (1997) as these are the variants which were examined most closely in their studies. Furthermore, these are the two which are used in the most similar way, despite some differences in terms of their semantics: there are occasions where the choice of one construction (say will + infinitive) rather than another (say be going to + infinitive) has a scarcely perceptible effect on meaning. (Quirk et al., 1985: 213) The interchangeableness of the two variants is commented upon in language teaching literature; for example, Soars and Soars (1987: 54) note that the ‘choice [between the two forms] depends on how the speaker sees the event and not on its certainty or nearness to the present’. For the most part, however, the English grammar books used in Switzerland present a fairly prescriptive view of the forms. Students are taught that will is used for prediction and going to is used for ‘future fulfilment of the present’ (Quirk et al., 1985: 214) and are tested on this. Despite this, students are not made aware that in native English register and formality have an effect on the use the variants (as found by Berglund, 1997). Taking as a starting point that variability is a difficult feature for nonnative speakers to acquire, this chapter will attempt to establish whether the formal teaching of variation can contribute to the acquisition of variable patterns. Will the fact that the Swiss speakers have been explicitly taught when to use will and when to use going to mean that their production of the two variants will match that of native speakers? On the surface, all three Swiss groups use both future forms in their English output, as is demonstrated in Examples (8)–(12) from emails below.

French speakers (8) I’ll see with Basel who is going to do the inscription. (*, French)

German speakers (9) I would really apreciate if you could prepare yourself for your standing committee you’re gonna join. (h, German) (10) About the Schoki, I will buy the 10 Kg and take them. (o, German)

Italian speakers (11) I’ll also contact our social programme group. (V, Italian) (12) Let me know who else is gonna come!? (P, Italian)

70

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

6.3 Future Forms in the Source Languages If the three source languages have variants and distributions that are related to the future variants in English, it may contribute to students’ acquisition of the form in a native-like manner.

Will form Unlike English, French and Italian have a morphological future tense. This tense is marked inflectionally on the verb itself; each person has a separate morphological future form tagged onto the verb root (Examples 13–16).

French (13) Demain j’irai à la plage (14) Demain tu iras à la plage

Italian (15) Domani andrò alla spiaggia (16) Domani andrai alla spiaggia German syntax also uses inflectional morphology but, unlike French and Italian, it does not have a separate construction for the future. Instead, as in English, one of the forms to express futurity is the use of a modal verb (werden) (Example 17). This is not, however, a very frequent form in spoken German and ‘in practice such forms are rather restricted in use’ (Fox, 1990: 183). A similar form exists in Swiss German as well, but as in the case of High German the werden construction is not frequently used. (17) Morgen werde ich an den Strand gehen

Going to form Italian and High German do not have a future form comparable to the English going to. French, on the other hand, has a construction similar to the English one (Bybee et al., 1991). This construction involves the verb aller (to go) and an infinitive, allowing a French speaker to say, for example, ‘je vais manger’ as well as ‘je mangerai’. The factors influencing the use of the going to form in French are similar to those of English (Emirkanian & Sankoff, 1985; Roberts, 2012), as they share a ‘future fulfilment of present situation’ sense discussed by Quirk et al. (1985: 214). Moreover, the aller + inf is the main form used in informal French at present. Studies of Canadian French (Emirkanian & Sankoff, 1985: 194; Poplack & Dion, 2009) revealed that it was used nearly 80% of the time, far more than the inflectional form. The use of the aller form is also said to be quite high in continental French (including Swiss French) (Ball, 2000: 87; Roberts, 2012). Swiss German dialects, too, have a similar construction ‘i gan go esse’ (I’m going to go eat).

Future

71

Table 6.1 Comparison of future structures Future forms English French Italian German Swiss German

Suffixation

Modal verb

‘Going’ based form

√

√ √

√ √ √ √

√

Other future forms The use of the simple present to convey the future is available in the three Swiss languages. Its function in French, German and Italian is similar to that in English. Unlike English, however, it is not restricted to dependent clauses. This form is the one most commonly used in German (Fox, 1990: 183). Table 6.1 summarizes the similarities and differences among the four languages.

6.4 Previous Studies Previous studies on future forms in native English (Berglund, 1997; Poplack & Tagliamonte, 1999; Tagliamonte, 1997, 2002; Torres Cacoullos & Walker, 2009a) have determined that the constraints on the choice of variant are style, register (oral versus written speech; Berglund, 1997), the point of reference, age of the speaker and personal pronoun used (Poplack & Tagliamonte, 1999). These studies found varying rates of going to use, ranging from 21% in Berglund’s analysis of the London Lund Corpus, to 30% in Tagliamonte’s York corpus, to 45% in Poplack and Tagliamonte’s Ottawa corpus and 51% in Torres Cacoullos and Walker’s Quebec English corpus.2 These are the figures this analysis will use as a benchmark for the non-native results. Poplack and Tagliamonte (1999) found that the internal factors which constrain the use of the variables were cross-dialectally shared for the most part, so British and North American varieties of English had very similar hierarchies and ranges in their variable rules for the selection of will and going to.

6.5 Setting Up the Analysis All the tokens of will and going to as future forms were extracted from the IFMSA corpus (both email and oral sections) and the native speaker email corpus. Tokens of going to that only conveyed directional meanings were, of course, excluded (Example 18). (18) All the people who are going to Denmark are kindly invited to a big party. (h, German email)

72

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

The tokens were coded for variant (will, going to and shall as a sub-variant of will), for grammatical person, for whether the form was contracted (‘ll or gonna versus will and going to) and, in the Swiss data, speaker identity, speaker native language and register. The contraction of the form is linked to orthographic convention and was not considered in the oral data. It was chosen as a factor because it can help establish the formality of the emails. Grammatical person was considered for two reasons; first of all, Poplack and Tagliamonte (1999: 335) had found that going to occurred less frequently with first person subjects than with other subjects. Secondly, this factor will help establish if different subjects occur at different rates in oral and email registers and make it possible to better understand the results. The non-native tokens were also given an additional code if they were felt to diverge in some way from native speaker English (Examples 19–20). This makes it possible to examine the learning-related variability as well. (19) A booklet that will include all the information I will get from you. (b, Italian, email) (20) They shall say the swiss have good organized hospitals on a high level. (h, German, email) As will become clear in the analysis, the nature of the variable in the Swiss speakers’ English meant that fewer factors were considered than in many previous studies.

6.6 Results Overall distribution There are nearly 1500 tokens of email and oral data combined for the three non-native groups, and a further 237 tokens from the native email corpus. Tables 6.2 and 6.3 give the overall distribution of will and going to for the non-native speakers for email and oral data, respectively. These tables reveal that the rates of going to are low for all groups in both the oral and the email data. They show, however, that the German speakers have significantly higher rates of going to use than the other groups in the email data, despite the fact that, in all three groups, most of the speakers with more than 10 tokens do use both variants (100% of Italian speakers, 60% of German and 71% of French ones). Although going to is in all the speakers’ linguistic repertoire, will nevertheless is used at very high rates. The results of the oral data must be taken with more caution, especially where the German speakers are concerned. The German distribution consists of only 20 tokens and all of the going to tokens in the German oral data stem from a single speaker (who uses both will and going to), whereas the other

Future

73

Table 6.2 Percentages of will versus going to in non-native speaker emails Native language

% of will

% of going to

Number of going to tokens

Total N

Italian French German Total

99 99 93.5 98

1 2 6.5 2

7 7 13 27

650 314 201a 1165

Notes: aChi-square = 19.7402324545435; p is less than or equal to 0.001. There is a significant difference across language groups.

Table 6.3 Percentages of will versus going to in non-native speech Native language

% of will

% of going to

Number of going to tokens

Total N

Italian French German Total

99 95 65 96

1 5 35 4

1 3 7 11

186 64 20 270a

Notes: aThe low numbers for going to in three groups make it impossible to use chi-square to determine if the distribution between the three groups is significant.

German speakers use only the will form. Except for this one speaker, then, the German speakers are quite similar to their French and Italian counterparts. The apparent lack of difference between the oral and the email data for the non-natives is an important finding, as previous studies (Berglund, 1997: 15) had found that register had a significant effect on going to. Tables 6.4 and 6.5 present the overall distribution of going to in native speaker studies. Table 6.4 provides figures from Berglund (1997), Poplack and Tagliamonte (1999), Tagliamonte (1997) and Torres Cacoullos and Walker (2009a), while Table 6.5 provides the figures from the native email data. The figures from the email data are lower than those for oral data – as was expected, since going to has been found to occur less frequently in writing and email than in speech. Although Berglund (1997: 15) found 21% of Table 6.4 Overall distribution of will and going to in earlier studies (oral data)

Quebec (Canadian English) Ottawa (Canadian English) York (Northern British English) London (LLC) (Southern British English)

% of will

% of going to

49 54 70 79

51 46 30 21a

Notes: aThe figures for Quebec come from Torres-Cacoullos and Walker (2009a: 328), those for Ottawa come from Poplack and Tagliamonte (1999: 326), those for York from Tagliamonte (1997) and those for London from Berglund (1997: 15).

74

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Table 6.5 Percentages of will versus going to in native speaker emails

Native English total

% of will

% of going to

Total N

89.5

10.5

237

going to in the oral data, the written corpora she considered (the Brown Corpus and the LOB) had significantly lower rates (around 5%). In the case of will versus going to, emails are much closer to writing than speech. How does this compare to the non-native data? The non-native email results are very low in terms of going to but, unlike the native data, there is no shift towards higher rates of going to in speech. Removing the one German speaker who had very different frequencies of use, the overall distribution of going to in the oral data remains very low in all three non-native groups. Figure 6.1 presents the oral/email distribution across the four linguistic groups.3 Figure 6.1 confirms two points: first of all, that the non-native rates are consistently lower than the English rates; secondly, that the non-native speakers do not have the register distinction native speakers have. Although there is a slight increase of going to in oral contexts for the French speakers, the difference is not statistically significant.4 These results underline the importance of using a comparable corpus of native emails with the non-native data: the difference between natives and non-natives is not due solely to register. Although it seems plausible to claim that the non-native emailers are more formal in their writing than the native emailers, the results do not fully support this conclusion, as the discrepancy between native and non-native patterns is maintained in speech.

Grammatical person The near categoricity of will use means that a further analysis into many of the factors examined in previous studies is impossible. One factor which it is possible to consider is grammatical person. Poplack and Tagliamonte 50 40 30 20 10 0 French

German E-mail

Italian Oral

Figure 6.1 Percentage of going to in oral versus email data

English

Future

75

Table 6.6 Percentage of going to by grammatical person

Italian German French English

First person (sing and plural)

Non-first person

% of going to

Total N

% of going to

Total N

1 1.1 2.4 8

309 88 126 76

1.2 10.6 2.1 12

341 113 188 161

(1999: 335) found that, although historically going to was linked to speaker attitude and hence was favoured in first person contexts, in most contemporary varieties of English going to had become grammaticalized to the point that it occurred more often with non-first person subjects. Table 6.6 examines the four linguistic groups in terms of the email data. The non-native oral data was not included in this analysis as the present results were not directly comparable with previous studies5 and because of a low number of going to tokens in non-native speech. The proportion of first person/non-first person tokens was very similar for all groups across the register, however, with first person subjects representing around 40% of tokens. There is no difference in the distribution of will and going to in terms of grammatical person for the Italian and the French speakers; their rates of going to remain very low regardless of person. This is likely to be in part due to the overall low rate of going to tokens. The English and German speakers, however, follow the predicted pattern in that they have higher rates of going to in non-first person contexts. The German pattern is especially plain, with a rate of 1% in first person contexts going up to nearly 11% in non-first person contexts.6 The German speakers have acquired the native English constraint uncovered by Poplack and Tagliamonte (1999: 335), but the Italian and French speakers have not.

Full versus contracted forms Another factor that may reveal further similarities or differences between the groups is whether the verb forms will and going to are contracted to ‘ll and gonna, respectively. The contraction of the two variants functions in slightly different ways, so we will consider full and contracted forms of the two variants separately. Although this factor cannot elucidate factors affecting the selection of one of the variants over the other, it enables an examination of the effect of the email register. Table 6.7 presents the contraction in the native emails, while Table 6.8 presents the non-native emails. The native speaker rates of contraction are 23% for will and 8% for going to; this is fairly low and underlines how even messages written in an informal style are affected by the register of email. Turning to the non-native

76

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Table 6.7 Percentages of full forms versus contracted forms in native speaker emails Native English

% of full form

% of contracted form

Total N

Will Going to

77 92

23 8

212 25

Table 6.8 Full and contracted future forms in non-native English emails

Italian Will Going to French Will Going to German Will Going to

% of full form

% of contracted form

Total N

81 71

19 29

643 7

89 86

11 14

307 7

88 77

12 23

188 13

groups, we can see that they differ considerably from the native group, although they are quite similar to each other. Their rates of will contraction are slightly lower than in natives, but their rates of going to contraction are higher. Figure 6.2 compares the percentages of will and going to contraction across language groups. 100 90 80 70 60 50 40 30 20 10 0

English

Italian Will

Figure 6.2 Percentage of contraction

French Going to

German

Future

77

Table 6.9 Contracted forms of will by grammatical person First person

French German Italian English

Other

% of contraction

Total N

% of contraction

Total N

20 17 29 41

123 87 306 70

4 7 10 14

184 101 337 142

The high use of gonna by the non-native speakers in emails underlines that is probable that it is being used as an idiomatic form in the English of the Swiss speakers and functions as a lexical chunk that only occurs in specific contexts. If it were due purely to the relative informality of the emails, then will would have been contracted more often as well. Additionally, because the proportion of going to increases in informal contexts in native speech, if gonna were truly linked to informality for the non-native speakers, then the overall use of going to forms would have been higher. Although only the German speakers had shown a significant difference between the first person and non-first person subjects in terms of the distribution of will and going to, this factor also needs to be considered for the contraction of will. If first and non-first person subjects do not contract in the same way across language groups then this could have an impact on our results, as Table 6.6 revealed that nearly half the future tokens in the data were first person. Table 6.9 considers this, examining the contraction of will in first person singular and non-first person contexts across the linguistic groups. Table 6.9 shows that all groups are more likely to contract will with first person singular subjects. This distribution was found to be statistically significant in all four groups. The non-native speakers have the same hierarchy as the native speakers, but lower rates of contraction in all contexts.

Learning-related variation In addition to comparisons with native speaker patterns, this feature also offers insight into LRV, which may help explain why the non-native speakers have not acquired the full variability natives demonstrate. A further analysis will also reveal whether any difficulties with the use of the future is shared across the three linguistic groups or restricted to a particular language. Given the overall low use of going to in the non-native data set, it is unsurprising that most of the tokens which were judged to be non-native in some way are of will. Generally, this involves cases where will is used instead of a continuous form, instead of would or instead of a present tense form. These types of non-native token are detailed below. These are not cases

78

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

where going to was the only variant suitable, but rather tokens where other issues occurred.

Will in place of a continuous form There were a number of cases where will was used in situations where Standard English uses a present (or future) continuous tense (Examples 21–22). This overuse of will was especially common when it was with the verb to come. (21) Just let us know how many of you will come, so that we buy enough food (c, Italian, email) (22) unfortunately I won’t meet you in Basel . . . because I will stay in New York at that time (d, German, email)

Will for would Some of the learning-related tokens found in the data were caused by the speakers using will instead of would. Standard English requires a conditional tense rather than a future tense in some sentences (Examples 23–24). (23) I am leaving for Malta on the 2nd March, so it will be nice of you, if you could send me these reports before Thursday (g, French, email) (24) In all cases you will have to send to me the REPORT of the activities of this past 6 months. I will be extremely grateful if you could sent them AGAIN to me, even if you sent them out already (b, Italian, email, emphasis in original)

Tense concord Learning-related variation (LRV) was commonly caused by tense concord (Examples 25–27), namely cases where the tense agreement rules followed were those of the native language rather than English. In conjoined future tense clauses, French, Italian and German tend to have a future form in both the main and the subordinate clause, whereas English has a future tense only in the main clause and a simple present in the subordinate clause (as was mentioned in Section 6.2). (25) This summer, when the intranet section will be available, an antivirus will maybe be in there for download (j, French, email) (26) All in all it seemed to be a good appointment and I will let you more about the story as soon as I will know more, alright? (d, German, email) (27) As soon as I’ll get them I will send them to you all (b, Italian, email)

Shall Shall was occasionally used with pronouns other than I and we (Example 28). This may in fact be a case of hypercorrection rather than learning-related difficulties, in that the non-native speakers are using a form which they have been taught as being more appropriate than will. Non-native English grammar books try to enforce the use of shall although it is not often used in colloquial Standard English. There are no instances of shall in the native speaker data.

Future

79

Table 6.10 Number and percentage of non-native sentences in three language groups

Total number of tokens judged non-native Percentage of non-native tokens

French

German

Italian

Total

20 6.3

16 7.9

59 9.1

95 8.1

(28) They shall say the swiss have good organized hospitals on a high level and they shall say, they are (h, German, email) None of the non-native error types was associated with a particular speaker group, so the three groups are very similar in terms of LRV as well as targetbased variation. Overall, about 8% of the tokens of will did not follow native speaker norms and the three groups have similar rates (between 6% and 9%) of learning-related tokens (Table 6.10).7

6.7 Discussion The above analysis has revealed the following points about the use of will and going to by the non-native Swiss speakers: • • • •

• • •

The rates of going to use are far lower in the non-native speakers than in native speakers. None of the non-native groups shows a change in going to use between email and oral data, whereas native speakers show a shift. Despite there being statistically significant differences between the nonnative groups in terms of overall distribution, they are far more similar to each other than they are to the native speakers. Grammatical person has no effect on the Italian and French speakers, but the German speakers have the native English pattern found by Poplack and Tagliamonte (1999: 335), whereby non-first person subjects have more going to. The non-native groups also show themselves to be considerably different from the native speakers in terms of contraction and yet similar to one another. Considering only will, all four groups contract more in first person contexts than elsewhere, but non-native overall contraction is lower, however. LRV tokens are found in all three non-native groups at similar rates; thus the use of future variants does not appear to be influenced by the source languages.

This feature presents some aspects where the non-native speakers match the native speakers, but there are also a number of ways in which the non-natives

80

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Table 6.11 Summary of similarities and differences in distribution Native English

French

German

Italian

Percentage of going to

12 (email) 20–45 (oral)

1

7

2

Oral versus email distinction

Yes

No

No

No

Grammatical person

Non-first person favours going to

No pattern

Non-first person favours going to

No pattern

Percentage of will contraction

23

11

12

19

Percentage of going to contraction

8

14

23

29

Will contraction

First person > Other

First person > Other

First person > Other

First person > Other

are not following native patterns (see Table 6.11). None of the non-native groups has the oral/email shift found in the native speakers and have far lower overall rates of going to than native speakers. Register is a very important factor in the use of going to in native English, so the overwhelming impression is that the non-native speakers have not acquired the native patterns. This is despite the fact that they pattern similarly to the native speakers in some respects. This feature cannot be considered one where the native speakers and non-native speakers share the same patterns. How do the three non-native groups compare with each other, then? As discussed, the three groups have low overall rates of going to and do not increase their use of it in speech. All three groups contract will less often than going to and show a number of similarities in terms of LRV. The German speakers have a grammatical person effect which did not operate for the Italian and French speakers, so there are some differences, but it does seem as if the non-matching of the native patterns is tied to general acquisition processes rather than to individual linguistic differences.

6.8 Conclusion In terms of the acquisition of non-native sociolinguistic competence, this chapter has revealed, first of all, that Swiss speakers of English do not make use of the full range of variability found in the future, rarely using one of the

Future

81

variants (going to) in their repertoire. It has also shown that this appears to be tied to general acquisition processes, rather than being language specific. Similarities to English in the structure of the future in some of the source languages do not appear to have influenced the acquisition: the French speakers who have a very similar construction to English going to do not use the English variant any more frequently than the other two groups. The speakers of Italian, which does not have a periphrastic structure comparable to English, are not very different from the other two groups. My findings are similar to previous studies on the acquisition of sociolinguistic competence by non-native speakers, as they also found that variability was often not acquired, particularly for features with variants which were marked stylistically. While going to is a Standard form, it is nonetheless used less often across written media than in speech. The results demonstrated that the Swiss students do not have this oral/written distinction. Overall the results underline that although Swiss English classes teach their students some of the differences between the two variants, the nonnative groups have not acquired this variability. If they were using the patterns natively then there would have been a higher rate of going to in their speech, which there is not. The implications of this finding for future pedagogical practice will be considered in Chapter 10.

Notes (1) Shall is also used in colloquial English English in first person questions consulting the needs of the listener (as in Example 2) (Quirk et al., 1985: 230). (2) The rates for Torres Cacoullos and Walker were calculated by considering only the tokens of will and going to in Table 1 (Torres Cacoullos & Walker, 2009a: 328). (3) The oral English figures come from the York speakers in Tagliamonte (1997) which represents the group most similar to the email corpus. (4) Chi-square = 1.24734978624342; p is less than or equal to 1. (5) Poplack and Tagliamonte (1999) had provided the factor weights for grammatical person but not percentages. (6) Chi-square = 7.2534847562222; p is less than or equal to 0.01. (7) Removing these does not substantially alter the results presented earlier.

7

Relatives

7.1 Introduction English native speaker relative pronoun choice has been the subject of a sizeable amount of research (Ball, 1996; D’Arcy & Tagliamonte, 2010; Quirk, 1957; Tagliamonte et al., 2005, among others). Relative pronoun choice is a ‘clear case of syntactic variation’ in native English which ‘from intuitive reflection [can be said to be] subject to strong constraints’ (Guy & Bayley, 1995: 148). As such, it is a useful feature to consider in terms of non-native variation, both because of previous research but also because the variation is conditioned in very specific ways. This analysis will establish the extent to which non-native speakers follow these constraints. Relative clauses are said to be ‘fairly frequent in speech as well as writing’ (Guy & Bayley, 1995: 148), and so offer an additional benefit when examining emails. Relative pronoun choice is a particularly intriguing site for study in nonnative speaker data as it involves both categorical and variable constraints, as demonstrated in Examples (1) and (2). (1) the people that/who/*which are registered. (modified from *, French, email) (2) the infos that/which/*who are in the booklets. (modified from b, Italian, email) The prescriptive and categorical aspects of use are taught to non-native speakers, but the variable rules are presented as being free variation. By this, I mean that the non-native speakers are simply taught that several pronouns are acceptable in the same situation. The aim of the present analysis is then two-fold; it will focus on establishing the extent to which the non-native speakers differ from the prescribed norm (so their level of learning-related variation), while at the same time judging how closely they follow the variable norms (i.e. target-based variation). 82

Rel at ives

83

Considering a feature such as this one, where the variability is only partially taught, will make it possible to determine to what extent non-native speakers are aware of and capable of dealing with variable rule patterns. Because the full variable rules are not taught, if the non-native speakers have acquired them, then it must be by modelling the native speakers. The categorical restrictions on relative pronoun choice are presented directly below, while those which are merely preferences in terms of relative pronoun selection, and as such belong to native speaker variable rules, will be examined in the Section 7.6, which also deals with the extraction and coding of the tokens.

7.2 Relative Clauses Relative clauses fulfil similar functions to conjunctions and serve to link two or more clauses together (Quirk et al., 1985: 1244ff). The two clauses in (3) are joined together in (4): (3) I saw the man. The man was wearing a red hat. (4) (a) I saw the man who was wearing a red hat. (b) The man whom I saw was wearing a red hat.1 English has a number of different relative pronouns (who, whom, which, that, when, why, where, whose), as well as a zero form. Examples 5–9 illustrate the use of some of these pronouns with data from the Swiss emails. (5) i’ve found someone who is willing to do the homepage. (h, German, email) (6) I need deans, professors, doctors to whom you have a good contact. (c, Italian, email) (7) I think this is a good opportunity to enlarge our activities, especially with the help of an organization which is already settled. (b, Italian, email) (8) Despite the incredible tiredness that fell over us in the last days. (A, German, email) (9) a document Ø we sent to our Doyen in Geneva. (f, French, email) The selection of relative pronouns by native speakers is governed by a number of prescriptive rules. That can be used with any type of subject, which is restricted to inanimate subjects, while who can only be paired with animate subjects (as demonstrated in Examples 1 and 2). Whether a relative clause is restrictive or non-restrictive also plays a role in the selection of pronoun. If the clause which is relativized provides no vital information, then the clause is classed as a non-restrictive relative

84

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

phrase (Example 10), while if it does provide necessary information it is classed as restrictive (Example 11) (Quirk et al., 1985: 1245). (10) My sister, who is fifteen, went to Rome last year. (11) The girl whom I was telling you about is his sister. In written English, convention has it that non-restrictive relative clauses are signalled by the use of commas, separating the additional information from the rest of the sentence (Quirk et al., 1985: 1245). The distinction between the two type of clauses has been said to be ‘crucial for quantitative studies of relative markers, because NRRs [i.e. non-restrictive relatives] strongly favour WH- forms in standard English whereas RRs [i.e. restrictive relatives] do so variably’ (Ball, 1996: 229). In addition to restrictiveness, there are two main properties affecting the selection of relative pronouns in English. Relative pronouns can: (i) show concord with [their] antecedent, i.e. the preceding part of the noun phrase of which the relative clause is a postmodifier [external relation]. (Quirk et al., 1985: 1244) and (ii) indicate [their] function within the relative clause either as an element of clause structure (S, O, C, A), or as a constituent of an element in the relative clause [internal relation]. (Quirk et al., 1985: 1244).

External relation The factor which is designated as external relation in Quirk et al.’s terminology is reflected in English relative clauses ‘on the basis of a two-term “gender” system, personal and nonpersonal, and applies only to the wh-series’ (Quirk et al., 1985: 1245). Although that and the zero form can both be used either with animate or inanimate objects (Examples 12 and 13), the two whforms are distinguished in that who (and its oblique form, whom) is used solely with animate subjects or objects, whereas which is used only with inanimates (Examples 14–17). (12) (13) (14) (15) (16) (17)

The man that/Ø I saw is very smart. The toy that/Ø I saw is new. The man whom I saw is very smart. *The man which I saw is very smart. *The toy whom I saw is new. The toy which I saw is new.

Internal relation For internal relation as described by Quirk et al., there are three main types of relative clauses: those where the relative pronoun is the subject (Example

Rel at ives

85

18), those where it is the direct object (Example 19), and those where it acts as a prepositional object (Example 20). (18) The man who won the prize is very smart. (19) The man I saw is very smart. (20) The man to whom I talked is very smart. That and which can be used in all three types of relative clause (Examples 21–23). Who can be used in all three types as well, although it is sometimes modified to whom in direct and prepositional object clauses (Examples 24–26). The use of whom is rather formal and it is usual, in informal styles, to find who used in non-subject relative clauses (Example 27). In Standard English, zero, unlike the other relative pronouns, can only be used in nonsubject clauses (Examples 28–30). (21) (22) (23) (24) (25) (26) (27) (28) (29) (30)

The toy that/which was broken was mine. The toy that/which I broke was yours. The box that/which that the toy was in was red. The man who won the prize is very smart. The man whom I saw is very smart. The man to whom I talked is very smart. The man who I talked to is very smart. The man I saw is very smart. The man I talked to is very smart. *The man won the prize is very smart.

Although all four relative pronouns are acceptable in prepositional relative clauses of certain types, the structure of the sentence is modified in different ways depending on which pronoun is used. Prepositional relative clauses can either be pied-piped or stranded and relative pronouns can be distinguished depending in which of these two structures they are grammatical. In piedpiped relative clauses the preposition is placed immediately before the pronoun (Example 31), whereas in stranded relative clauses the preposition is left until after the verb in the subordinate clause (Example 32) (Quirk et al., 1985: 1246; Tagliamonte et al., 2005: 95).2 (31) The man to whom I talked was very nice. (32) The man Ø I talked to was very nice. The relative pronoun that and the zero form can only occur in prepositional clauses that are of the stranded type, while which and who(m) can appear with both stranded and pied-piped clauses. Stranded prepositions are generally avoided in writing and in formal speech despite the fact that they are frequently found in everyday conversation (Quirk, 1957: 100).

86

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Table 7.1 Possible relative pronouns

Subject Personal Non-personal Object Personal Non-personal Prepositional Personal Non-personal

Who

Which

That

Zero

Number of pronouns considered grammatical

√ *

* √

√ √

* *

2 2

√ *

* √

√ √

√ √

3 3

√ *

* √

√ √

√ √

3 3

Table 7.1 presents the relative pronouns used in Standard English depending on animacy and the syntax of the embedded sentence. It demonstrates that, although the internal and external relations place a number of constraints on the choice of relative clauses, there is nonetheless leeway, as there are always at least two possible variants for any given type of relative clause. These prescriptive rules of Standard English for relative pronoun selection are the ones that are taught to non-native students of English. Swiss students will have learnt which pronouns not to use depending on the context, but there is no focus on which of the remaining possible forms are most likely to be used by native speakers. The next section will briefly present a few points about the syntax of relative clauses in the source languages, as this may, at a later stage, provide some insight into how the non-native speakers acquired and dealt with English variability.

7.3 Relative Clauses in the Source Languages The structure of relative clauses in the three source languages is rather different from that of English, as well as from one another. As there are so many factors at play in the choice of relative pronoun in English, the factors that have a role in relative pronoun selection in the source languages as well need to be noted, in order to gain better insights into their potential crosslinguistic influence on English. The aim of this section is not to fully present the syntax of relative pronouns in the three Swiss languages, but rather to briefly introduce the main aspects of each in order to establish what the similarities or differences in terms of relative pronouns are between each language and English.

Rel at ives

87

French In French, the main relative clauses are qui, que and dont. The choice of pronoun in French is determined almost entirely by the grammatical function of the relative; presenting the matter very schematically, qui is used as a subject relative, que as an object relative and dont as a prepositional object relative (see Table 7.2). Qui can also be used in prepositional relative clauses when in conjunction with a preposition. Animacy does not dictate the form of the relative pronoun selected.3 In fact, in French, there is no possible variation in relative pronoun choice, as for any given sentence with a relative clause only one variant is acceptable. Furthermore, there is no comparable form to the English zero form in Standard French.

Italian Generally speaking, there is one main relative pronoun in Italian, che. This pronoun is used for animate and inanimate objects and for direct and indirect objects. For prepositional clauses, a determiner marking the gender of the relative subject with quale, or the form cui can also be used (see Table 7.3). As in the case of French, Italian has no zero form of the relative pronoun.

German German, like French, makes a difference between direct and indirect objects and does not distinguish animate from inanimate. The case and gender of the relativized noun is important, however, because the relative pronouns are in fact determiners. See Table 7.4.

Table 7.2 Examples of French relative pronouns Subject relative – QUI La fille qui est venue (animate subject relative – feminine subject) Le courier qui est arrivé (inanimate subject relative – masculine subject) Object relative – QUE La fille que j’ai vue (animate object relative – feminine object) Le courrier que j’ai reçu (inanimate object relative – masculine object) Prepositional relative - DONT La fille dont je t’ai parlé (animate prepositional relative – feminine object) Le livre dont je t’ai parlé (inanimate prepositional relative – masculine object) Prepositional relative – QUI + preposition La fille à qui j’ai donné une rose (animate prepositional relative – feminine object) Le courrier auquel j’ai répondu (inanimate prepositional relative – masculine object)

88

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Table 7.3 Examples of Italian relative pronouns Subject relative – CHE la ragazza che è venuta il libro che è caduto Object relative – CHE La ragazza che ho visto Il libro che ho comprato Prepositional relative – Det + QUALE, CHE, Prep + cui La ragazza la quale/che mi piace Il fiore il quale/che mi piace La ragazza a cui ho dato il libro Table 7.4 Examples of relative pronouns in (Swiss) German Subject relatives – DER, DIE, DAS Der Mann, der gekommen ist Die Frau, die gekommen ist Das Mädchen, das gekommen ist Object relatives – DER, DIE, DAS (+ declension) Der Mann, den ich gesehen habe Die Frau, die ich gesehen habe Das Mädchen, das ich gesehen habe WO dr Maa wo mer geschter gseh hei (Penner & Bader, 1995: 135) dr Maa wo mer geschter mit-em gschpiut hei (Penner & Bader, 1995: 135)a Notes: aTranslation of the Swiss German sentences ‘the man that we saw yesterday’, ‘the man whom we played with yesterday’.

In Swiss German, the main form of relative pronoun is wo, and this is used for all noun cases. This is used alongside the German forms. German has no zero form. Most important for the analysis of sociolinguistic competence is the fact that none of the source languages has a zero form comparable to English: if the Swiss students use it and use it at similar rates and in similar ways to native speakers then this will demonstrate that they have acquired the variability.

7.4 Previous Studies Quirk et al. (1985) provides a useful starting point for the discussion of factors influencing the selection of relative pronouns and deals with a number that will be considered in the subsequent analysis. However, as its focus is

Rel at ives

89

mainly typological and its discussion of variability is based on intuition rather than empirical study,4 I will present the work of other researchers who have looked into relative pronoun selection. There is a considerable body of literature on relative pronouns studied using variationist methods. These studies consider relative pronouns in terms of native speakers from a number of different regions, including North American English (Ball, 1996; D’Arcy & Tagliamonte, 2010; Guy & Bayley, 1995), early African American English (Tottie & Harvie, 2000; Tottie & Rey, 1997), educated British English (Quirk, 1957), ‘Dorset’ English (Van den Eynden, 1993), varieties of northern British English (Tagliamonte et al., 2005), non-native English (Flanigan & Inal, 1996; Karstadt, 1996) and second generation Mexicano-English (Bayley, 1999). The methodology of some of these studies makes them less directly comparable to others; however, it is useful to mention them all here to underline the extent of work that has been done on this feature in English, even if it is only the studies which have analyzed their data in a manner that can be used to establish the distribution of forms according to the various constraints on use in terms of restrictiveness, syntax type and animacy which will be considered in more depth. An analysis of the findings of previous studies is particularly useful in establishing which factors have been shown to influence (or potentially influence) the choice of relative pronouns. All the factors examined in the present analysis have been found to be significant in other research, which means the extent to which the non-native speakers are comparable to native speakers is verifiable. Before going into detail about which factors were previously found to have an effect on relative pronoun choice, I will first discuss what was excluded from the analysis, as this too is tied to previous findings. Only the relative pronouns which can (under the right circumstances) be substituted for one another were considered in this analysis. The pronouns which fulfil this requirement are who, (whom), which, that and the zero form. Adverbial relative pronouns such as where, when, why have been excluded from consideration as they cannot be as readily exchanged with the other variants.5 Furthermore, because non-restrictive relative clauses do not allow for the same range of pronouns, they were also excluded.6 Both were excluded from analysis in a number of previous studies (Guy & Bayley, 1995: 150; Tagliamonte et al., 2005: 85, for example) so this will make it easier to compare the present results with them. The factors discussed in the introduction to relative clauses, i.e. those placing prescriptive rules on pronoun selection, will need to be considered by examining the results of each category separately. The main factor groups which restrict the pool of potential relative pronouns are: (1) animacy (2) the function of the relative pronoun (Quirk et al., 1985: 1248)

90

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Beyond the prescriptive restrictions detailed already, these two factors affect which variant is more likely to be used.

Animacy Beyond restricting some pronouns to animate relative clauses and others to inanimate ones, animacy can affect which pronouns are more or less likely to be selected: previous research has found that animate subjects are more likely to occur with the wh- form than with that or zero (Guy & Bayley, 1995; Quirk, 1957; Tagliamonte et al., 2005: 17). This was broken down as human for individuals or groups of people (Example 33) and non-human for everything else in the present analysis. (33) In particular the people who will come to Denmark. (b, Italian, email)

Syntactic function The type of relative clause also restricts the variants that are allowed, with subject and object relatives being the main categories. Because the main distinction in terms of syntax and relative pronoun selection is a twoway division between subject and object and the data sample overwhelmingly contained subject relatives, I chose not to break down the various types of object relatives further within this factor group. Object relative types were considered, however, in an additional factor group (see Type of object relative, below). Syntactic function variably constrains pronoun choice as follows ‘when the antecedent is personal and the pronoun is the subject of the relative clause, who is favoured, irrespective of style and the occasion’ (Quirk et al., 1985: 1250).

Type of object relative This factor group distinguished the type of object relative: direct object relatives (Example 34), prepositional object relatives with stranding (Example 35) and prepositional object relatives showing a pied-piped construction (Example 36). (34) I won a flight which I cannot refuse! (d, German, email) (35) write a kind letter to one of the profs Ø you are in good contacts with at the hospital. (b, Italian, email) (36) I need deans, professors, doctors to whom you have a good contact. (c, Italian, email) Stranded and pied-piped relatives do not allow the same relative pronouns. Furthermore, as noted by Quirk et al. (1985: 1251), when the antecedent is

Rel at ives

91

personal but the object of a verb or prepositional complement ‘there is a much stronger preference for that or zero, perhaps to avoid the choice between who and whom’.

Definiteness This factor focuses on the definiteness of the antecedent NP head, whether it is a definite noun phrase (Example 37) or an indefinite noun phrase (Example 38). (37) here is the address of the man who was in charge of organizing. (b, Italian, email) (38) How we must speak with someone who wants to stop smoking. (f, French, email) Van den Eynden (1993: 113) notes that ‘the restrictive wh- and th- forms occurred most frequently with indefinite antecedents’ and that ‘zero, on the other hand, occurred exclusively in definite contexts’. The coding followed Van den Eynden’s division, whereby ‘nouns modified by the definite article, nouns preceded by a demonstrative, possessive, or interrogative (what/which) determiner, proper names, and personal, demonstrative, possessive and interrogative pronouns’ (Van den Eynden, 1993: 116) were considered to be definite, whereas ‘nouns modified by the indefinite article, the zero article, by a cardinal number, or by a quantitative determiner, and indefinite or quantifier pronouns’ (Van den Eynden, 1993: 116) were considered indefinite.

Adjacency A number of studies found adjacency to be important for the selection of relative pronouns. The zero form has been found to be more likely to occur in adjacent position than in other positions. Indeed, Quirk et al. (1985: 1252) stated that ‘when complex phrases or clauses intervene between the antecedent head and the relative pronoun, which is generally preferable to that and very much preferable to zero’ and Guy and Bayley (1995) also found this factor to be significant in their analysis. This factor was divided into three groups, following Guy and Bayley (1995): relative pronouns adjacent to their antecedents (Example 39), relative pronouns separated from their antecedents by another relative clause (Example 40) and relative pronouns separated from their antecedents by a phrase other than a relative clause (Example 41). (39) If anybody knows somebody who is keen in design. (b, Italian, email) (40) for those who are arriving on the 3. or 4. march in malta. (or who are staying longer) (h, German, email)

92

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

(41) She also adviced to have a nurse or somebody working with older people that can bring us. (b, Italian, email) The final factor considered is register: Quirk et al. (1985: 1252) state that ‘all things being equal, more informal discourse will tend to have a preference for zero’. Moreover, the wh- forms are more commonly used in writing than orally (for example, Tagliamonte et al., 2005: 87 find that the wh- forms are virtually inexistent in their corpora of vernacular dialects). The non-native tokens for this feature come from the email corpus, so the native email corpus is the main point of comparison to ensure true comparability. Although the tokens from the IFMSA oral corpus were extracted as well, there were not enough of them to enable us to compare the two registers adequately and thus they will not be considered in this analysis.

7.5 Analysis and Results The extraction of the native and non-native email corpora yielded over 800 tokens. As can be seen in Table 7.5, each group has at least 100 tokens of various relative pronouns. The overall number of tokens per group does not reveal much in terms of how relative pronouns are selected. Table 7.6 provides the overall distribution of the four most frequent relative pronouns in the data by language group. The fifth relative pronoun considered in this analysis, whom, did not appear in all the language groups and was used rather infrequently so is not included here; however, it will be dealt with in the discussion of Table 7.5 Number of relative pronoun tokens by native language English French German Italian Total

266 149 109 289 813

Table 7.6 Overall distribution of the most frequent relative pronouns in data by language group (%)

English French German Italian

who

that

which

zero

Total N

41 43 41 40

26 24 14 29

6 6 17 9

26 27 28 24

266 149 109 289

Rel at ives

93

Figure 7.1 All relatives

object relatives.7 Figure 7.1 presents the same data in graph form to make them more readily interpretable. All four groups use who most frequently and, except for the German speakers, which least frequently, and so are similar in terms of frequency of use. The distribution of relative pronouns is highly constrained in terms of clause type and in terms of the animacy of what is relativized, so these will need to be examined to see if the similarities are borne out. It is only when these effects have been disentangled that it will be possible to determine how similar the native and non-native groups really are. Before turning to this, however, learning-related aspects of relative pronoun selection need to be examined. If the results reveal that a large proportion of the data in the analysis is different from Standard English, then this would significantly affect the results.

Learning-related variation The analysis revealed that there were in fact very few cases of LRV in the Swiss data; only 11 in out of a total of 547 tokens (2% overall). Six tokens were found where which was used with animate subjects (all in the output of Italian speakers; Examples 42 and 43), while five tokens of who were used with inanimate subjects (three by a French speaker, two by an Italian speaker; Example 44). (42) A NEO is not only a responsible person, but also one which regularly gives time to the associations, that knows. (c, Italian, email) (43) There where a lot of new Members, especially from Asia, Eastern Europe and South America, which had to be introduced into the. (c, Italian, email) (44) Ex. I forgot to mention that the magical letter who went directly into the hearts of the director . . . (b, Italian, email)

94

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

The initial analysis found another relative pronoun that created a considerable amount of difficulty for the non-native speakers. Whose was not selected as one of the relative pronouns under study because it did not occur frequently enough in the data and also because it was not fully interchangeable with the other variants. There are, however, a number of points to be made in terms of its non-native use. In fact, it is more a case of non-use in terms of the non-native speakers, as the extraction did not yield a single token of whose for them, whereas 13 tokens were found in the native data. It is difficult to establish if it is a case of the non-native speakers avoiding the use of a form they felt uncomfortable with or whether more data, or different data, would have produced more tokens of it. A few of the sentences discarded from the analysis, however, showed instances of sentences where a native speaker would have used whose but where the non-native speakers used another strategy (Examples 45–46), so it is likely they did not know how to use it in a native English manner. (45) The only way to look for a computer which address is not indexed / known (such as yours) is, when you are. (j, French, email) (46) the person where he live phone me to say she doesn’t want to lodge him. (f, French, email)8 These non-native tokens represent only 2% of the overall tokens of relative pronouns in the data, which may explain why it is not often a feature considered by SLA researchers. From a purely prescriptive perspective, the nonnative speakers have acquired the feature natively. It is only a more in-depth analysis which will be able to reveal how similar or how different the native and non-native groups are, however. The factors found to affect native variability will be considered in turn below, considering first the prescriptive rules which dictate which pronoun is acceptable in a given circumstance (i.e. the factor groups dealing with animacy and relative clause syntax) and then the variable rules which have been found to influence relative pronoun choice (adjacency, definiteness). The few tokens of non-native-like use have been left in the analysis as this may make it possible to uncover what factors triggered their use and the low number of them will not affect the results to a great extent. Animacy will be considered within the analysis of subject and object relatives as they are interlinked with respect to the range of relative pronouns acceptable. Subject relatives allow only who and that for animate and which and that for inanimate, while object relatives allow who(m), that and zero for animate and which, that and zero for inanimate. For both subject and object relatives, all tokens of each type will first be considered together, and then animate and inanimate subject/object relative are considered separately. This will allow me to establish in what ways the linguistic groups are similar.

Rel at ives

95

Subject relatives Subject relatives make up 63% of the overall tokens of relative pronouns in the data (513 out of 813). Table 7.7 and Figure 7.2 provide the overall distribution of subject relatives by native language. The French, Italian and English groups share the same hierarchy (who > that > which), whereas the German group use which slightly more often than that. Like the other three groups, the German speakers have who as their main form. Because the relative pronouns used in subject relative clauses are closely linked to the animacy of the subject, human and nonhuman subjects have to be disentangled before assessing how similar the groups are. The following tables and charts consider this, looking first at human subjects (Table 7.8 and Figure 7.3) and then at non-human subjects (Table 7.9 and Figure 7.4) to determine whether the surface differences in distribution between the groups are reflected in the more detailed analysis. Differences in these tables will be a far more accurate way of judging whether the native and non-native groups have substantially different processes to deal with the Table 7.7 Overall distribution of subject relatives in data

English French German Italian

who %

that %

which %

Total N

63 69 62 63

29 24 18 24

8 7 19 12

172 91 72 178

Figure 7.2 Subject relatives only

96

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Table 7.8 Overall distribution of subject relatives in data (human only)

English French German Italian

who %

that %

which %

Total N

87 94 88 87

13 6 12 8

– – – 5

126 63 51 129

Figure 7.3 Subject relatives (animate) Table 7.9 Overall distribution of subject relatives in data (non-human only)

English French German Italian

that %

which %

who %

Total N

72 64 33 65

28 21 67 33

– 14 – 2

46 28 21 49

relative pronoun choice than the first tables (such as Table 7.5) presented in the analysis. Except for the tokens of which found only in the Italian speakers, and which are due to LRV, the patterns of the four groups are quite similar; in all four groups who is predominantly used and that is used to a much lesser extent. As with the human subject relatives, there are a few tokens of non-native forms (who) used by the French and the Italian speakers, but what is most striking about these figures is the difference between the German speakers

Rel at ives

97

Figure 7.4 Subject relatives (inanimate)

and the other groups. While the native speakers and the other non-native groups use that as the main pronoun for non-human subjects, the German speakers are far more likely to use which. Indeed, in terms of actual percentages, the tendencies are reversed, in that the native English speakers use that around 70% of the time and which the remaining 30%, and the German speakers use which 70% and that 30%. It might be due in part to the lower number of tokens found for the German speakers (21 versus 28 and 49 for the other two non-native groups), but it is nonetheless an intriguing finding and I will return to this later.

Object relatives As mentioned already, the object relatives occur less frequently than subject relative clauses, and make up only 37% of the total number of relative pronouns (300 out of 813). This is fairly typical; Tagliamonte et al. (2005: 89) found object relative clauses made up 33% of their data. Table 7.10 and Figure 7.5 demonstrate that there are fewer differences between language varieties in terms of hierarchies for the object relatives than the subject relatives. The English, French and Italian groups all show the same hierarchies of zero > that > which > (whom) and while the German speakers reverse that and which, the percentage difference between the two is not as considerable as in the case of the inanimate subject relatives and they, like the other three groups, have zero as the main form. The low use of that for the German speakers is similar to the findings for the subject relatives, and shows that the German speakers are avoiding its use and are instead going for the next

98

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Table 7.10 Overall distribution of object relatives

English French German Italian

zero %

that %

which %

whom %

Total N

75 69 84 61

21 24 5 30

4 5 11 5

0 2 0 4

94 58 37 111

Figure 7.5 Object relatives

most frequent option. In the case of object relatives, the zero form is used overwhelmingly by the German speakers (84%), far more than by the other language groups. The high use of the zero form across linguistic groups is interesting from a language acquisition perspective, as none of the source languages has a zero form comparable and it underlines the extent to which the non-natives are following the native patterns. The use of whom by the non-native speakers is also noteworthy; the native tokens revealed no cases of this oblique form, whereas the non-native data provided one token for the French speakers and three for the Italian speakers. This demonstrates that the non-native emails are partially closer to standard written English, but overall the number of tokens for human object relatives is extremely low (around 20 tokens across the four groups) and this is a factor as well. Because there are only 21 tokens across the four linguistic groups of human object relatives, the results for the two types separately will not be presented as individual patterns of distribution as they could not be determined. What have the results thus far shown us? Although the variable rules governing relative pronoun selection were not explicitly taught to non-native

Rel at ives

99

speakers, they nevertheless adhere to them to a great extent. The following results underline this: • • •

for human subject relatives all four linguistic groups share the same who > that pattern; for non-human subject relatives the English, French and Italian groups share a that > which pattern; for object relatives all the groups have zero as the main variant, followed by that then which for the English, French and Italian groups and which then that for the German group.

These results demonstrate that the non-native groups are very similar to the native group and most importantly share the hierarchies and percentages of the native control group. This is very different from the results found for the future, where going to was discarded and will was essentially the only form in use for all three non-native groups. While at this stage it is not possible to determine whether the variable rules are exactly the same, this has shown that the patterning is in the same direction as native speakers. For the most part, the non-native speakers use the relative pronoun variants in a similar way to native speakers. There are of course a few deviations from the native norm, with unsuitable clauses used for animate or inanimate objects, but these represent a very small percentage of the overall number of tokens. Although the German speakers use some of the forms (which) differently from the other two Swiss groups and the native group, by and large the main form is the same everywhere; furthermore, the main divisions, object versus subject relatives, are the same. The differences found in the German speakers are difficult to account for at this stage; they could be due to transfer from their mother tongue or they could be linked to the factors which have not been examined yet. I will come back to this in the discussion section. How do these results compare to previous studies? Ball (1996) provides a summary table of the results of subject relative pronoun choice for a number of earlier studies of Standard English (British and American), divided into human and non-human subjects. Comparing the results from the email data collected for this study to the results from other studies will allow for a greater measure of verifiability to help establish the extent to which the nonnative speakers do indeed share the native patterns. As discussed before, although the native emails were collected to form a control group, as a counterpart to the non-native email corpora, non-native speakers would have had the most contact with Standard English orally or through formal writing. It may turn out that the few differences which were found between the native and non-native groups are linked to the fact that the native corpus is comprised of emails and that the non-native results are in fact closer to native speech.

100

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Table 7.11 Human subject relative pronouns: comparison of Ball (1996) and present data who %

that zero % %

other %

Total N

98 92 81

1 8 19

– – –

– – –

50 50 114

Standard British English

1970s Ball (1996) 1970s Ball (1996) 1970s Guy & Bayley (1995) 1950s Quirk (1957)

91

9

0.45

–

222

British native email Non-natives French German Italian

2000 2000 2000 2000 2000

87

13

–

–

126

94 88 87

6 12 8

– – –

– 58 – 51 5 (which) 129

Watergate Hearings Terkel MC interviews Nixon

Table 7.11 presents the results for human subject relatives, while Table 7.12 provides those for non-human subject relatives, considering both American data from the 1970s and British data from the 1950s. All four studies discussed by Ball come from Standard, middle-class varieties of English, most likely closely approximating what the Swiss students would have been exposed to during their acquisition of English. There are no major differences between the email and oral results in terms of human subject relative pronouns; in all cases, who is the variant Table 7.12 Non-human subject relative pronouns: comparison of Ball (1996) and present data which that zero % % %

other %

Total N

1970s Ball (1996) 1970s Ball (1996) 1970s Guy & Bayley (1995) Standard British English 1950s Quirk (1957)

24 12 13

76 84 87

– – –

– – –

50 50 130

48

52

0.33 –

304

Native email Non-natives French German Italian

28

72

–

–

46

21 67 33

64 33 65

– – –

14 (who) – 2 (who)

28 21 49

Watergate Hearings Terkel MC interviews Nixon

2000 2000 2000 2000 2000

Rel at ives

101

used most frequently. In terms of the non-human subject relatives, the hierarchy is the same for all the groups except the German emailers: that is the primary variant with which as the alternate, but there are greater divisions in terms of percentages than for human subject relatives. Leaving the German speakers aside, that is selected between 52% and 87% depending on the group under study. In terms of the native speaker groups, it could be seen to represent a change in progress as the results span nearly 50 years. Quirk’s results show the highest use of which (48%), while the native British emails collected in 2000 and later show only a 28% use of which. Backing this hypothesis up further, Tagliamonte et al.’s (2005) research found the wh- forms to be almost totally absent from their data. Because they had focused primarily on relic varieties, I have chosen not to include their results in the comparison with Ball’s research. The non-native speakers are approximating current relative pronoun patterns, not those of 50 years ago. The native results also demonstrate differences between American and British English patterns as the three American studies show more that use than the later British email study. The French and Italian distribution is very similar to the distribution in the native groups. Having compared the native and non-native groups’ distributions in terms of animacy and relative clause type, the focus will now turn to the factors of adjacency and definiteness to establish whether the native patterns are replicated by the non-native speakers here as well. Note, however, that object and subject relatives are considered together for these two factor groups and this may partly obscure what is occurring. They must be considered together at this stage to avoid getting extremely low token numbers in some of the cells, which would make it difficult to establish any patterns. The later multivariate analyses will determine the relative strength of the different factors, however.

Adjacency Tables 7.13–7.16 present the distribution of the three variants in terms of whether the relative pronoun is adjacent to the subject or separated by another relative clause or separated by something else. Each table considers a linguistic group separately. Previous studies have found that the zero form is far more likely to be found in an adjacent position than elsewhere.

Table 7.13 French relative pronoun distribution by adjacency

Adjacent Sep. relative Sep. other

% of wh-

% of that

% of zero

Total N

46 50 68

23 50 32

31 0 0

128 2 19

102

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Table 7.14 Italian relative pronoun distribution by adjacency

Adjacent Sep. relative Sep. other

% of wh-

% of that

% of zero

Total N

51 57 39

21 43 55

28 0 5

237 14 38

Table 7.15 German relative pronoun distribution by adjacency

Adjacent Sep. relative Sep. other

% of wh-

% of that

% of zero

Total N

55 100 67

14 0 13

30 0 20

92 2 15

Table 7.16 English relative pronoun distribution by adjacency

Adjacent Sep. relative Sep. other

% of wh-

% of that

% of zero

Total N

46 25 61

26 50 26

28 25 13

239 4 23

The results for both adjacency and definiteness make a three-way distinction of the relative pronouns, combining who and which into a single form. This is to partly conflate the animacy distinction for this feature and also because it is how a number of earlier studies examined relative pronouns and as such this allows for an easier comparison. All four tables reveal, first of all, that zero relative pronouns are found more frequently in an adjacent position than when separated from the main clause. Indeed, there are no tokens of zero in non-adjacent positions whatsoever for the French speakers. Although the English, Italian and German speakers have some instances of zero in these positions, it is still lower than for adjacent tokens. The low number of tokens separated by another relative clause in all four groups means that for the multivariate analysis there will be only a two-way distinction: adjacent or non-adjacent. These results match what had been predicted by previous studies, i.e. the zero form is most likely in adjacent positions. The non-native speakers have acquired this constraint.

Definiteness Tables 7.17–7.20 present the three variants in terms of definiteness; unlike adjacency, the distribution of tokens between definite and indefinite relative pronouns is better balanced.

Rel at ives

103

Table 7.17 French relative pronoun distribution by definiteness

Definite Indefinite

% of wh-

% of that

% of zero

Total N

37 60

21 27

42 13

71 83

Table 7.18 Italian relative pronoun distribution by definiteness

Definite Indefinite

% of wh-

% of that

% of zero

Total N

41 68

27 25

32 7

193 96

Table 7.19 German relative pronoun distribution by definiteness

Definite Indefinite

% of wh-

% of that

% of zero

Total N

49 70

9 20

42 9

65 44

Table 7.20 English relative pronoun distribution by definiteness

Definite Indefinite

wh-

that

zero

Total N

41 55

25 28

34 18

146 120

The four tables show remarkably similar hierarchies comparing definite to indefinite relative clauses: • •

for all four linguistic groups there is a higher rate of wh- forms in indefinite contexts than in definite ones. the zero variant occurs far more frequently in definite context than indefinite ones for all four groups.

Both of these results are in line with what has been found in earlier studies on native relative pronoun distribution (Tagliamonte et al., 2005; Van den Eynden, 1993) as they, too, had found that definite contexts favoured the use of the zero variant. In terms of the factors of adjacency and definiteness, the overwhelming impression is again that the non-native groups are similar to the native control group (and the native speakers considered in previous studies) and that they are following the variable rules of relative pronoun selection that had been uncovered in previous studies.

104

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

7.6 Multivariate Analysis Having examined the factors affecting the use of relative pronouns individually, this section will focus on the combined effect of the factors together in order to disentangle any interaction there might be and uncover which factors can be shown to play a significant role in the occurrence of the variants. To establish what the variable rules of each linguistic group are, a separate multivariate analysis has to be conducted on each group and then the four analyses are compared to one another. This will allow me to examine whether they share hierarchies of constraint and ranges and ultimately to determine how similar their variable rules are. I have followed Guy and Bayley’s method for the division of relative pronoun types for my own multivariate analysis; this involves categorizing relatives into one of three groups: wh-clauses, that and zero. By considering who and which together we will be able to examine the factor of animacy across all the variants. Like them and rather differently from later studies such as Tagliamonte et al. (2005), I will be considering subject and object relative clauses together in the runs examining wh- and that. This will allow me to determine whether the non-native groups share the same constraints in terms of syntax as the native group for the variants which are used both with subject and object relative clauses. When considering the factors affecting the use of zero I will only take object relative clauses into account as zero relative pronouns do not occur in subject relatives in Standard English. The constraints governing the selection of wh- forms, that and zero forms are considered in the three tables below, each table presenting four multivariate analyses, one for each linguistic group. Each variant will be discussed on its own and Section 7.7 will focus on an overall discussion into how native variability compares to what was found for the non-native speakers. Table 7.21 presents four independent multivariate analyses of factors contributing to wh- forms. Syntax of relative clause, animacy, adjacency and definiteness were considered in all four analyses. As noted earlier, for adjacency, tokens were grouped into adjacent and non-adjacent. In a multivariate analysis, the distribution of factor groups has to be ‘orthogonal’ (Guy, 1988: 126–127), ‘this means that the factor groups should be independent of each other’ (Tagliamonte, 2006: 181).9 Although some degree of overlap (i.e. interaction between groups) is acceptable, ‘some distortion of the results is probably occurring’ if this overlap is above 90% (Guy, 1988: 131). Earlier it was noted that there were very few human object tokens in the data and crosstabulations of the two factors revealed that the overlap between them was above 90% in all four linguistic groups. Consequently, the analyses below will combine the factors of syntax and animacy, giving me a single factor group that considers human subject, non-human subject and object relatives separately.10

Rel at ives

105

Table 7.21 Four independent multivariate analyses into the probability of whEnglish Input

0.41 FW

Syntax and animacy Subj/human 0.90 Subj/non-human 0.37 Object 0.06 Range

French

German

0.50 % N FW 47 266

0.59 % N FW 49 149

87 126 0.94 46 0.35

36

28 4

84

94 0.07 87

94 7

63 0.84 28 0.58 58 0.08

Italian 0.52 % N FW % N 58 109 50 289 88 51 67 21 11 37

76

Adjacency Adjacent

[0.50] 46 239 [0.46] 46 128 [0.44] 55 92

Non-adjacent

[0.51] 56

27 [0.74] 67

21 [0.78] 71 17

[0.47] 41 146 [0.42] 37

71 [0.45] 49 65

Indefinite article [0.54] 55 120 [0.57] 60

78 [0.57] 71 44

Range Definiteness Definite article Range

0.92 92 129 0.35 35 49 0.07 8 110 85 0.55 51 237 0.27 44 52 28 0.41 41 193 0.67 68 96 26

Before proceeding, I will reiterate how the results are interpreted as this is the first feature to use multivariate analysis. Examining the results factor by factor will allow me to determine which factors favour the use of wh(those with weights over 0.50) and which disfavour wh- (those with weights under 0.50). The input represents the ‘overall tendency of rule application’ (Tagliamonte, 2006: 156). In Table 7.22 the inputs show that all the groups except the English speakers favour wh- forms as they are above 0.50. Having examined the factors individually, I will then use the ‘three lines of evidence (statistical significance, constraint ranking and relative strength of factors)’ (Tagliamonte, 2006: 245) to establish whether the four groups share the same patterns. The factor groups which were found not to be statistically significant are given in square brackets. Syntax and animacy is the only significant factor for three of the groups and the most significant factor for the Italian speakers. All four groups show very high ranges, which underlines how important this factor is in the conditioning of wh-. The constraint ranking is identical in the four groups, with human subject tokens very strongly favouring wh- forms, followed by non-human subject tokens and finally object tokens which very strongly disfavour wh- forms. Non-human subject tokens favour wh- only in the German multivariate analysis; this is likely due to the higher overall rate of wh- in their data. As the two factors were combined, the results are not

106

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Table 7.22 Four independent multivariate analyses into the probability of that English Input

0.24 FW

Syntax and animacy Subj/human 0.34 Subj/non-human 0.89 object 0.47 Range 58 Adjacency Adjacent [0.49] Non-adjacent [0.55] Range Definiteness Definite [0.51] Indefinite [0.49] Range

French

German

Italian

0.19 % N FW 26 266

0.11 % N FW 24 149

0.20 % N FW 14 109

% N 26 289

14 126 0.23 72 46 0.89 21 94 0.58 66

6 64 24

12 33 5

51 0.23 21 0.87 37 0.65 64

9 129 65 49 29 110

26 239 [0.50] 23 128 [0.51] 14 30 27 [0.53] 33 21 [0.46] 12

92 0.43 17 0.80 37

21 237 52 52

25 146 [0.44] 21 28 120 [0.56] 27

65 [0.50] 27 193 44 [0.51] 25 96

63 0.51 28 0.80 58 0.31 49

71 [0.44] 9 78 [0.59] 21

completely comparable to previous studies but to a great extent they follow the expected constraint ranking; subjects favour wh- more than objects and animates favour it more than inanimates (Guy & Bayley, 1995). The nonnative groups are very similar to the native group in terms of this factor. The other factors considered in this analysis are not as important, but nevertheless show further similarities between the native and non-native groups. These factors were not found to be significant for the English, French and German groups, while they were for the Italians; this is likely due to the fact that the Italian speakers had a higher number of overall tokens. It is by examining the constraint ranking, alongside the aspects just examined, that I will be able to establish how similar the non-natives are to the native speakers. In terms of adjacency, previous studies had found that non-adjacent tokens were more likely to be used with wh- forms than adjacent tokens (Guy & Bayley, 1999: 154). This is borne out by the English, French and German groups, where non-adjacent tokens favour wh-, while adjacent tokens do not.11 This factor was not found to be statistically significant for any of them, however. The run considering the Italian speakers, on the other hand, found this factor to be significant and that the distribution was reversed; adjacent favours wh-, whereas non-adjacent disfavours it. Although definiteness was only found to be significant for the Italian speakers for the selection of wh-, the hierarchies are identical in all four

Rel at ives

107

groups, as had been predicted by Quirk et al. (1985). Indefinite articles are more likely to be found with wh- relative clauses than definite articles. How do the groups compare overall in terms of wh-? Although two factor groups are found to be significant only for the Italian speakers, the factor group combining syntax and animacy was significant in all groups. Moreover, it shows very high ranges in all four groups and an identical constraint ranking (subj/human > subject/non-human > object). The hierarchy is also the same in terms of definiteness: wh- is favoured more with indefinite than with definite. The factor of adjacency shows differences between the Italian speakers and the other three groups; unlike the other groups, Italian speakers favour wh- in adjacent contexts. This is the main difference between the groups. The overall impression is then one of similarity; the non-native groups share the patterns of the native groups. Turning now to that, Table 7.22 considers the factors found to be significant in its selection. The inputs of the four groups show that that is quite strongly disfavoured; this is to be expected given the high rates of who found in all the speaker groups. As for wh-, only one factor group is statistically significant in all four linguistic groups – the one combining syntax and animacy. The German groups have a slightly different constraint ranking from the three other groups, however. The native speakers and the Italian and French groups show that human subject contexts are those most disfavoured, while non-human subject are most favoured with that (subj/nonhuman > object > subj/human). The German speakers favour non-human subject most strongly but this factor is then followed by human subject and not object contexts (subj/non-human > subj/human > object). Adjacency is statistically significant only for the Italian speakers. The expected hierarchy, whereby non-adjacent contexts are favoured with that, is found for the English, French and Italian groups. The German speakers have the opposite hierarchy. Although definiteness was not found to be significant in any of the four languages, the factor weights in the English and Italian groups are very close to each other (slightly above or below 0.50). The French and the German groups favour that in indefinite contexts. As with wh-, the patterns of the native English group are closely matched by the non-native groups in terms of that, demonstrating again that the non-native speakers mirror the variable rules of native English speakers. The German results show slightly different hierarchies from the other three groups, but the overwhelming impression is one of similarity. Again, the non-native speakers have acquired the patterns of native English speakers. Like the native speakers, the non-natives show that the combined group of syntax and animacy most strongly affects the use of that, and that within this group subject inanimates contribute to the occurrence of that the most. The factor of adjacency shows us that the native, French and Italian groups favour that more in non-adjacent contexts. The few

108

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

differences found for the German group may be partly due to the fact that they have fewer tokens overall. Table 7.23 presents the final set of multivariate analysis runs, focusing on the constraints on the use of zero as a relative pronoun. Because zero can only occur in object relative contexts, all subject relatives were removed from these analyses. As a consequence of this, the number of tokens is quite low, which means that in the case of the French and German speakers multivariate analyses could not be run. With only 58 and 37 tokens, respectively, the runs would have been misleading. Their marginals will be compared to the English and Italian speakers’ multivariate analyses. In object relative contexts, the zero form is selected most frequently; it is highly favoured in the input of all four groups and represents between 63% and 84% of the object relative tokens. Animacy was not found to be significant in the two groups where it was run and the constraint ranking is not shared. The English speakers favour animates more strongly than inanimates, whereas this hierarchy is reversed in the Italian group. Because of very low tokens for human objects, the pattern of the French and German speakers cannot be established. Adjacency was found to be statistically significant for the Italian speakers but not for the English speakers. All four groups seem to share the same constraint ranking, however, as adjacent tokens favour zero and non-adjacent tokens disfavour it. This follows the results of Guy and Bayley (1999: 154). Table 7.23 Four independent multivariate analyses into the probability of zero English Input

Animacy Human Non-human Range Adjacency Adjacent Non-adjacent Range Definiteness Definite Indefinite Range

0.75 FW

French % 75

N 94

[0.57] [0.50]

80 74

[0.52] [0.32]

[0.53] [0.45]

FW

German

% 69

N 58

5 89

50 70

76 57

87 7

77 70

64 30

FW

Italian

% 84

N 37

2 56

100 82

3 34

74 0

54 4

90 50

81 48

37 21

96 44

0.63 FW

% 63

N 110

[0.41] [0.51]

50 64

12 98

31 6

0.56 0.06 50

67 11

101 9

28 9

[0.60] [0.19]

71 30

87 23

Rel at ives

109

In terms of definiteness, the findings match what had been found by Van den Eynden (1993: 113); in all groups, the zero form is favoured in contexts where there is a definite article and disfavoured when an indefinite article is present. Overall, the groups are not as similar in terms of zero relative pronouns as for wh- and that, although they all have high rates of zero in contexts where it can be used. The differences are due primarily to the fact that in considering only object relatives the runs for zero had a considerably lower number of tokens and the results are less conclusive. There are a number of similarities between the groups, however. All four groups have very high percentages of zero in object relatives and share the same hierarchy for definiteness and for adjacency, and although the hierarchies differ in terms of animacy, it is most likely due to a very low number of human object relative tokens. Although I cannot establish with absolute certainty that the non-native groups fully share the patterns of the native speakers, for the most part the patterns have been acquired.

7.7 Discussion The multivariate analyses for the three relative pronoun variants show a number of similarities and differences between the groups, both in terms of native/non-native but also across the non-native groups. The overwhelming consensus, however, is that the non-native groups are remarkably similar to the native speakers; although there are some differences in terms of constraint ranking and in terms of magnitude of effect, overall the non-native speakers appear to be following the strategies of the native emailers or those found in other native speakers in other studies. The differences can generally be explained in part by uneven data distribution. The multivariate analyses considering wh- and that show far more similarity between the groups than the ones for zero; they also have a much higher overall number of tokens. It is not only in terms of the factors which highly restrict relative pronoun selection that the non-natives match the natives, but also for the factors that are largely unconscious, such as adjacency and definiteness. By and large, the non-native users have the same constraint ranking for these as native speakers. This similarity between the native and non-native speaker groups is not only on the surface, but also in terms of the underlying variability. Not only have the prescriptive rules affecting the distribution been acquired, but the variable constraints have as well.

7.8 Conclusion As well as showing that the native English constraints governing relative pronoun selection have been acquired by the three non-native speaker

110

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

groups, this section has also been able to demonstrate that differences in terms of syntactic structures of the source and target languages do not necessarily inhibit native-like acquisition of variable rules. This is rather different from previous studies and underlines the importance of examining a range of variable features beyond those that are affected primarily by standardness. Furthermore, the fact that much of the variability was not taught to the non-native speakers was found to have no effect on their variable patterns as they match the native speakers. This is considerably different from what was found in terms of the future tenses which were not acquired in a native English-like manner by the non-native speakers. Here the non-native speakers were able to deal with the variability present in native English. By focusing on the target-based variability found in relative clauses rather than looking solely at the learning-related variability, this section has been able to uncover the extent to which unconscious variable rules may be acquired by non-native speakers.

Notes (1) The examples which do not give the speaker code and language were made up for illustration purposes. (2) Studies of spoken data reveal that tokens of pied-piping are very rare; whom accounted for a mere 0.05% of the tokens in Tagliamonte et al. (2005: 87). (3) Note that it does, however, affect the verb form in some tenses, as in the examples in Table 7.2. (4) Although it is reasonable to assume that Quirk’s own empirical research (1965) into relative pronouns may well have been the basis of a number of the comments in the later work. (5) Whose has also been excluded because it cannot be substituted by the other relative pronouns. However, it did appear to present some difficulty for the non-native speakers and will be touched on briefly in the analysis section of this chapter. (6) Although Standard English constrains the use of that with non-restrictive relative clauses, there are some cases of this in the non-native data: – as well as explaining to ‘f’ or ‘g’, that will be in Malta, what to do for you during the SCORP Working Committee (b, Italian, email) This token and ones similar to it were excluded from the analysis as they are nonetheless non-restrictive. (7) This is why the numbers in Tables 7.5 and 7.6 do not match for the French and Italian speakers. (8) It is difficult in this example to ascertain precisely what was meant; however, a native speaker could possibly have used a construction such as ‘the person whose house he lives in’. These tokens were of course not included in the analyses presented here. (9) An example given by Tagliamonte (2006: 182) of non-orthogonality is a case where none of the oldest group of speakers in her sample was highly educated. Education was not considered in her analysis. (10) The log-likelihoods of the runs with the combined factor groups were found to provide a better fit to the data. (11) The range between the two factors is very low for the English speakers.

8

Complementizers

8.1 Introduction1 The third linguistic variable I will consider is the alternation of that or zero complementizers, as in Examples (1) and (2): (1) I hope Ø you enjoyed the day and liked the city and the bears. (b, Italian, email) (2) I hope that in the future in all Switzerland we’ll have some common projects at national level. (f, French, email) Complementizer variation has been the subject of both synchronic (Elsness, 1984; Tagliamonte & Smith, 2005; Thompson & Mulac, 1991; Torres Cacoullos & Walker, 2009b) and diachronic (Rissanen, 1991; Warner, 1982) research on native speaker English. These studies provide an excellent backdrop from which to gauge the use of this variable in non-native speech, specifically with regard to frequencies and constraints on the use of the zero form. There are a number of reasons why this feature will be useful in order to gain an understanding of the non-native acquisition of variation: first of all, unlike some of the other features considered in this book, both complementizer forms are nearly always fully acceptable to native speakers. Consequently, this variable is not given any focus in teaching, so if the nonnative speakers match the native speaker patterns, then it is purely due to their ability to notice and acquire native patterns unconsciously. Secondly, as will be discussed more fully in Section 8.3, two of the source languages do not have a zero complementizer form similar to the English form, which means that any similarities between the native and non-native groups cannot be due to transfer from the source languages. 111

112

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

8.2 Complementizers Syntactic structures, such as those introduced with verbs such as think, say, mean, can either have a that complementizer between the verb and the following clause or zero (as in Examples 1 and 2 above). Previous studies on the topic have reported two main findings. First of all, the use of the zero form has grown through the history of English. Although the zero form was rarely used in Old and Middle English, its use has increased over time, achieving a near categorical use in the 20th century for some specific verbs (Tagliamonte & Smith, 2005: 301; Thompson & Mulac, 1991: 244). Secondly, previous studies have found that in present-day English style exerts a considerable effect on the selection of zero complementizer forms. Elsness (1984: 521) found that in formal writing zero complementizers are used far less frequently than in informal writing. Looking at informal oral data, Thompson and Mulac (1991: 242) found that overall the zero complementizer was used at a rate of 86%, Tagliamonte and Smith (2005: 300) found a rate of 84%, Torres Cacoullos and Walker (2009b: 16) found 79% and Kolbe (2008: 112) found 90%.2 Elsness (1984: 521), looking at a subsection of written data from the Brown corpus, found rates of 52% and 58%, respectively, in his two informal text categories (Press Releases and Fiction: Adventure & Western) and far lower rates of 15% and 1.3% in his two formal text categories (Belles Lettres & Biography and Learned & Scientific Writing) (see Ellegard, 1978; Kucera & Francis, 1967 for further information about the Brown corpus). The non-native speakers of all three linguistic backgrounds use both variants, as can be seen from Examples 3–8 below. French speakers (3) I think that there’s a virus in the past document I send to you. (f, French, email) (4) I will take some copies [. . .], because I think Ø there is a virus. (f, French, email) Italian speakers (5) I guess that there will not be very many new people. (b, Italian, email) (6) I guess Ø you’re back in Switzerland! (b, Italian, email) German speakers (7) So I also think that we need to think carefully about the division of expenses. (h, German, email) (8) I think Ø it is very important to have such a useful booklet. (h, German, email) In terms of the teaching of complementizer forms to Swiss speakers, a survey of the grammar books reveals that the fact that there are two variants is

Complement izers

113

never explicitly made clear. The two forms are used, however, both in the grammar books (Soars & Soars, 1987; Spencer, 1999) and other teaching materials3 the Swiss students use and in the speech of their teachers.

8.3 Complementizers in the Source Languages As in the case of relative clauses, French and Italian do not have a zero complementizer variant. Complementizer forms in these two languages are somewhat similar to relative pronouns, in that the complementizer particle is also the more frequent of the possible relative pronouns (that is to say que for French, and che for Italian) (Examples 9–10). German and Swiss German, however, have both overt and zero complementizer forms (Examples 11–12). Moreover, similarly to English, the dass form is seen to be more formal than the zero form. The variable patterns of complementizers have not been studied in German, so we cannot know if some of the other factors found to be significant in English complementizer use operate in German as well. (9) Je pense que tu as presque fini. (10) Penso che hai quasi finito. (11) a. Ich glaube, dass du schon fertig bist. b. Ich glaube Ø du bist schon fertig. (12) d Ruth glaubt, dass/Ø d Susann het s gmacht. (Penner & Bader, 1995: 103)

8.4 Previous Studies Studies which have focused on present-day varieties of English (Elsness, 1984; Kolbe, 2008; Tagliamonte & Smith, 2005; Thompson & Mulac, 1991; Torres Cacoullos & Walker, 2009b) can help establish how close the nonnative speakers are to native norms. Although studies which focus on the diachronic side of complementizer variation (Rissanen, 1991; Warner, 1982) are useful in terms of establishing the factors that might influence variability, they cannot provide a suitable comparison to the non-native speakers. Tagliamonte and Smith (2005) provide an in-depth presentation and summary of earlier studies. Their analysis focused on the patterns of zero complementizer use of the oldest generation of speakers in several relatively isolated northern British communities (in Cumbria, Lowland Scotland and Northern Ireland). They found very high percentages of the zero complementizer (around 90%) and established that the zero form had become nearly categorical in some contexts (such as I think, I mean) (Tagliamonte & Smith, 2005: 299). In the variable contexts of use, they also uncovered a number of internal factors which conditioned the use of the variants; first and second

114

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

person subjects versus other subjects (Examples 13–14), present tense versus past tense (Examples 15–16), additional elements in the verb phrase (VP) (Example 17) and finally adverbials between the VP and complementizer (Example 18) (Tagliamonte & Smith, 2005: 301). (13) I guess Ø I will be the only one from Bern who will join the meeting. (m, German, email) (14) he explain that there will be a meeting in Geneva. (f, French, email) (15) I think Ø it would be a great opportunity also to show what is going on in the other cities. (b, Italian, email) (16) But he told us, that he can speak German! (r, German, email) (17) I don’t think Ø I’m talking about that girl from Catalan country. (h, German, email) (18) I really hope that everybody is feeling fine. (a, Italian, email) Formality also exerts a considerable effect on the use of the complementizer variants. Tagliamonte and Smith (2005: 300) and Thompson and Mulac (1991: 242) found very high rates of zero complementizer in their analyses which considered varieties of informal spoken English. On the other hand, Elsness (1984: 521), who examined written data, found low rates of the zero form in the more formal texts. He found higher rates of the zero form in the less formal texts but these rates were still rather lower than the oral data in Tagliamonte and Smith (2005) and Thompson and Mulac (1991). These effects of formality and of register will need to be taken into account when considering the email data in this analysis. The place of email on the continuum between oral and written data is debatable (see Herring, 2001) so we cannot know a priori if the rates for the zero complementizer in our email data will be high, as in spoken English, or low, as in formal written English. The native email control data will play a crucial role here as it is directly comparable to the non-native data. If the native data show high levels of zero complementizer then we would expect high levels from the non-natives as well, if we are to prove that the non-native speakers pattern like native speakers. On the other hand, if the native data show low levels, then we would expect the same from the non-natives.

8.5 Analysis and Results The present analysis will focus on comparing the two email corpora and not consider the oral non-native data because, as shown above, there are considerable differences between oral and written uses of complementizers and the tokens of the oral section of the corpus were too unbalanced in terms of language group. It will thus be more revealing to establish whether there are any comparable patterns between the native and non-native speaker

Complement izers

115

groups. For the analysis, every instance where either that or zero could have been used was extracted from the data.4 The factors which were coded in this analysis and which will be studied in detail are very similar to those examined by Tagliamonte and Smith (2005) and Torres Cacoullos and Walker (2009b). As well as speaker and speaker’s native language, which are the external factors in this analysis, I will focus on subject of the matrix clause, the tense of the verb, whether there are any additional elements to the verb phrase, whether the VP and the complementizer are separated by adverbs or adverbials and finally the lexical verb which the complementizer follows. Previous research on complementizers found first and second person subjects have a higher proportion of zero complementizers than third person subjects. The tokens have been coded for whether they are first person singular (19) or plural (20), second person (singular or plural) (21), third person pronouns, he, she, it and they (22–24) or third person noun phrases (singular and plural) (25). Because of a low number of tokens for some of these factors, all the first person subjects are later combined into a single group. In the multivariate analysis, all non-first or -second person subjects are combined regardless of whether they are pronouns or noun phrases. This was possible because these subjects had been shown to pattern in a similar manner. (19) I guess that I still the NORE5 for this year. (l, French, email) (20) We knew that X didn’t speak German. (o, German, email) (21) Do you also think that we should buy a Firewall program? (c, Italian, email) (22) But he told us, that he can speak German! (r, German, email) (23) it’s about time that things get clear in this meeting story. (&, French, email) (24) they told me that they can give us some sample materials. (b, Italian, email) (25) Young teenagers (girls) think Ø their physical appearance is very important. (f, French, email) The tense of the matrix verb was coded with a three-way distinction, verbs in the present tense (29), verbs in the past tense (30) and sentences with no verb (31). Tagliamonte and Smith (2005: 304) had found that verbs in the present tense favoured the zero complementizer more than past tense verbs. There are relatively few tokens of sentences with no verb, so these will be excluded in the multivariate analysis. (26) I think also that X is in Geneva, isn’t it? (w, French, email) (27) In the meantime I found that he wrote his notes for fundraising on the IFMSA web page. (c, Italian, email)

116

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

(28) The fact that the computer is not always on minimizes greatly the possibilities for anyone to access it. (j, French, email) In terms of additional elements in the VP, the tokens were coded for whether there were no additional elements (29), whether the additional element was a modal (30) or whether the additional element was a negation form (31). The tokens which had both a modal and a negation form were given a separate code (32). Tagliamonte and Smith (2005: 304) found that ‘simpler constructions’ (i.e. those without additional elements in the VP) favoured the zero complementizer and Torres Cacoullos and Walker (2009b: 21) found the same. (29) I think Ø you’ll understand why. (c, Italian, email) (30) To sum it up it can be said that SCOME-CH has to build a solid structure for concrete projects. (f, French, email) (31) I don’t think Ø it would be necessary to buy a multi-user license. (j, French, email) (32) I cannot promise that I can attend. (*, French, email) Like the previous factor, the tokens in the factor considering other additional elements in the matrix clause were coded for absence (33) or presence (34) of additional elements. Again, Tagliamonte and Smith (2005: 304) found that tokens without additional elements favoured the zero complementizer. (33) if you Ø thought that you had already won the portwine bottle. (p, German, email) (34) I really hope that everyone arrived in Kopaonik as per travel-timetable! (a, Italian, email) A number of specific verbs were analyzed to determine how they affected the variability; any verb occurring frequently enough to allow it to be analyzed on its own was considered (think, hope, tell, say and know). The cut-off point for this was at least 24 tokens in the whole of the IFMSA data set. This is a far lower cut-off point than that used by Torres Cacoullos and Walker (2009b: 16) in their analysis of spoken Canadian English. There, verbs with more than 200 occurrences were frequent and those between 10 and 49 occurrences were deemed to be very infrequent. Except for hope, which was not examined on its own by Tagliamonte and Smith (2005), the other specific lexical verbs are the same ones considered in Tagliamonte and Smith (2005) and similar to those found most frequently in Torres Cacoullos and Walker (2009b).

Overall distribution The non-native speaker email corpus yielded 576 tokens and the native corpus 328. The breakdown of the tokens, by native language and by that or zero complementizer is provided in Table 8.1.6

Complement izers

117

Table 8.1 Overall distribution of zero use in complementizers (emails)

IFMSA French German Italian Natives

% of zero

% of that

Total N

33 26 46 34 38

67 74 54 66 62

576 197 87 292 328

The occurrence of the zero complementizer form is far lower in all four groups than previous studies on native English use, as their research had found near categorical rates of zero in certain contexts. This is not entirely surprising, however, as it had been noted in the research of both groups that speech is more likely to show the zero form than written data. The rates found in the present analysis are at a mid-point between the informal and formal texts that Elsness (1984) had considered. This further underlines how emails are a separate medium from both oral and written data and why it is crucial to use a native English control group of emailers to compare with the Swiss data. It might have been hypothesized that the non-native speakers would have much higher rates of that given their mother tongues do not have forms comparable to the English zero form and the fact that they are not explicitly taught, but this is not the case. The native group is not substantially different from the three non-native groups in terms of percentage of zero complementizer forms, being at a mid-point between the German speakers and the French and Italian speakers as can be seen from Figure 8.1. A chi-square test considering all four linguistic groups reveals that the differences between them are significant,7 so at this stage we cannot state that the four groups are identical in their complementizer use. The three non-native groups make use of both variants, however. By examining the 80 70

% of zero

60 50 40 30 20 10 0

French

Italian

Figure 8.1 Percentage of zero complementizer

English

German

118

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

distribution of the variants across the various factor groups in more detail, we may be better able to establish how similar the non-native groups are to the native speakers. Note that the German speakers use more of the zero complementizer than the native group; this may be due to interactions with the lexical verb. If the German speakers have a higher proportion of think than the other groups, for example, then this could explain the difference in zero complementizer use. Alternatively, this higher zero use could signal a type of generalized that avoidance: German speakers were less likely than other groups to use that as a relative pronoun as well. In terms of individual speaker use, some of the IFMSA members contribute a far higher number of tokens than others; what is crucial, however, is that all the speakers are variable to some extent, so there is no risk that a single member might be skewing the results of one of the language groups. In terms of speakers per individual linguistic group, there are 14 different French language speakers who have provided tokens of complementizers, 15 German language speakers and six Italian language speakers. Because, as was the case for relative pronoun selection, there are a number of factors which highly constrain the variability, an in-depth examination of these factors will allow us to gain a greater perspective into any similarities or differences between the linguistic groups.

Lexical verb The first factor considered separately was the specific lexical verb which preceded the complementizer clause. As mentioned above, some verbs (namely think, mean, know) showed far higher rates of zero complementizer than other verbs. This difference between specific verbs was explained in two ways; first of all, some verbs, such as think, were more likely to have an epistemic meaning and were more likely to use the zero form, and secondly higher frequency verbs were also more likely to show high rates of zero complementizer (Torres Cacoullos & Walker 2009b: 15). The present analysis will consider think, hope, tell, say and know, as these were the verbs which occurred most frequently in the non-native data. The other verbs which previous studies had found to be more likely to use the zero complementizer did not occur frequently enough to warrant being considered individually.

Think The two email corpora (non-native and native) examined provided almost 200 tokens of think. This represents more than 20% of the overall tokens considered in this analysis, so think is one of the most frequently occurring verbs, as was the case in previous research. The percentages of think with a zero complementizer are lower for all four groups than that found in earlier research where it was nearly categorical

Complement izers

119

Table 8.2 Distribution of complementizer forms for think

IFMSA French German Italian Natives

% of zero

% of that

Total N

52 51 68 54 90

48 49 32 46 10

161 51 22 80 39

(Table 8.2). It is, however, the lexical item with the highest rate of the zero variant in previous studies. The native speaker group is far closer to the near categorical average found in previous studies (with 90% zero) than the three non-native groups (with ranges between 50% and 70%). While it will only be through an analysis of the multivariate analysis of this feature that I will able to establish for certain whether the non-native groups are truly following different patterns, these results nevertheless signal that there are some differences between the native and non-native groups.8 The lower rates found in the non-native speakers might also demonstrate that their emails are generally more formal than natives; for the relative pronouns, only the non-native speakers had tokens of the highly formal variant whom. I will return to this point in the discussion of the feature. Despite the difference from the native group, the three non-native groups show very similar rates; there is not a statistically significant difference in their use.9

Hope There were nearly 100 tokens of hope for the non-native speakers and 25 in the native emails. Although most studies did not consider it separately from other verbs, Torres Cacoullos and Walker (2009b: 16) found that the complementizer was deleted at a rate of 89%. As in the case of think, hope demonstrates a high proportion of the zero complementizer in all four groups, with the German speakers being closest to the native percentages and the French speakers furthest away (Table 8.3). Table 8.3 Distribution of complementizer forms for hope

IFMSA French German Italian Natives

% of zero

% of that

Total N

62 56 81 60 88

38 44 19 40 12

95 27 16 52 25

120

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Here again, although all the groups vary and show percentages of zero complementizer forms above 50%, there are differences between the groups. In this instance, the German speakers are very close to the native speakers whereas the other two groups delete the complementizer about 20% less often. The low number of tokens for the German speakers makes it difficult to know how important this is. The differences between the four groups are significant but, as for think, when only the three non-native groups are considered, the difference is found not to be significant.10

Tell The next high-frequency lexeme is tell. Tagliamonte and Smith (2005: 301) had found rates of 64% of the zero form with tell in their data, while Torres Cacoullos and Walker (2009b: 16) had found 54%, both of which was considerably lower than the other verbs they examined and lower than the overall distribution (which was 80%). Directly reported speech has to be separated from indirectly reported speech with tell and was not included because in direct speech there is no complementizer (Examples 35–37). (35) He told me ‘I’m happy’ – direct speech (36) He told me that/Ø I was happy – indirect speech (37) He told me that/Ø he was happy – indirect speech While the number of tokens for tell is rather lower than for think and hope, the distributions of the various linguistic groups can still be analyzed, and once again there are considerable differences between the natives and the German speakers on one hand and the French and Italian speakers on the other (Table 8.4).11

Say Tagliamonte and Smith (2005: 301) found that the zero complementizer occurred at a rate of 85% with say in their data, while Torres Cacoullos and Walker (2009b: 16) found a rate of 73%. There are slightly over 40 tokens of it in the two email corpora. The distribution of the tokens is not ideal as the French and native English speakers provide the majority of the tokens, so the

Table 8.4 Distribution of complementizer forms for tell

IFMSA French German Italian Natives

% of zero

% of that

Total N

23 11 44 15 45

77 89 56 85 55

31 9 9 13 11

Complement izers

121

results of the German and Italian speakers (with totals of three and four tokens, respectively) cannot be considered to be truly indicative of the situation (Table 8.5). As was the case for hope and tell, say occurs at a higher rate with the zero complementizer for the German and native English speakers. The French rate (18%) is considerably lower than the native group (37%). The native rate is much lower than that which was found in speech.

Know Similarly to say, Tagliamonte and Smith (2005: 301) found that know occurred with the zero complementizer at a rate of 85%, while Torres Cacoullos and Walker (2009b: 16) find 66% deletion. There are around 50 tokens of know in the email corpora, and in this case it is the French and the German speakers that have a far lower number of tokens than the other two groups (Table 8.6). Table 8.7 presents a summary view of the percentages of zero complementizer for all the verbs studied individually, while Figure 8.2 presents these results in graph form. The rates of the different verbs alongside each other must be considered, as this will allow us to establish whether the groups share the same patterns despite having different overall distributions. Figure 8.2 also plots the results from Tagliamonte and Smith (2005: 301) and Torres Cacoullos and Walker (2009b: 16) so it can be compared to the email data.

Table 8.5 Distribution of complementizer forms for say

IFMSA French German Italian Natives

% of zero

% of that

Total N

21 18 33 25 37

79 82 67 75 63

24 17 3 4 19

Table 8.6 Distribution of complementizer forms for know

IFMSA French German Italian Natives

% of zero

% of that

Total N

43 25 40 47 42

57 75 60 53 58

28 4 5 19 24

122

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Table 8.7 Percentage of zero complementizer by verb

English French German Italian

Overall

Think

Hope

Tell

Know

Say

38 29 49 32

90 51 68 54

92 56 81 60

45 11 44 15

42 25 40 47

37 18 33 25

100 90 80 70

60 50 40 30 20

10 0

Overall

Think

Hope

Tell

Know

Say

English

French

German

Italian

Tagliamonte and Smith

Torres Cacoullos and Walker

Figure 8.2 Percentage of zero complementizer by verb

What is most striking when examining Figure 8.2 is that the four email groups show remarkably similar patterns; despite differences in percentages, the four email groups have the highest rates for the same verbs. The hierarchy they all show is hope > think as the verbs with the highest rates of zero complementizer, with tell, know and say showing lower rates. The French and Italian groups have patterns that are marginally more similar to each other than to the other two groups and the same holds for the English and German groups; nevertheless, the overall picture is that the three non-native groups have similar patterns to the native control group. Because Tagliamonte and Smith (2005) did not provide the rate of zero complementizer use with hope in their results, there is a gap in Figure 8.2. Although the rates of zero complementizer found by them are considerably higher than for any of the email groups, there are several points of similarity. Tell is the lexical verb with the lowest rate of zero complementizer in all five groups and think has one of the highest rates. Similarly to the French and Italian groups, the results of studies examining native speech show that know is used considerably more with the zero complementizer than with tell. Overall, in terms of specific verbs, these results provide us with two major findings. First of all, the native emails show a distribution similar to the findings in previous studies in terms of specific verbs. The rates of zero

Complement izers

123

complementizer are slightly lower than that which was found in spoken data, but the near categoricity of the zero form with think is replicated here. The email medium did not affect the variable rules underlying the distribution of the zero complementizer. Secondly, although the three non-native groups have lower percentages than the native speaker control group, Figure 8.2 demonstrated that their patterning is very similar to the native speakers. Their percentages fall between those of the formal and informal text categories that Elsness (1984: 521) had examined, again underlying how difficult it can be to place email in terms of written or oral registers.

Subject of matrix clause Previous studies found that the subject of the matrix clause influenced complementizer choice in that first and second person subjects were more likely to occur with a zero form than third person forms. The data in the present study will be considered in terms of a four-way distinction with singular and plural subjects considered together: first person subjects, second person subjects, third person pronoun subjects and third person NPs. This follows previous studies. The patterns found in previous studies are matched by the non-native and native emailers as is demonstrated in Table 8.8. First person subjects have the highest rates of the zero complementizer in all four groups and, except for the German speakers, these are followed by second person subjects.12 There are a number of differences in terms of the hierarchy in terms of the third person categories, however: the English and the French speakers have higher percentages of zero with NPs than with third person pronouns whereas it is the opposite for the German and Italian speakers. Figure 8.3 plots the percentages of zero by subject in a graph to make the distribution more apparent; despite some differences in terms of third person subjects, the four groups are quite similar. The non-native groups share the hierarchies of the native speakers and are varying their use of the zero complementizer according to the subject of the matrix clause in a very similar way to the native emailers. Table 8.8 Percentage of zero complementizer by subject of the matrix clause French % of zero

Total N

German % of zero

Total N

Italian % of zero

Total N

English % of zero

Total N

First person

38

119

60

60

44

192

52

143

Second person

29

7

22

9

18

22

36

75

4

49

25

8

16

44

18

56

Noun Phrase

18

22

0

10

6

34

22

54

Total

26

197

46

87

34

292

38

328

Third person pronoun

124

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

100 90 80 70 60 50 40 30 20 10 0 French

German

Italian

English

First person

Second person

Third pers. pronoun

Noun Phrase

Figure 8.3 Percentage of zero by subject

Tense Previous studies (Tagliamonte & Smith, 2005; Thompson & Mulac, 1991) found that verbs in the present tense are more likely to be used with the zero complementizer form than past tense verbs. The results for the present study are shown in Table 8.9 and Figure 8.4. Table 8.9 and Figure 8.4 show that, unlike the factors of lexical verb and the subject of the matrix clause, the groups are very different in terms of verb tense. Rather unexpectedly, the English group shows a higher proportion of zero deletion with past verbs than with present ones, which is at odds with both the non-native emailers and with the findings for native speakers in previous studies. It may be because this analysis includes tokens of I thought, whereas Tagliamonte and Smith’s 2005 study did not. This discrepancy could also be linked to a relatively low proportion of past tense forms as opposed to the present tense tokens. The multivariate analysis should be able to determine if this is the case or whether there are other factors involved in this difference. Table 8.9 Percentage of zero complementizer by verb tense French German Italian English % of zero Total N % of zero Total N % of zero Total N % of zero Total N Present 30 Past 12 No verb 0

162 34 1

49 42 0

65 19 3

36 12 17

262 24 6

38 42 13

269 43 25

Complement izers

100 90 80 70 60 50 40 30 20 10 0

French

German Present

Italian Past

125

English

No verb

Figure 8.4 Percentage of zero by verb tense

No conclusions about the tokens found in sentences without verbs can be drawn. Except for the English natives who have 25 tokens with no verb, the number of tokens is far too low to enable us to establish whether this factor favours or disfavours the zero complementizers. As mentioned earlier, ‘no verb’ will not be considered in the factor of verb tense in the multivariate analysis.

Additional elements in verb phrase Previous studies found that the presence of modal and negation forms lowered the likelihood of the zero complementizer being used. The results presented in Table 8.10 and Figure 8.5 consider this in terms of the email corpora. Table 8.10 presents the percentages and figures of modal and negation forms separately, while Figure 8.5 collapses them together as there are too few tokens otherwise. This allows us to examine the rates of zero complementizer in terms of additional elements versus no elements. Table 8.10 Percentage of zero complementizer depending on elements in VP French Total N German Total N Italian Total N English Total N % % % % Negation 80 Modal 0 Modal and negation 0 Any element 29 No element 26

5 6 3 14 183

0 33 0 20 48

0 3 2 5 82

25 10 0 12 36

4 21 1 26 266

100 8 0 16 41

4 40 1 45 283

126

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

100 90 80 70 60 50 40 30 20 10 0

French

German

Italian

Any element

English

No element

Figure 8.5 Percentage of zero deletion by elements in VP

Although the hierarchy is very similar for the German, Italian and English groups, in that they all, as predicted, have higher rates of zero complementizer in clauses without additional elements, the overall low rate of tokens containing additional elements means that it is difficult to attribute too much importance to these findings. The low number of tokens may also account for the divergent distribution of the French speakers.

Additional elements in matrix clause Tagliamonte and Smith (2005) and Torres Cacoullos and Walker (2009b) found that additional elements in the matrix clause, such as adverbials, decreased the likelihood of a zero complementizer being used. Table 8.11 and Figure 8.6 analyze the distribution in the email data. All four email groups show the predicted distribution; additional elements in the matrix clause lower the use of the zero complementizer form. Having examined the factors one by one, the next step is to consider the factors alongside one another in a multivariate analysis to determine with greater accuracy which factors have a significant effect on the rates of the two complementizer forms. Before turning to the multivariate results, Table 8.11 Percentage of zero complementizer with additional elements in the matrix clause

Something Nothing

French %

Total N

German %

Total N

Italian %

Total N

English %

Total N

0 30

23 174

30 48

10 77

8 36

24 268

17 41

48 280

Complement izers

100

127

Something

90

Nothing

80 70 60 50 40 30 20 10 0

French

German

Italian

English

Figure 8.6 Percentage of zero complementizer with additional elements in the matrix clause

however, the main findings of each factor in terms of similarity between oral and written data (by comparing my results to previous studies) and between the native and non-native groups will be briefly summarized. •

•

•

•

Overall rates. the main differences in zero complementizer use are between the email and the oral data. Whereas previous studies examining oral data had rates between 70% and 90% of zero, the email groups, native and non-native, are all around 35%. The English and the German groups have higher rates of the zero form than the French and the Italian groups, however. Although higher than the rather low rates (15% and 1%) Elsness (1984) had found in his two categories of formal texts, all four groups have lower overall rates than those he had found (52% and 58%) in two categories of informal texts (Elsness, 1984: 521). In terms of complementizers, the register of emails is more formal than oral data and informal written data, but this affects native and non-native speakers alike. Lexical verb. Despite some differences in overall distribution of the zero form, the four email groups have very similar patterns for the different verbs; hope and think are considerably higher than the other verbs. This matches the findings in Tagliamonte and Smith (2005) and Torres Cacoullos and Walker (2009b). Subject of matrix clause. The four email groups have the pattern found for oral data in Tagliamonte and Smith (2005); first person subjects have the highest rate of the zero variant, followed by second person subjects, then other subjects. Tense. The three non-native groups show the expected favouring of the zero complementizer in present tense contexts. For the native speakers,

128

•

•

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

on the other hand, there is no statistically significant difference between the two contexts, mostly likely due to low Ns. Additional elements in verb phrase. Here the order predicted was an increase of the zero complementizer in cases where there was no additional element in the VP; all but the French speakers showed the expected order. Additional elements in matrix clause. As for the previous factor group, the absence of additional elements was found to favour the zero complementizer. This was found to be the case for all four email groups.

Examining the whole range of factors influencing the use of complementizers, the overwhelming conclusion one can draw is that the non-native groups are very similar to the native emailers. The multivariate analysis will now be able to verify if it is indeed the case that the non-native groups have managed to acquire the native variable patterns of this feature. Before turning to the multivariate analysis, the possibility that learning-related processes affected the results needs to be considered. As mentioned previously, French and Italian do not have a form comparable to the English zero complementizer, whereas both High German and Swiss German do. Is this noticeable in the results? The German speakers have higher overall rates of zero complementizer than the two other non-native groups. This may signal that the zero complementizer form in German helped this group acquire it in English. Before ascribing too much importance to this fact, however, remember that the German speakers had rates closer to native speakers than the other groups for the future variable as well. The higher rates found in the German group could be due to a higher level of English competency rather than structural similarities. Furthermore, more significant than mere percentages are the patterns which are matched not only by the German speakers, but by the Italian and French speakers as well. These patterns are those of English, not of German, so the German speakers would have had to acquire them, just as the French and Italian speakers did.

8.6 Multivariate Analysis Four separate multivariate analyses considering the probability of the zero complementizer being used were run; each analysis examining a single linguistic group. The four analyses are presented in a single table (Table 8.12) to permit an easier comparison. The five separate factor groups of lexical verb, subject of matrix clause, tense, additional elements in VP, and additional elements in matrix clause were considered. Because of differences in terms of overall distribution in the

0.36 FW

%

Total N 328

Lexical verb Think 90 0.94 Hope 88 0.90 Know 42 0.58 Say 37 0.55 Tell 45 0.64 Other 21 0.30 Range 64 Subject of matrix clause First [0.52] 52 Second [0.60] 36 Other [0.41] 20 Range Additional elements in VP Some 16 0.27 None 41 0.54 Range 27 Additional elements in matrix clause Some 17 0.26 None 41 0.55 Range 29 Tense Present [0.50] 38 Past [0.42] 42 Range

Input

English % 38

10

[0.42]

n/a n/a

n/a n/a

[0.51] [0.47]

45 283

48 280

269 43

30 12

0 30

29 26

37

[0.55]

143 75 110

51 56 [25 18 11 7

%

0.81 0.84 n/a 0.47 0.34 0.23 61

0.20 FW

Total N 197

39 25 24 19 11 210

N

French % 26

162 34

23 174

14 183

79

118

51 27 4] 17 9 89

N

[0.49] [0.53]

0.17 0.55 38

n/a n/a

0.20 45

0.65

0.70 0.81 n/a n/a 0.65 0.19 62

0.42 FW

German % 46

49 42

30 48

20 48

15

60

68 81 [40 [33 44 16

%

Total N 87

Table 8.12 Four independent multivariate analyses into the probability of zero complementizer

65 19

10 77

5 82

27

60

22 16 5] 3] 9 32

N

[0.52] [0.27]

0.18 0.53 36

[0.34] [0.52]

[0.53] [0.29] [0.49]

0.75 0.78 0.68 n/a 0.31 0.22 56

0.27 FW

Italian % 34

36 12

8 36

12 36

44 18 12

54 60 47 [25 15 10

%

Total N 292

262 24

24 268

26 266

192 22 78

80 52 19 4] 13 123

N

Complement izers 129

130

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

different linguistic groups, some factors were excluded from one or more of the analyses: • •

•

•

•

For tense, the tokens which been coded as having no tense were excluded in all the analyses as there were generally too few of them to allow for an accurate analysis. For subject of the matrix clause, all third person subjects, NP and pronoun, were merged together for the English and the Italian analyses, creating a three-way division. For the French and the German data, a two-way distinction was made, considering first person subjects against all other subjects due to the low number of tokens of non-first person subjects. The factor of additional elements in the matrix clause was not run for the French group as all the tokens with additional elements used the that variant. Goldvarb cannot run a multivariate analysis on factor groups where one of variants is used 100%. The factor group examining additional elements in the VP was not considered for the German or the French speakers are there were too few tokens of complementizers with additional elements to make an analysis conclusive. For lexical verb, verbs with fewer than nine tokens were excluded from the individual analysis as their inclusion could have skewed the data. For the French speakers know was excluded, for the Italian speakers say was excluded and for the German speakers both know and say were excluded.

This will make the comparison across varieties slightly more complicated, but factor groups whose distribution is deemed to be linked solely to token numbers must be removed as this can obscure the actual patterns. All four groups disfavour the zero complementizer, which is rather different from what had been found in Tagliamonte and Smith (2005). This is most likely due to the fact that the data in the present analysis come from emails rather than speech. Lexical verb was selected as being significant in all four variable rule analyses; moreover, all four linguistic groups show similarly high ranges for this factor. In terms of the hierarchy of this factor, there are a few differences but the overall impression is that the groups pattern in a very similar fashion. Think and hope are the two most favoured forms overall and the category of ‘other verbs’ is the most disfavoured factor in all four groups. There are some differences in terms of the hierarchy of the other verbs studied, especially given that some verbs were excluded from some analyses, but their weights generally are between think/hope and ‘other’. Table 8.13 presents the differences in hierarchy between the various linguistic groups in terms of lexical verb.13 The verbs given in parentheses are

Complement izers

131

Table 8.13 Hierarchy of lexical verbs for zero complementizers English French German Italian

Think Hope Hope Hope

> hope > think > think > think

> tell (> know) > tell > know

> know > say (> know ) (> say)

> say > tell (> say) > tell

> other > other > other > other

those which were not included in the multivariate analysis; their position in the table is related to their percentage of zero complementizer. The subject of the matrix clause was found to be significant only for the German speakers. However, as predicted by Tagliamonte and Smith, first person subjects are favoured in all four groups. First person has the highest factor weight in the three non-native groups, but second person subjects have a higher rate for the native speakers. This is somewhat unexpected as it does not match Tagliamonte and Smith’s results. Additional elements in the VP was found to be significant only for the native speaker group. The hierarchy matches the results that had been found in previous studies of complementizers, in that additional elements disfavour the use of the zero complementizer. Moreover, Italian speakers, the other group this factor was run for, also share the expected hierarchy. Not only was the factor additional elements in the matrix clause found to be significant for the three groups where it was included in the variable rule analysis, but the hierarchy of factors was found to be the same in them as well. The distribution is what was predicted by Tagliamonte and Smith (2005): elements in the matrix clause disfavour the zero complementizer. Although this factor could not be examined for the French speakers, this was due to the fact that all of the tokens with additional elements in the matrix clause appeared with a that complementizer, so they can be said to share the same hierarchy. Tense was not found to be significant for any of the groups, in terms of hierarchy; however, the English, French and Italian groups show the hierarchy (present > past) found by Tagliamonte and Smith (2005). This tendency is reversed in the German speakers, where past forms have a higher factor weight than present forms. Note that the difference between these two factors is very small and that the lower overall number of tokens for the German speakers might have influenced this result.

8.7 Discussion There are a number of differences between the four groups in terms of the multivariate analysis results. Although lexical verb has the highest range in all groups, there is a switch in terms of the hierarchy of think and hope between the native and non-native groups. For the subject of the matrix

132

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

clause, the native group shows a higher factor rate for second person subject than first person ones, while this tendency is reversed for the non-natives and matches what had been found in previous studies. Finally, for tense, the German speakers do not share the hierarchy of the other groups. While these differences are important, the overall picture is one of similarity. The non-natives, for the most part, match native speaker patterns (either those of the email control group or those considered in previous studies). Think, tell, know and say have higher rates of zero complementizer than other verbs, first person subjects are more likely to be used with a zero complementizer than other subjects, and elements, either in the VP or the matrix clause, inhibit the use of zero. This demonstrates that the non-native speakers have acquired many of the constraints which operate on English complementizer patterns. The (Swiss) German speakers have a zero form used in a similar way to the English zero complementizer form in their native language and they are the non-native group with the highest rates of zero complementizer. They are not the only group to show variability in the feature, however, as both the Italian and French groups also use the zero complementizer form. While the similarity between German and English might have benefited the German speakers in some ways, the fact that they and the other two nonnative groups share native speakers’ hierarchies of constraints and ranges is due to more than surface similarity. Although we do not know what constraints operate on German complementizers, it is unlikely that they are the same constraints as in English. The German speakers (as well as the French and Italian speakers) are applying English variable rules for their use of the zero complementizer form. The rates of zero complementizer are lower across the board than in previous research. The main explanation for this is that most had examined oral data, whereas this analysis considered emails.

8.8 Conclusion This chapter has shown that despite the constraints underlying that and the zero complementizers form selection not being explicitly taught to nonnative Swiss speakers of English, these speakers have nevertheless acquired the variable rules of native speakers. The multivariate analyses of the various linguistic groups have shown that the hierarchies and ranges of the nonnative groups are closely matched to those of the native control group. These patterns also match the constraints that have been uncovered in previous analyses of English zero complementizers. Despite there being no similar zero complementizer form in two of the three source languages, the French, German and Italian speakers have been able to integrate the variable rules of English complementizer distribution.

Complement izers

133

This underlines the fact that there are a number of underlying syntactic distribution patterns which non-native speakers can acquire without explicit teaching.

Notes (1) A reduced version of this chapter appeared as Durham, M. (2011) I think (that) something’s missing: Complementizer deletion in non-native e-mails. Studies in Second Language Learning and Teaching 1 (3), 421–445. (2) Note, however, that what was included in the analysis did vary from study to study. Kolbe chose to examine solely the verbs think, say, know and see which goes some way towards explaining why her rates are higher than the other studies. Tokens such as I think, you know and I mean which categorically selected the zero complementizer were excluded in Tagliamonte and Smith’s results. (3) Much of the literature read by students of English would have contained instances of both that and zero complementizer. (4) Rather differently from Tagliamonte and Smith (2005), there were no cases of parentheticals which were in a position where the that form was not a possible option (i.e. ‘She’s very nice, I think’). (5) NORE = National Officer for Research Exchanges. (6) In this table and the other tables in this chapter, the native data come solely from my own email corpus. (7) df = 3; χ2 = 12.1882144658936; p < 0.01. (8) A chi-square calculation of the four groups confirms this is a significant result: χ2 = 18.0340485792825; p < 0.001. (9) A chi-square calculation of the three non-native groups is found to be not significant: χ2 = 1.9266286702428; p < 1. (10) All four groups: χ2 = 9.3976; p < 0.025. Three non-native groups: χ2 = 3.1209; p < 1. (11) The low number of tokens per cell means that it is not possible to test whether these figures are significant. (12) Although Tagliamonte and Smith (2005) had found a considerable number of interaction terms of first person subjects and think, this analysis did not uncover the same categorical distribution of I think with the zero form for any of the linguistic groups. (13) Verbs in bold are those which the multivariate analysis found to favour the use of the zero complementizer.

9

Lexical Variation: Also, As Well and Too

9.1 Introduction Adverbial placement is often seen as a difficulty for learners of English (Osborne, 2008; White, 1991): not only can specific adverbial type affect placement within a clause, but in many cases individual adverbs have their own sets of restrictions. The extreme range of variability is not only a challenge for non-native speakers but also for researchers, as this makes it difficult to consider adverbials as a group and instead it is often necessary to consider only a few at a time. This latter method is adopted here and the focus will be on three additive adverbials: also, as well and too (Examples 1–3), which are synonymous and which Fjelkestam-Nilsson (1983) has studied previously, finding that their distribution is conditioned by style and register as well as clause structure in native English. (1) maybe there’s also some Porto wine. (c, Italian, email) (2) I ordered as well some shiny days to make your stay in Zurich as comfortable as possible. (h, German, email) (3) it was a situation that I shared, too. (f, French, email) The analysis in this chapter will examine the non-native speakers’ use of these forms in order to establish whether they have rates of variation similar to native speakers. Because the variants are interchangeable and semantically equivalent there is a possibility that the non-native speakers will select just one of the variants and not use the other two, as Chapter 6 showed occurring for the future tense. This feature is rather different from the other target-based variation features studied thus far because the difference between the variants is linked purely to lexis rather than to morphosyntax. The analysis will also consider aspects of learning-related variation (LRV) in the non-native speakers as the 134

Le x ical Var iat ion : Al so, As Well and Too

135

variants are sometimes found in positions that are not acceptable to native speakers (Examples 4–6). (4) he looked at as well the budget. (h, German, email) (5) She agrees, but she can’t take also this job. (c, Italian, email) (6) they have to go also for two months. (u, French, email)

9.2 Also, As Well and Too Also, as well and too, the three variants, are additive adverbials. Additive adverbials belong to the focusing class of adjuncts in Quirk et al.’s (1985: 609) categorization of adverbials. This category also includes either, even, neither, nor and in addition. I have decided to focus only on also, as well and too as they are the most semantically similar and thus function as lexical variants. Furthermore, Fjelkestam-Nilsson (1983) concentrated primarily on these three variants in her research on additive adverbials in British and American English corpora. A surface analysis of the Swiss data revealed that French, German and Italian speakers use the whole range of variants available, as the data revealed examples of also, as well and too in the output of the three linguistic groups (Examples 7–15). Also (7) Actually, this message is also a big HELP request. (j, French, email) (8) ‘c’ also created some wonderful posters. (p, German, email) (9) The fifth of the IFMSA Committees is also existing in Switzerland. (c, Italian, email) As well (10) she’d really like to visit those three cities as well. (V, Italian, email) (11) I think the grammar is changing as well now. (S, French, interview) (12) I will try to get a video player as well for the meeting. (h, German, email) Too (13) it was a situation that I shared, too. (f, French, email) (14) we discussed that we could do it in CH , too and compare the results. (h, German, email) (15) and I hope it was interesting for you, too. (b, Italian, email) A first aim of the analysis, then, will be to establish whether the three variants are used at rates similar to those of the native speakers. By comparing how also, as well and too are used by the non-native and the native groups, this section will be able to uncover whether, similarly to the future, one of the forms is used instead of the other two variants.

136

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Table 9.1 Possible positions of also, as well and tooa Sentence

Positions

(Also) John (also) created some wonderful posters (also) John (, too,) created some wonderful posters (too) John (, as well,) created some wonderful posters (as well)

Initial, mid, final Final Final

Notes: aEven in the cases when too and as well seem to be placed in a more central position within the sentence, i.e. ‘John, too, created some wonderful posters’, they were still considered to be in final position and coded accordingly; this has to do with the fact that they are modifying only the NP in the clause and not the whole clause and, thus, do indeed come after it.

Although the adverbials are semantically similar, there is a difference in terms of the positions where they are acceptable within the clause. While also can be used in a wide range of positions (Jacobson, 1964: 226; Jacobson, 1975), the other two adverbials tend to be found only in clause final position (Jacobson, 1964: 353) (Table 9.1).1 Comparing the non-native placement of the variants to that of the natives will allow us to analyze the learning-related variability present. Moreover, in the case of also which can grammatically be placed in several positions, the distribution of placement between the native and Swiss groups can be compared. This will reveal how frequently the Swiss speakers use the variants in unacceptable positions, but also show which variants and positions are most likely to be a source of difficulty.

9.3 Additive Adverbials in the Source Languages In terms of the source languages, French, German and Italian have only one main form equivalent to the three English adverbials. That is not to imply that the three non-native languages do not have synonyms for the main term, but that in each case there is one term which is the one predominantly used. This is a quite different situation from English, where all three forms have been found to be used frequently both in speech and in writing (Fjelkestam-Nilsson, 1983). The main form in French is aussi, with également used considerably less frequently; Italian has anche as the main form, with pure used alongside it (as well as the very literary form altresi); German has auch and a much less frequently used ebenso and Swiss German has au or öi, depending on the dialect. One point that it is important to consider is that the German word also means therefore and this may affect these speakers’ use of the homophonous English variant. Table 9.2 presents the differences between the Swiss languages and English schematically. When selecting additive adverbials in English, the Swiss speakers will have three variants to choose from whereas there is only

Le x ical Var iat ion : Al so, As Well and Too

137

Table 9.2 Also, as well and too: equivalents in the three Swiss languages English

French

German

Italian

also

as well

aussi

auch

anche

too

one main form in their native languages. Consequently, it may be that they use one form more than the other two and more than native speakers would. Although the three variants are taught in Swiss schools, their distribution is not. Moreover, while the positioning of the three forms is different, this is never explicitly stated in English language teaching, despite the fact that Swiss students are expected to place the variants appropriately.

9.4 Previous Work on Additive Adverbials Although to date there is no research specifically on non-native variation in the usage of the three adverbials in English, their distribution in native English has been studied quite extensively. Brita Fjelkestam-Nilsson’s 1983 study is based on results obtained in three corpora which were first made available in the late 1970s. One of the corpora consists of American written texts (Brown University Corpus, henceforth BUC; Kucera & Francis, 1967), one of British written texts (Lancaster-Oslo/Bergen Corpus, henceforth LOB; Johansson, 1978), and one of oral British data (London Lund Corpus, henceforth LLC; Svartvik & Quirk, 1980). For her study, Fjelkestam-Nilsson compared the distribution of the three additive adverbials (only also and as well were considered for the oral data) in terms of text type and origin of the speakers. Her results will be very useful for the present study, even if there is a considerable time lapse, as it will help determine whether the non-native groups are similar to native speakers. The main results of Fjelkestam-Nilsson’s research will be presented alongside those for the Swiss and native email data, but it should be noted that the main differences that she found in the choice of also and too were stylistic: also was favoured in ‘informative’ texts more than in ‘imaginative’ ones, although it was used more often than too in all text types (Fjelkestam-Nilsson, 1983: 16).2

138

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

There were also some differences between American and British usage, and between writing and speech, but these were not as significant as the difference between types of text. She also found that the adverbial as well was at a mid-point between the two other adverbials – less formal than also but not as informal as too.

9.5 Analysis and Results Data were extracted from both the oral and email components of the IFMSA corpus and from the native speaker email corpus. Instances where the position of the adverbial was not felt to be native-like were flagged so they could then be considered separately. Obviously, cases of too as an intensifier rather than as an additive adjunct were excluded from the analysis (Example 16), and similarly tokens of also where it functioned as a subjunct rather than an adjunct were excluded (Example 17).3 (16) I cannot do too many things at the same time. (o, German, email) (17) Also don’t forget about the town club on Friday evening at 7PM. (native English email) A total of 311 tokens of also, as well and too were extracted from the nonnative corpus (Table 9.3). A chi-square test revealed that the differences between the oral and written segments of the corpus were not significant,4 so they are combined in the rest of the analysis. Table 9.4 considers the overall distribution of also, as well and too in terms of the three non-native language groups. In terms of the overall distribution, the Italian speaker tokens make up roughly two-thirds (205 out of 311) of the total. The number of tokens for the other two groups, although small, is sufficient to gain some idea of their distribution of the variants. The main variant used by all three groups is also, although the German speakers use it considerably less than the other two (at a rate of 51% opposed to 81% and 92% for the French and Italian speakers, respectively). The French and Italian speakers have fairly low rates of the other two variants, whereas the German speakers use as well a full 40% of the time. The differences between the three

Table 9.3 Non-native distribution of the three additive adverbials for source type (all groups combined)

Emails Oral data Total

% of also

% of as well

% of too

Total N

82 90 84

11 7 10

7 3 6

221 90 311

Le x ical Var iat ion : Al so, As Well and Too

139

Table 9.4 Non-native distribution of the three additive adverbials by speaker group

German French Italian Total

% of also

% of as well

% of too

Total N

51 81 92 84

40 6 4 10

9 13 4 6

43 63 205a 311

Notes: aThe difference between the three language groups is significant with p being less than or equal to 0.001.

groups are significantly different, which underlines that the German speakers are not selecting the three variants in the same way as the other two groups. The native email corpus provided 47 tokens of the three adverbials. Table 9.5 considers the results for the native email control group in order to establish whether the non-native patterns are typical of native speakers. The results for the native emails show that, similarly to the three nonnative groups, also is the form used most frequently. Table 9.6, which gives the figures from Fjelkstam-Nilsson’s study for both or and written data, will allow us to establish whether the email results are comparable to previous findings. Although we do not have figures of too for the Fjelkstam-Nilsson’s LLC data, the rest of the results look very similar to what had been uncovered for the native and non-native groups; namely, that also is the main form. Figure 9.1 plots the results for the three non-native groups alongside the findings of the native email corpus and of Fjelkstam-Nilsson’s research (English oral and English written in Figure 9.1). From this it is clear that: •

also is the main variant for every single group and is selected more than 50% of the time, although the German speakers use it less than the other groups;

Table 9.5 Overall distribution of additive adverbials (native email corpus)

Emails

% of also

% of as well

% of too

Total N

68

17

15

47

Table 9.6 Overall distribution of additive adverbials in Fjelkstam-Nilsson’s results (1983)

Written data (BrEng and AmEng) ( LOB and BUC) Oral data (LLC)

% of also

% of as well

% of too

Total N

73

24

4

2392

65

35

N/A

367

140

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

100 90 80 70 60 50 40 30 20 10 0

Also

As well

Too

German

French

Italian

English e-mail

English written

English oral

Figure 9.1 Overall distribution of additive adverbials by linguistic group

• • •

the lower rate of also found in the German group may be linked to a homophonous word in German with a different meaning; as well is the next most favoured form for the native speakers and the German speakers; too is the second most favoured form for the French speakers and the Italian speakers use as well and too equally.

Although the French and Italian speakers do not use the as well and too variants as much as the native speakers, they use them frequently enough to establish that all three adverbials are part of their linguistic repertoire. The German speakers use also far less than the other groups and as well far more; this is most likely in part due to the presence of the lexical item also in German, but because Fjelkstam-Nilsson had found higher rates of as well in her oral corpus, it may be that the higher rate of as well in the German output reveals a closer similarity to native norms than the other two nonnative groups. In order to establish how similar the non-native groups are to native speakers, the learning-related variability aspects of the feature need to be considered. First, in terms of how frequently the non-native speakers use the variants in a way that is ungrammatical to native speakers; secondly, if there are non-native tokens, what is triggering them?

Learning-related variation An analysis of the non-native usage of the additive adverbials in the nonnative corpora revealed two different types of problems involving also, as well and too. As well as tokens of the adverbials which were deemed to be positioned in an ungrammatical manner (Examples 18–19), there were also cases

Le x ical Var iat ion : Al so, As Well and Too

141

of tokens where one of the variants was used with a meaning which would be considered ungrammatical to a native speaker (Example 20). Of the two types of problems, the one associated with problems of placement was more common than the second type. (18) Also return date will be problem. (p, German, email) (19) It is as well an occasion to thank ‘a’ another time. (b, Italian, email) (20) Another thing: I can’t come to the meeting also on the 23–24 feb.5 (c, Italian, email) Because the type which is associated with lexical choice was less common, it will be dealt with first, before focusing on the findings in terms of problems in position placement. The analysis found only seven tokens of problems linked to lexical choice but it was of two different types: five tokens where also was used in place of either or neither (Example 20 above), and two tokens where as well is used in place of as well as (Example 21). These tokens have been removed from the subsequent analysis. The low number of tokens of this type suggests that this aspect did not create many problems for the non-native speakers. (21) a link between the international world and the local world and as well to the Swiss medical students’ association. (b, Italian, email) In terms of the additive adverbials, there were problems in placement for only two of the variants. Too was always placed at the end of the clause so presented no difficulty for the non-native speakers, but as well was used by the Swiss speakers in other positions than final position. Specifically, 10 out of the total 29 tokens of as well for all three groups were placed in a midposition (as in Example 19 above) and one was placed in initial position.6 Also, which has a far greater range of acceptable positions, was nevertheless occasionally placed in such a way as to make it sound unnatural to a native speaker, occurring either in an inappropriate medial position (Example 22), or clause finally, in this case generally alongside another adverbial in the inappropriate order (Example 23). (22) I hope that also the other folks out there will do the same. (b, Italian, email) (23) and ZH can take people also for two months. (u, French, email) In all, 18% (55 out of 311) of the tokens were judged to be non-native in terms of positioning. Table 9.7 provides the breakdown of deviant tokens by speaker group. The overall rate of problematic tokens for the non-native speakers is fairly high, although the numbers are low. This demonstrates that the non-native

142

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

Table 9.7 Problematic positioning of additive adverbials

Italian French German Total

Also N

As well N

Too N

Percentage non-native

Total N (non-native)

34 8 3 45

2 0 8 10

0 0 0 0

18 13 26 18

36 8 11 55

groups are using the additive adverbials in positions unacceptable to native speakers despite using the three variants at rates similar to those found in English writing. The German speakers have the highest rate of non-native tokens at 26%. In terms of the specific variants placed in a non-grammatical fashion, also accounts for 80% of the non-native tokens. There are differences between the non-native groups, however, as the German speakers are far more likely to place as well in a position where it is ungrammatical to native speakers. Not only are the non-native speakers using the variants differently from native speakers, but the German group does not share the patterns of the other two Swiss groups. This may be due to their overall higher usage rate of as well; by using it more frequently than the French and Italian speakers, there are more opportunities for them to use it in an ungrammatical position.

Position of also As the only variant which can be found in several clause positions, also will now be considered on its own, in order to determine whether the non-native speakers follow the distribution patterns of native speakers. The positioning of adverbials is in large part linked to focus, so it may be that differences between groups are tied to that; however, because the type of messages were similar across the groups, this will still give us an idea whether the groups share similar patterns. The ungrammatical tokens of also have been removed from consideration, so that only the tokens which have been judged native-like are compared to the native speakers’ output. Table 9.8 and Figure 9.2 provide the results for the different linguistic groups. The native and non-native rates are very similar in terms of positioning; the mid-position is the most frequently selected position (around 75%), the initial position is the next most frequent position (around 20%) and the final position is very rarely used. There is no statistically significant difference between the four groups,7 so it seems to be the case that they share the same distribution. Although there are a number of tokens in ungrammatical position in the non-native data, the overall distribution of positions is otherwise very similar in the four groups.

Le x ical Var iat ion : Al so, As Well and Too

143

Table 9.8 Overall distribution of also by clause position

Italian French German Non-native speakers combined Native speakers (email corpus)

% of Initial

% of Mid

% of Final

Total N

23 16 22 20 22

75 78 78 77 75

2 6 0 3 3

152 42 18 212 32

90 80 70

60 50 40 30 20 10 0 Initial

Italian

Mid

French

Final

German

English

Figure 9.2 Distribution of also by sentence position

9.6 Discussion The results have revealed a number of contradictory points, so before discussing the reasons why this might be the case this section will briefly summarize the various findings. • • • • • •

also is the main form used in all the linguistic groups; as well is used far more frequently by the German speakers than the other groups, however; the non-native speakers all show high rates of LRV as their positioning of the adverbials does not follow native norms; the French and the Italian speakers deviate from native norms mainly in their positioning of also, whereas the German speakers deviate in their positioning of as well; Excluding the ungrammatical tokens, the three Swiss groups are very similar in terms of their positioning of also, with the mid-position as the overwhelmingly favoured position; this matches what was found for the native emailers.

144

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

In some ways, this variable has not been acquired in a native-like manner by the Swiss groups, but at the same time there are also a number of factors which reveal similarities between them and the native group. Ultimately, the aspects which differentiate the non-native speakers from the native speakers outweigh those which are similar. Although all three groups have similar problems, it would be misleading, nevertheless, to suppose that these similarities are due to ELF processes rather to the inherent difficulties of the features. The fact that the German speakers show considerably different patterns from the other Swiss groups underlines this. Interestingly, of the features examined in this book, it is the single feature where the German speakers are further away from the native speakers than the French and Italian groups.

9.7 Conclusion The analysis of the three additive adverbials has highlighted the fact that a single feature may provide us with contradictory findings as to whether it has been acquired natively or not. Some aspects of a feature may be used in a native-like manner while other aspects remain very different. The nonnative speakers do use all three forms, but not at the rates found in native speakers. Once the problematic sentences have been removed, the non-native speakers show that their positional distribution of also is very similar to native speakers. In terms of the additive adverbials, although one variant has not been unduly favoured over the others, as was the case for the future tenses, the native patterns cannot be said to have been fully acquired because there is still a high rate (18%) of tokens placed in a position not acceptable to native speakers.

Notes (1) Jacobson presents his results in terms of the constituent referred to as well as position, but the positions in Table 9.1 can be constructed from his figures. (2) The text types are those given in the corpora Fjelkestam-Nilsson used. (3) Almost all of the tokens of also as a subjunct were found in the native data. (4) Chi-square = 3.26124765559715; p is less than or equal to 0.20. The distribution is not significant. (5) The context makes it clear that this speaker meant either and not also. (6) Some native speakers of Canadian English do this as well. (7) Chi-square = 2.22848503245692; p is less than or equal to 1. The distribution is not significant.

10 Discussion

10.1 General Considerations The analyses provide an insight into the acquisition of sociolinguistic competence that goes beyond previous findings. Although some features showed no evidence of native variation patterns in the non-native groups, and could be flagged as difficult to acquire, other features show native-like patterning in the French, German and Italian speakers. These contrasting findings can help us understand which types of feature non-native speakers are most likely to learn to use natively. Unlike much previous research on the acquisition of sociolinguistic competence, this study did not specifically focus on stylistically constrained features (although style was often one of the factors that affected variation) and I will examine how this broadens an understanding of the acquisition of sociolinguistic competence. Furthermore, by examining the variation patterns of three non-native linguistic groups at the same time, I have been able to establish how nonnative speakers deal with native variability, particularly in terms of the relative effect of the source language on English variability, in a way that is not possible when examining a single non-native group. For example, the results indicated that native-like patterns were not necessarily linked to similar structures in the source language: native English complementizer patterns are found in French, German and Italian speakers, although only German speakers have a zero complementizer form comparable to English. The nonnative speakers follow the English patterns, not because they have similar ones in their own languages but because they are sensitive to the English patterns and are able to replicate them. The study of non-native acquisition of inherent variation can elucidate broader processes of SLA, as it separates the variable features which are likely to be acquired with ease from those which are not. The various analyses revealed that features that are variable in the target language can help understand aspects that cannot be discovered by examining only features linked to learning processes. For example, the results found that not only do nonnative speakers make relatively few errors in selecting relative pronouns, but 145

146

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

also that they are able to acquire the less tangible aspects of relative pronoun selection. This provides insight into how they acquire deeper structures within language and is particularly useful because the rates of learningrelated variability appear to affect how non-native speakers deal with targetbased variation aspects. Another finding is that, although there are cases where the non-native speakers have not acquired a feature in a native-like manner, there is far less directly ascribable transfer than might have been expected. Although the structures of the source languages affected the ease or difficulty of acquisition in some cases, there is not much evidence of specific constructions being transferred from the source language: the periphrastic future form in French did not help the French speakers use the comparable English variant and their rate of going to was far lower than native English speakers. Additionally, although the periphrastic form is stylistically constrained in French in a way similar to English, this was not transferred, as the French speakers’ rate of going to did not increase in the interviews or the meeting from their emails. I will first review the main results of each feature in order to discuss the overall findings more clearly (in Section 10.2) and then in Section 10.3 offer suggestions as to why some features are learned while others are not. Section 10.4 will consider the implications of this research for the wider understanding of the acquisition of sociolinguistic competence and of how ELF affects the acquisition of sociolinguistic competence.

10.2 Features In order to establish to what extent sociolinguistic competence has been acquired, the three categories of variability presented in Chapter 5 must be examined. Features where non-native speakers had the same variable rules, hierarchies and constraints as native speakers have been acquired in a nativelike manner (case A); those where the non-native groups’ variability was different from native speakers, but in which the three Swiss groups shared the same variable rules, were examples of focused variables (case B); and finally, features which were shown to have entirely different variation patterns in the four groups were judged to be variables in which the patterning was influenced solely by the native language (case C). Table 10.1 (repeated from Chapter 5) presents this in schematic form. Four distinct morphosyntactic and lexical variables were studied in this book and all three of the above-mentioned cases were found in the results. The main findings for each feature are repeated briefly below.

Future tense – will versus going to Although non-native speakers had been taught that will is used for prediction and going to for the near future, they did not demonstrate patterns

Discussion

147

Table 10.1 Potential cases of variation Case Case A Case B Case C

Patterns found French, German and Italian present similar variation patterns to native speakers French, German and Italian vary unlike native speakers but like each other French, German and Italian vary unlike native speakers and unlike each other

Outcome Sociolinguistic competence and variation acquired Focused form (or common problems of acquisition) Variation influenced by native language only

similar to those of native English speakers. For all three groups, will was selected almost categorically rather than going to. The near categorical use of will made it impossible to examine many of the factors which previous native-speaker studies had found to affect the variability. However, the groups showed similar instances of learning-related variation (LRV): despite the low use of the going to variant, the analysis also revealed that 8% of the will tokens were used in a way that could be judged to be non-native. Moreover, the groups showed similar rates of contraction of will and going to to ‘ll and gonna. Because of the similarity between the nonnative groups, this feature was judged to be a possible case of focusing. Case B

French, German and Italian vary unlike native speakers but like each other

Focused form (or common problems of acquisition)

Relative pronouns The analysis of relative pronouns revealed two main findings. First of all, there were very few cases of learning-related variability in the data, under 2%. Secondly, variable rule analyses revealed that variability in the nonnative groups was conditioned in the same way as the native group. Like native speakers, non-native relative pronoun choice was found to be linked to syntax, animacy, adjacency and definiteness. Moreover, the constraint ranking of the factor groups generally matched that of native speakers, suggesting that the non-native speakers shared the same rules as the native speakers, making it an example of a feature where the variation had been acquired in a native-like manner. Case A

French, German and Italian present similar variation patterns to native speakers

Variation acquired

Complementizers Like the relative pronouns, the non-native speakers very closely matched the variable rules of the native speakers for complementizers. The non-native speaker patterns were found to be linked to lexical verb, grammatical person,

148

The Acquisit ion of Soc iolinguist ic Competence in a Lingua Franca Conte x t

tense, and the presence of additional elements in the verb phrase and matrix clause. Again, their constraint rankings and ranges corresponded to those of native speakers for the most part. The similarity between the native and non-native speakers is noteworthy given that the zero and the that variant are acceptable in all situations and that the variability is not overtly taught. This feature was judged to be another case where non-native speakers had been able to acquire inherent variability in the target language. Case A

French, German and Italian present similar variation patterns to native speakers

Variation acquired

Also, as well and too The analysis of the three additive adverbials showed that the non-native speakers had a number of patterns in common with the native group; also was the form selected most often in all groups. However, this feature showed a high rate of LRV: the non-native speakers were not always able to position also (and as well for the German speakers) in a native-like manner. Overall, the differences between the non-native speakers and the native group and between the groups themselves were such that the feature was judged to be involved in variation affected by the native language only. Case C

French, German and Italian vary unlike native speakers and unlike each other

Variation influenced by native language only

The additive adverbials fall into case C, the future falls into case B, whereas relative pronouns and complementizers fall into case A. To understand why some features have been acquired while others have not, and to predict the possible outcome of other features, it is necessary to examine in what ways the features differ and then interpret what consequences this may have had on the variability.

10.3 Interpretations The features analyzed can be differentiated in various ways, but the ways which offer the best chance of explaining the divergent results, and which may suggest the outcome of other features, have to do with feature type, frequency of occurrence, rate of LRV and whether the feature is overtly taught.

Feature type The analysis of lexical variation (the distribution of also, as well and too) showed the most distinct patterns between linguistic groups. It was not used

Discussion

149

in a native-like manner by the non-native groups and the three Swiss groups were very different from each other. On the other hand, the features in which the variants were syntactically conditioned, the relative pronouns and the complementizer forms, showed patterns very close to those of native speakers in all three non-native groups, while the future tense use, which is constrained both lexically and syntactically, falls at an intermediate point between the other two types. Although it was not acquired natively, it was found to be very similar across non-native speaker groups. Based on the features considered here, lexical variation is more difficult for non-native speakers to acquire than features of morphosyntactic variation. However, as I will discuss below, this aspect of feature type correlates with frequency. Feature type can also be used to account for the degree to which a variable is affected by style: some of the features demonstrate more marked variation with respect to formality than others. Earlier studies into the acquisition of sociolinguistic competence overwhelmingly found that informal forms were acquired later than the formal variants. In fact, many of the studies on non-native French speakers found that they had substantial difficulties acquiring native variable patterns (Mougeon et al., 2010; Regan et al., 2009): most of the features in the earlier studies were of variables in which one of the variants was more informal (for example, the rates of ne deletion by students of French in Regan, 1996). These studies examined the variation in terms of whether the learners matched native speakers’ knowledge of when and how frequently to use the more vernacular variants and found the non-native speakers did not match the native rates in informal contexts. In this book, only some of the features selected for analysis here had genuinely formal and informal variants and while the variation was sociolinguistically constrained, it was often in terms of internal, rather than external, factors. The difference of feature type between these earlier studies and the present research explains why more features here demonstrate native-like sociolinguistic competence. The features selected here, however, tend to be variable primarily in terms of internal, not external factors. In complementizer use, for example, the main factors conditioning variability are factors such as lexical verbs and grammatical person, not gender or age. Overall, this suggests that internal linguistic constraints may be easier to acquire than external, social constraints. Research on first language acquisition of sociolinguistic competence backs up this conclusion as well: Smith et al. (2013) found that very young children (